Manappuram Finance · Gold Loan

Where Gold Loan enquiries leak, and how to win them back

A map of every place a visitor gives up before sending an enquiry, the proof for each, and what to change first.

Prepared by Zyxware Technologies · 17 September 2026

1

What we are trying to grow, and what is stopping it

What we are trying to grow, and the biggest leaks ranked.

1A The goal in one line

More people who want a gold loan should reach the enquiry form, fill it in, and send it. Everything in this report is judged against that.

1B About this report

What we looked at

Every way people reach a Gold Loan enquiry on manappuram.com, not only the Gold Loan page: the rates page, home page, branch and contact pages, and the 13 language campaign pages. Plus what customers write in complaints and feedback.

What we did

Walked the live site as a phone and as a computer, and ran a borrower's visit on a slowed mid-range phone. Read Google Analytics from January 2024 to August 2026. Sorted 6,622 complaint and feedback messages. Compared the gold loan pages of IIFL Finance and Rupeek, which we could load, and checked each fix against RBI rules. Muthoot Finance and Bajaj Finance block automated visits, so we could not measure their pages. We never sent an enquiry or asked for the one-time SMS code (OTP).

What you will find

Ten leaks, each with the journey today, the proof, the fix, the journey after and how to measure it (chapter 4). Six customer journeys from complaints (5), the tracking problems and the correct setup (6), six bigger ideas (7), and a ranked list of where enquiries will come from (9).

What to expect

More enquiries from people who already visit: 208,385 rates page visits in three months with no way to enquire, and language campaign pages converting at half the rate of others. Every fix comes with a number to track. We do not promise a multiple of today's enquiries until the counting in Leak 1 is in place.

The full ranked list of leaks, and how many more enquiries each could bring, is in chapter 9, Where more enquiries will come from.

2

Which pages people arrive on, and where they give up

The Gold Loan page is only one door. The biggest leaks are on other pages.

People reach Manappuram through many doors, not only the Gold Loan page. We looked at every big one over a full year. Red doors get a lot of visitors and have no enquiry form or button at all.

2A Every page people arrive on, and what it brings in

Why a year: it evens out festivals and short ad pushes, and shows whether a leak is steady. The small bars under each page show its visits month by month. The campaign pages only started in December 2025, so theirs cover nine months. Other chapters use June to August 2026 where they describe the site as it is today; each says so.

Paid campaign page (English)Facebook, Google ads. Live since December 2025Phones 96% · Computers 3%
Visits1,614,789
Enquiries209,434
Has its own form on the opening screen, 480 pixels down. The best performing door. 12.97% of visits ended in an enquiry.
Home pageSearch, typed addressPhones 78% · Computers 21%
Visits981,329
Enquiries1,904
Has its own quick enquiry form, which the reports cannot count, so its real number is higher. 14,417 visits in June to August 2026 went on to the Gold Loan page. 0.19% of visits ended in an enquiry.
Scheme rates pageThe "9.90%" banner on the home page links herePhones 99% · Computers 0%
Visits829,532
Enquiries91
No form, no enquire button. Visitors stay about 6 seconds. 0.01% of visits ended in an enquiry.
Branch locator pageSearch for a branch near mePhones 24% · Computers 76%
Visits282,108
Enquiries9
Almost no visits go on to an enquiry. Unusually, most visits are on computers, which may include Manappuram staff looking up branches. The reports are not set up to tell staff and customers apart (chapter 6). 0.00% of visits ended in an enquiry.
Branch network pageSearch for branchesPhones 28% · Computers 72%
Visits295,104
Enquiries91
Almost no visits go on to an enquiry. Mostly on computers too, so part of this may be staff. 0.03% of visits ended in an enquiry.
Contact pageSearch, typed addressPhones 96% · Computers 4%
Visits244,239
Enquiries30
Has its own enquiry form on the second phone screen, which the reports cannot count. 0.01% of visits ended in an enquiry.
Gold Loan pageMostly Google searchPhones 79% · Computers 21%
Visits123,313
Enquiries945
Has its own form, but the reports cannot count enquiries sent on it (Leak 1), so its real number is much higher. 0.77% of visits ended in an enquiry.
The harm: the rates page and the two branch pages had 1,406,744 visits in a year, and none of them has an enquiry form or button. Only 191 of those visits ended in an enquiry.

What a red door looks like from the visitor's side

The same walk on any of the three red doors above. Drawn from the live pages on 17 September 2026, with the numbers for the rates page, the biggest of the three.

What happens now on a red door
9.90%
Learn more
1. Taps the rate banner or a search result, and lands on the rates pageProof829,532 visits a year land on the rates page, 577,212 more on the two branch pages
→
Schemes and Rates
Rates cut off, no way to enquire
Doubt
2. Reads a yellow table of scheme codes, hunting for the rateProofLive test: the rate columns run off the side of a 360 pixel phoneHarmThe average visit lasts about 6 seconds
→
Gold Loan interest
13% to 24% a yearloans from Rs 10,000
From 9.90% a yearRs 5 lakh or more
No form here
3. Looks for a way to ask. There is no form, no button, and no phone number on the page at allProofLive check, 17 September 2026: no form, no enquire button, and the phone numbers are not tappableHarm91 of 829,532 rates page visits ended in an enquiry anywhere
→
←
Back to Google
Lost here
4. Leaves, or goes hunting for another pageProof1,988 visits went from the Gold Loan page to the rates page in three months, and did not come back to itHarmA visitor who came to ask a price leaves without asking anything
What should happen instead
Schemes and Rates
Rates cut off, no way to enquire
1. Lands on the same page, with the same table
→
Mobile number
Pincode
Send
CallWhatsAppMy amount
2. A short enquiry sits under the first table: mobile number and pincode
→
Branch
680567
Valapad
Triprayar
3. The branch fills itself in from the pincode
→
✓
Enquiry sent
Success
4. Sends it from the page that answered the questionOutcomeThe pages that answer the price question can also take the enquiry, without buying a single extra visit.How we will knowEnquiries sent from the rates page and the branch pages, which is zero today.

2B The scoreboard is broken, and it is sending money the wrong way

The "Enquiries" numbers above only count forms sent on paid campaign pages. Forms sent on Manappuram's own pages are never counted, so those pages look far weaker than they are.

Two warnings before anyone adds these up. The "Enquiries" column above counts thank you page views in the website reports, and chapter 6 shows those views are counted more than once when a page is reloaded, so the campaign totals are higher than the number of real enquiries. Manappuram's own records, by contrast, show 14,489 Gold Loan enquiries for the same twelve months. The two counters do not agree and cannot be added together or divided into each other. Which of them is right, and what each one includes, is the first question we would put to whoever owns the enquiry report.

Counted by the website reports

Forms sent on paid campaign pages, because they open a thank you page.

Never counted

The Gold Loan page form, the home page quick form, the contact page form, and the forms on the per gram, calculator, interest rate, city and article pages.

14,489Gold Loan enquiries in the monthly count Manappuram gave us, September 2025 to August 2026. We did not measure this ourselves, and the file does not say which forms it covers
945enquiries the website reports can link to visits that started on the Gold Loan page, in the same 12 months
The harm: anyone using the website reports to decide where to spend sees paid campaign pages bringing thousands of enquiries and the Gold Loan, home and contact pages bringing almost none. That pushes money towards paid ads and away from the pages people find for free, purely because those are the only ones counted. It also means no page in this report can be fairly compared with another until every form is counted (Leak 1).

The two numbers above measure different things and should not be divided: Manappuram's records count Gold Loan enquiries from all its forms; the reports count only campaign page forms reached from the Gold Loan page. For the red doors, visitors would first have to go to another page to enquire.

2C Four language pages get half as many enquiries

December 2025 to 28 September 2026, the whole time the pages have run

English13.0%
Assamese12.8%
Hindi12.6%
Bhojpuri12.3%
Gujarati11.2%
Marathi9.7%
Bengali9.6%
Odia9.5%
Kannada7.2%
Punjabi7.1%
Telugu6.4%
Tamil5.7%
Malayalam3.3%

What the percentage means: of the people who clicked an ad or link and landed on that language page, the share who sent the form. It is not how many people saw or tapped the ad; that is inside Facebook's ad account, which we asked for and have not received. December 2025 (when the pages went live) to 28 September 2026. Red: Tamil, Telugu, Malayalam, Kannada.

5.6%of 202,812 visits to the Tamil, Telugu, Malayalam and Kannada pages sent an enquiry
11.0%of 267,366 visits to the other regional language pages did
The harm: about 10,900 fewer enquiries in nine months than if these four pages did as well as the others. These visits were paid for.

Is it the ads, the pages or the people?

Not only the ads

With the same kind of Facebook ad, on phones, June to August 2026: about 8 in 100 enquired on the four pages, against 17 in 100 on the others.

The pages have problems

The Tamil page labels the phone box "Required amount" (Leak 11), which explains Tamil but not the other three. The English parts of the form are the same on every language page, including the two that convert best, so they cannot explain this gap on their own.

Maybe the people

These are Manappuram's home states, where many may already have a gold loan. The ad account data would show this; we asked for it.

2D Ads sent people to a page they could not read

People tapped an ad meant for their state, and landed on a page they most likely could not read. 18,839 of 51,360 Malayalam page visits (37%) came from ads meant for other states.

Ad meant forMaharashtraLanded onMalayalam pageVisitors actually in Maharashtra1,792 of 1,943Sent an enquiry0.9%
Ad meant forTelanganaLanded onMalayalam pageVisitors actually in Telangana2,349 of 2,494Sent an enquiry1.0%
Ad meant forUttar PradeshLanded onMalayalam pageVisitors actually in Uttar Pradesh978 of 1,107Sent an enquiry1.0%
Ad meant forKeralaLanded onMalayalam pageVisitors actually in Kerala7,035 of 7,309Sent an enquiry5.8%

How we know: Google Analytics records which ad each visitor came from and which state they were in. All from Facebook. We did not see the ads themselves. Which state an ad was meant for comes from the ad label on each visit, the field Facebook fills with names such as "remarketing-TN-Tam-Sep2026", not from the campaign name. That matters, because every language ad on these pages reports its campaign name to the reports as "instant-state-wise-en", whatever language page it points at, so a check done on campaign names alone would show nothing wrong. Re-checked on the live pages on 21 September 2026.

Was it because of where people were? No.

They were in their own state

1,792 of 1,943 people from the Maharashtra ad were in Maharashtra, not Kerala. The same for Telangana (2,349 of 2,494) and Uttar Pradesh (978 of 1,107).

The ad decides the page

Each ad link opens one fixed page. A visitor in Maharashtra still got the Malayalam page, so the website did not choose it by location; the ad's link did.

Same page, different result

On the same Malayalam page and form, people from the Kerala ad enquired at 5.8%; people from the other states' ads at about 1%. The only difference is whether they could read it.

April to June 2026Only 753 of 5,823 Malayalam page visits in April were from Kerala
From July 2026About 95% of visits come from Kerala
TodayStopped, but not proven fixed. Whatever checks exist did not catch it for three months
The harm: 28,215 paid visits across the language pages came from an ad meant for a different state or language. At the normal enquiry rate, that is about 1,660 lost enquiries in nine months.

What the wrong language ads cost

Money spentRs 8.1 lakhsending 28,215 paid visits to a page in the wrong language
Money lostRs 4.5 lakhon top, 1,660 enquiries those visits did not send

If we fix itRs 8.1 lakh a year stops being spent sending people to a page they cannot read, and about Rs 4.5 lakh of enquiries comes back. Checking each ad link before it goes live is the whole fix.

How we worked this out, step by step
  1. 28,215 paid visits landed on a language page their ad was not meant for.
  2. One paid visit from a Google ad costs Rs 29 on average. 28,215 x Rs 29 = Rs 8.1 lakh. That money is already gone.
  3. Visits that land on the right language page send the form 8.5% of the time. Mismatched ones send it 2.7% of the time.
  4. That difference across 28,215 visits is 1,660 enquiries.
  5. 1,660 x Rs 273 an enquiry = Rs 4.5 lakh, on top of the money already spent.

Would switching the language by location fix this?

Not on its own. On 17 September 2026 we did not find a switch by location on the campaign pages we checked, so we cannot confirm how it works. If one is added or already live, three things still hold:

The ad link names the page

Each language has its own page address, and the ad points straight at one. A location switch usually respects a page that was asked for by name, so the wrong page would still open.

Place is not language

A Malayali working in Bengaluru would get Kannada; a Hindi speaker in Chennai would get Tamil. The ad already knows its language; guessing it from location is less accurate.

The check still covers both

Checking each ad's link before launch catches the mismatch and anything the location switch gets wrong.

To settle it, Manappuram's web team can confirm where the location switch is live and whether it overrides a page asked for by name.

The fix

Before any ad goes live, one person checks that it opens the page in the ad's own language.

For the marketing team: every mismatched ad
PageAd labelVisitsEnquiries
Malayalamremarketing-telangana-english-82paise-apr262,49424
Malayalamremarketing-maharashtra-marathi -82paise-apr261,94318
Malayalamremarketing-tamilnadu-english-82paise-apr261,31913
Malayalamremarketing-up-hindi-instant-apr261,10711
Malayalamremarketing-tamilnadu-english-instant-apr261,0249
Malayalamremarketing-tamilnadu-tamil-82paise-apr269554
Malayalamremarketing-andhra-pradesh-english-82paise-apr268484
Malayalamremarketing-tamilnadu-tamil-instant-apr267437
Tamilremarketing-TN-english-August202673454
Malayalamremarketing-odisha-english-instant-apr2671718
Malayalamremarketing-karnataka-english-instant-apr266990
Kannadaremarketing-KA-english-August202669942
Malayalamremarketing-KL-english-August202664826
Malayalamremarketing-kerala-english-july202655717
Kannadaremarketing-Karnataka-english-july202653136
Malayalamremarketing-kerala-english-82paise-apr2652612
Malayalamremarketing-telangana-telugu-instant-apr265184
Malayalamremarketing-karnataka-kannada-82paise-apr264480
Teluguremarketing-AndhraPradesh-english-july202635621
Hindiremarketing-MH-english-August202635245
Teluguremarketing-TG-english-August202634233
Punjabilp-interest-pondicherry-tamil-instant-apr263240
Punjabilp-interest-jharkhand-hindi-instant-apr263220
Malayalamremarketing-karnataka-kannada-instant-apr263210
Malayalamremarketing-mp-english-82paise-apr263100
Malayalamremarketing-andhra-pradesh-english-instant-apr263080
Malayalamremarketing-karnataka-english-82paise-apr262970
Punjabilp-broad-haryana-hindi-82paise-apr262957
Assameseremarketing-AS-english-August202627922
Hindiremarketing-UP-english-August202626441
Teluguremarketing-Telangana-english-july202624012
Hindiremarketing-MP-english-August202623133
Bengaliremarketing-WB-english-August202622720
Punjabilp-broad-bihar-bihari-instant-apr262247
Hindiremarketing-Maharashtra-english-july202622425
Odiaremarketing-OD-english-August202621716
Malayalamremarketing-maharashtra-english-instant-apr262120
Malayalamremarketing-haryana-english-82paise-apr262060
Bhojpurilp-bengal-bengali-low-interest-mar262024
Malayalamremarketing-maharashtra-marathi-instant-apr262010
Malayalamremarketing-gujarat-gujarati-instant-apr261960
Hindiremarketing-RJ-english-August202618524
Malayalamremarketing-mp-hindi-82paise-apr261840
Hindiremarketing-UttarPradesh-english-july202618424
Malayalamremarketing-gujarat-gujarati-82paise-apr261740
Malayalamremarketing-odisha-odiya-82paise-apr261640
Odiaremarketing-Odisha-english-july202615613
Bengaliremarketing-WestBengal-english-july202615524
Punjabilp-broad-haryana-haryanvi-instant-apr261480
Hindiremarketing-MH-Hin-Sep202614517
Punjabilp-broad-uttarkhand-hindi-82paise-apr261390
Hindiremarketing-MadhyaPradesh-english-july202613714
Punjabilp-broad-assam-assamese-instant-apr261273
Malayalamremarketing-haryana-haryanvi-82paise-apr261253
Malayalamremarketing-maharashtra-english-82paise-apr261210
Malayalamremarketing-gujarat-english-82paise-apr261170
Punjabilp-broad-uttarkhand-hindi-instant-apr261160
Punjabilp-broad-haryana-eng-instant-apr261154
Malayalamremarketing-mp-english-instant-apr261130
Punjabilp-broad-delhi-eng-instant-apr261130
Malayalamremarketing-kerala-english-instant-apr261123
Hindiremarketing-rajasthan-english-july20261089
Malayalamremarketing-odisha-odiya-instant-apr261010

Labels that name a different state or language from the page. Labels with no state or language, such as "panindia", are not counted either way. Remarketing ads in English sent to a regional page in the same state are counted as mismatched because the language differs.

For the marketing team: how we read the ad labels

When Manappuram runs an ad, the link behind it carries a short name tag, so the reports can tell which ad each visitor came from. The tags follow a pattern that says where the ad was aimed:

remarketingwho it targets: people who visited beforemaharashtrastatemarathilanguage82paisemessageapr26month

So we can check whether each ad sends people to the page in its own language. Here are two ads that both sent visitors to the Malayalam page:

Right page

KL-August2026
A Kerala ad, landing on the Malayalam page.
7,846 visits, 5.8% sent an enquiry.

Wrong page

remarketing-maharashtra-marathi -82paise-apr26
A Maharashtra ad in Marathi, landing on a page written in Malayalam, which the visitor most likely cannot read.
1,943 visits, 0.9% sent an enquiry.

2E What the Google Ads money buys

1 September 2025 to 31 August 2026

Search word ads

Text ads above the free Google results. Manappuram picks the words people type and pays for every tap.

Rs 84of every Rs 100 spent
brings
53of every 100 enquiries

Rs 437 per enquiry

Ads Google places for you

Manappuram sets a budget and one goal, a form sent. Google decides where each ad shows, including Google search, YouTube, Gmail, Maps and other websites, and what to bid.

Rs 11of every Rs 100 spent
brings
46of every 100 enquiries

Rs 68 per enquiry

Banner ads chasing past visitors

Picture ads on other websites, shown to people who have already been to manappuram.com.

3 of every 4 paid taps never opened the page. Rs 3.3 lakh spent in the year.

Why this matters: the two are bought in different ways, but after the tap they are the same: both land on the same campaign pages with the same enquiry form. So the gap is about how the visit was bought, not about different pages. One costs about 6 times more for each enquiry and takes most of the money. The ads Google places can also appear on Google search, so the two overlap, which is why this is a reason to test rather than proof.

Every kind of ad Manappuram runs, and what each one is

Kind of adWhat it is, in plain wordsSpent in the yearEnquiriesEach
Google search resultsText ads above the free Google results. Manappuram picks the words people type, and pays for every tap.Rs 2.97 crore67,851Rs 437
Google places the ad automaticallyManappuram sets a budget and one goal, a form sent. Google chooses where each ad appears, including search, YouTube, Gmail, Maps and other websites.Rs 40.1 lakh59,319Rs 68
YouTube: subscribers and video viewsBought to get videos watched and the channel followed, not to get enquiries.Rs 7.7 lakh2not the goal
Ads in YouTube, Gmail and Discover feedsAds that sit inside the feeds people scroll on YouTube, Gmail and the Google app.Rs 3.9 lakh729Rs 531
Banner ads shown again to past visitorsPicture ads on other websites, shown to people who have already visited manappuram.com.Rs 3.3 lakh880Rs 371
Google Maps listingPromotes the branch listing on Google Maps. Its job is to bring people to a branch, not to the website.Rs 0.3 lakh0not the goal

The whole account, so nothing is hidden behind the three cards above. YouTube and Maps campaigns are bought for views and listings, not enquiries, which is why their cost per enquiry is not a fair number.

Google Ads account→linked to Google Analytics→cost of each ad, next to the enquiries sent after it

How we know. 1 September 2025 to 31 August 2026. Cost is what Google charged; an enquiry is a form sent on a campaign page. Facebook and Instagram spend is not linked, so it is not included.

Why search ads cost more

People searching "Manappuram" by nameRs 375
Other gold loan searchesRs 355
People searching competitor namesRs 599
State by state general searches, from June 2026Rs 875
Ads Google places automaticallyRs 68

Cost per enquiry, by what the campaign name says it was bought for.

Search pays per tap

Every tap on a chosen search word costs money, whether or not the person enquires.

Automatic ads chase form sends

Google picks where they show and what to bid, to get a form sent as cheaply as possible. Manappuram's team still sets the budget, the goal, the words, images and page.

Rs 61.9 lakh on people already searching "Manappuram"

Some of them would have reached the site through the free search listing anyway.

Does this mean search ads are not working?

No

Search ads brought 67,851 enquiries, half of all enquiries from Google ads. They cost more for each one, and the comparison has three limits:

Overlap

Automatic ads also show on Google search, so they may take credit for enquiries search would have brought.

Last tap wins

An enquiry counts only for the last ad tapped, not ads seen before it.

Enquiry is not a loan

A dearer search enquiry may be more likely to become a loan. Only branch records can tell.

So the gap is a strong reason to test, not proof that search ads should stop.

Is Rs 437 a lot? Against the market, and against Manappuram's own best

Published range, financial services, Google ads, IndiaRs 400 to 1,200
Manappuram, ads Google places automaticallyRs 68
Manappuram, search result ads, allRs 437
Manappuram, state search campaigns from June 2026Rs 621 to 1,472
Rs 0Rs 400Rs 800Rs 1,200Rs 1,600
Against the market: search ads at Rs 437 sit at the low end of the published range.
Against Manappuram's own best: about 6 times the Rs 68 of automatic ads.
New state search campaigns: up to Rs 1,472, above the top of the published range.

Competitors do not publish what they pay per enquiry. The range is from Indian agencies for all financial services, not audited data: OwlClaw Technologies (Google ads, Rs 400 to 1,200) and GrowEasy (all channels, Rs 500 to 3,000).

What to do: test before moving the budget

Aim for: automatic ads near Rs 70, search ads under Rs 250 (our suggestion), no campaign above Rs 600.
The four week test
  1. Pick two states with similar enquiries today
  2. In one, move a fifth of the search budget to automatic ads for four weeks
  3. Compare loans given at branches for every rupee spent
  4. Scale up only if the test state wins

What today's ad money buys, and what it burns

Money spentRs 3.3 lakhon banner ads shown again to people who had already visited
Money lostRs 2.5 lakhof that same money bought taps where the page never opened

If we fix itRs 2.5 lakh a year stops being burnt the day the banner line is fixed or stopped. The bigger prize, up to Rs 2.51 crore a year, needs the four week test above before anyone can claim it.

How we worked this out, step by step
  1. Rs 3.3 lakh of the year's ad money went on banner ads shown again to past visitors. This is that one line, not the whole account.
  2. Those ads were tapped 48,332 times and every tap was paid for.
  3. Only 12,042 of those taps actually opened the page.
  4. So about 75 in every 100 rupees of this line, around Rs 2.5 lakh, bought nothing. The loss is part of the spend, not on top of it.
  5. Separately, on the same account: search ads bring an enquiry for Rs 437, and the ads Google places automatically bring one for Rs 68. If search matched that price, the same 67,851 enquiries would have cost Rs 2.51 crore less. That is a reason to test, not proof, for the three reasons given above.
  6. Facebook and Instagram are not linked to the reports, so none of their spend is in these figures.
See every kind of ad and every campaign
Kind of adSpentTaps on the adTaps where the page openedEnquiriesCost per enquiry
Google search resultsRs 2.97 crore947,21076%67,851Rs 437
Google places the ad automaticallyRs 40.1 lakh734,92162%59,319Rs 68
YouTube: subscribers and video viewsRs 7.7 lakh293,5604%2not the goal
Ads in YouTube, Gmail and Discover feedsRs 3.9 lakh39,45058%729Rs 531
Banner ads shown again to past visitorsRs 3.3 lakh48,33225%880Rs 371
Google Maps listingRs 0.3 lakh9,7682%0not the goal

Cheapest per enquiry

CampaignSpentEnquiriesEach
#GoldLoan_PMax_PANIndiaRs 27.7 lakh43,541Rs 64
Sok_LeadGen_Pmax_ZLB_States_050626Rs 8.8 lakh8,836Rs 100
# GoldLoan_Hindi States_ScalableRs 12.8 lakh6,069Rs 211
# GoldLoan_Brand_LI__Hindi States_ScalableRs 8.0 lakh2,979Rs 268
# GoldLoan_Brand_LI_9 Cities_South ClusterRs 10.5 lakh3,910Rs 269
# GoldLoan_LI_3 States_ScalableRs 15.9 lakh5,670Rs 281

Most expensive per enquiry

CampaignSpentEnquiriesEach
Sok_LeadGen_Search_HPS_Generic_KL_170626Rs 9.0 lakh576Rs 1,561
Sok_LeadGen_Search_HPS_Generic_TG_170626Rs 5.9 lakh432Rs 1,355
Sok_LeadGen_Search_HPS_Generic_KA_170626Rs 12.9 lakh962Rs 1,344
Sok_LeadGen_Search_HPS_Generic_OD_170626Rs 3.4 lakh290Rs 1,168
Sok_LeadGen_Search_HPS_Generic_TN_170626Rs 10.3 lakh973Rs 1,057
Sok_LeadGen_Search_HPS_Competitor_AL_310726Rs 3.2 lakh407Rs 775

Google ad campaigns that spent at least Rs 3 lakh in the year, not counting YouTube channel and Maps campaigns. Names are as set up in the ad account.

2F What Rs 3.52 crore bought

Spent on Google adsRs 3.52 crore
→
Taps on the ads2,073,241
→
Enquiries sent128,781
→
Loans givenNot knownonly branch records have this
Rs 267average cost per enquiry, leaving out YouTube and Maps ads
Rs 8.0 lakhon YouTube and Google Maps ads. These aim to grow the YouTube channel and the map listing, not to bring enquiries, so they are left out of cost per enquiry. Not wasted, just a different goal.

Profit or loss: what one loan cost in ads

We cannot say yet whether the ads made money, because we do not know how many enquiries became loans. So here is the ad cost of one loan for four possible answers. Finance picks the row that matches reality, and compares it with what a loan earns.

5 in 100 enquiries become a loan
Search ad: Rs 8,750 per loanAutomatic ad: Rs 1,353 per loan
10 in 100 enquiries become a loan
Search ad: Rs 4,375 per loanAutomatic ad: Rs 676 per loan
20 in 100 enquiries become a loan
Search ad: Rs 2,187 per loanAutomatic ad: Rs 338 per loan
30 in 100 enquiries become a loan
Search ad: Rs 1,458 per loanAutomatic ad: Rs 225 per loan

How to read it

If 1 in 10 enquiries becomes a loan, a loan from a search ad cost about Rs 4,375 in ads, and a loan from an automatic ad about Rs 676. If a loan earns more than that in interest and fees, the ad paid for itself. Some enquiries are counted twice when a thank you page is reloaded (chapter 6), so real costs are a little higher.

2G What the ads promise, and what people actually get

Interest of 82 paise210,781
Low interest313,250
Instant192,438
School reopening27,705
Other or none1,341,645

Visits to the language campaign pages, December 2025 to 28 September 2026, by the promise in the ad name. We have not seen the ads themselves.

What happens today
82 paise
Learn more
1. Sees an ad promising interest of "82 paise"Proof210,781 visits came from ads with "82 paise" in their name, Dec 2025 to 28 Sep 2026
→
Enquire Now
Mobile number
Pincode
Submit
2. Sends an enquiry on the campaign pageProofCampaign page form
→
82 paise? Only Rs 5 lakh+
Lost here
3. At the branch: 82 paise is only for loans of Rs 5 lakh or moreProofRates page: 9.90% needs a loan of Rs 5 lakh or moreHarmA paid enquiry turns into a let down at the branch
→
Offered 13% to 24%
Doubt
4. Needs Rs 40,000, so is offered 13% to 24% a yearProofRates page: the DS scheme for loans from Rs 10,000 runs 13% to 24% at 90 days, and Manappuram's own EMI calculator quotes 23.90% a year for a Rs 40,000 loanHarmThe customer doubts everything else the ad said
→
←
Leaves or complains
5. Feels misled, leaves or complainsProof20 customer messages about itHarmPaid enquiry and branch time wasted
The harm: 210,781 visits came from ads promising 82 paise (about 9.9% a year). Anyone borrowing less than Rs 5 lakh cannot get that rate. 20 customers wrote in about it, for example: "they involve me n my gold in 15% rate of intrest plan". Each one is an enquiry paid for, a branch visit wasted, and trust lost.

What the 82 paise promise costs

Money spentRs 60.5 lakhbringing 210,781 visits from ads that promise interest of 82 paise
Money lostRs 24.0 lakh8,788 enquiries, because these visits enquire less than every other ad's

If we fix itUp to Rs 24.0 lakh a year if these ads brought visits that enquire like the rest, on ad money already being spent. Plus fewer people arriving at a branch expecting a rate they cannot have.

How we worked this out, step by step
  1. 210,781 visits came from ads with "82 paise" in their name, December 2025 to 28 September 2026.
  2. One paid visit costs Rs 29 on average, so bringing them cost about Rs 60.5 lakh.
  3. Those visits sent the form 7.9% of the time. Every other ad averaged 12.1%.
  4. If they had matched the rest, that is 8,788 more enquiries. 8,788 x Rs 273 = Rs 24.0 lakh.
  5. We cannot prove the promise is the reason they enquire less, only that these ads bring visits that enquire less often than the rest. It is the clearest number we have for the cost of the promise gap on the website.
  6. The bigger harm is not on the website at all. It happens at the branch, where 82 paise turns out to need a loan of Rs 5 lakh or more. 20 customers wrote in about exactly that, and nothing on the website records how many walk away. That part stays uncounted.

2H The promise and the catch are on the same page, twelve screens apart

On the campaign pages built for this scheme, the promise is the first thing on the screen: "Pay only 82.5 paisa per 100 Rs.", at 147 pixels down, in 24 pixel type, translated into each language. The condition sits at about 9,438 pixels down, on screen 12 of 13, in 13.7 pixel grey English that no Tamil or Malayalam reader was promised: "Minimum loan limit to avail the MT+5 Scheme is 5,00,000/-. Annualised Rate of interest for 12 month scheme is 26.82%". Checked live on 21 September 2026.

The page says 82.5 paise, not 82 paise. Where this report quotes the page, it uses the page's own figure.

What to change

Every ad and page that shows 82.5 paise also says, in the same size: "for loans of Rs 5 lakh or more; smaller loans 13% to 24% a year". Better still, lead small borrower ads with the rate they will actually get.

After the fix
Rs 40,000: 23.90% a year
Learn more
1. Sees "Rs 40,000 costs about 23.90% a year"
→
✓
Enquiry sent
2. Sends an enquiry knowing the real rate
→
Same rate as the ad
3. The branch offers what the ad said
→
Takes the loan
Success
4. Takes the loan, trusts Manappuram

Whether each ad also said "for loans of Rs 5 lakh or more" we cannot tell without the ads, which we have asked for.

3

The customer we are designing for

A real picture of the visitor, and one borrower we follow through every fix.

Meet Lakshmi

38Karimnagar, TelanganaAndroid phoneWhatsApp every dayReads English slowly, prefers TeluguNever taken a gold loan

Her daughter's school fees of Rs 40,000 are due on Monday. She owns two gold bangles. She searches "gold loan" on Google and taps Manappuram.

Taken from the data

Telangana is a top four state for Gold Loan page visits. Most visitors are on Android phones, and 360 by 800 is the most common screen. Three in four are new to the site. Customers write in Telugu and other Indian languages. Every step of her visit in this report was measured on the live site.

Made up to tell the story

Her name, age, town, the exact amount and the bangles. She is one person built to stand for the most common visitor, not a real customer.

3A Who actually visits the Gold Loan page

Visits that started on the Gold Loan page, 1 March to 31 August 2026: 54,784 visits by 44,178 people. A visit is one sitting; a person who comes back three times is one person and three visits.

71%of visits were on phones (39,656 visits, 32,071 people). Computers: 16,168 visits.
35,054people were new to the site, and 11,339 had been before. Most are still deciding whether to trust Manappuram.
31%of visits (18,108) came before 9 am or from 7 pm. Branches are closed then, so nobody can call these people back. Only the website, WhatsApp or a call line can help them at that moment.
360x800is the most common screen size, so we tested at exactly this size. (June to August 2026, 3,825 people.)

Top states, 1 March to 31 August 2026

StateVisitsPeople
Maharashtra6,6005,450
Karnataka5,9184,939
Telangana5,7034,693
Tamil Nadu5,4304,405
Uttar Pradesh3,5662,888
West Bengal3,5192,646
Gujarat2,7852,169
Madhya Pradesh2,7532,012
Odisha2,6351,997
Kerala2,4402,018
Delhi Division2,0261,619
Andhra Pradesh2,0241,682

Top cities, 1 March to 31 August 2026

CityVisitsPeople
Bengaluru4,9874,195
Hyderabad4,9474,085
Chennai3,3712,714
Mumbai2,0371,724
Bhubaneswar1,8601,423
Kolkata1,8601,444
Pune1,8191,530
Delhi1,7361,372
Indore1,4741,102
Kochi1,2181,006
Ahmedabad1,162943
Patna1,148912

People are counted once in each row, so a person who visited from two cities appears in both rows. Do not add the people column up. "(not set)" rows, where Google could not tell the location, are left out.

When they visit, by hour, 1 March to 31 August 2026

036912151821

Red: before 9 am or from 7 pm. Monday is the busiest day (9,340 visits), Sunday the quietest (4,793). Hours are as recorded in the reports.

Phones, age and gender

June to August 2026, people who viewed the Gold Loan page: 26,566 on Android and 4,216 on iPhone. After those, the most common device makes are Vivo, Apple, Samsung, OPPO and Xiaomi; "Google" is left out because it includes computers using Chrome. Across the whole site, 80% of visits are on phones.

Age is unknown for 69% of people and gender for 66%, so treat these as a rough guide. Where known, the largest age groups are 25 to 34 (4,725), 18 to 24 (3,925) and 35 to 44 (2,480). 81% of people with a known gender are recorded as male.

3B Lakshmi's visit, screen by screen

A simple drawing of each screen she meets on her phone, with the real numbers for that step. The red tags show which leak from chapter 4 happens on that screen; they follow her journey, so they are not in number order. Tap a tag to jump to the full evidence.

gold loan
Gold Loan Up to Rs 5 Crore | Apply Online
Leak 5
1. Google search
21,197visits came from free Google search, Jun to Aug 2026
Leak 5The title promises "Apply Online". A first loan needs a branch.
→
Gold Loan interest rate starting at 9.90%
Apply Online for Loan Against Gold
Enquire Now
no price shown
no call button
Leak 2Leak 5Leak 4Leak 6
2. Gold Loan page, first screen
49,932visits opened the Gold Loan page, Jun to Aug 2026
Leak 2Nothing here says what the loan costs. The rate banner is on the home page, and it leads to the rates page, which has no form.Leak 5The heading says "Apply Online".Leak 4No price anywhere on the page.Leak 6No call button, and WhatsApp opens an unnamed number.
→
Best Gold Loan Schemes and Rates
no enquire button
Leak 2
3. Taps the 9.90% banner: rates page
208,385visits started on the rates page
48of them ended in an enquiry
Leak 2On a phone the rate columns run off the right of the screen, and there is no enquire button. Visitors stay about 6 seconds.
→
form is 8 screens down
Leak 3
4. Back on the Gold Loan page, scrolling
4.6%of phone visits scroll to the bottom (1,640 of 35,928)
Leak 3The form is 8 phone screens down the page. See all 9 real screens in Leak 3.
→
Enquire Now
Pincode abc keyboard
Branch
Phone
Send OTP
Overrides DNC / NDNC
Leak 7Leak 16Leak 7Leak 7
5. The enquiry form
2 of 19city pincodes return a wrong branch list or none
not countedhow many give up here: the SMS code figures were not shared
Leak 7Letters keyboard for numbers, side menu over Send OTP, and a box that gives up Do Not Call protection.Leak 16For some cities, the branch list offers branches in other states.
→
Thank you
Website reportsGold Loan page enquiries: 0
Leak 1
6. She sends it
3,501Gold Loan enquiries in Manappuram's records, Jun to Aug 2026
0seen by the website reports
Leak 1Her enquiry is saved, but the website reports never count it.
See the real screens from the live site
Gold Loan page, first phone screen
Gold Loan page, first phone screen
Rates page: rate columns cut off
Rates page: rate columns cut off
Form: side menu over Send OTP
Form: side menu over Send OTP
WhatsApp: unnamed number
WhatsApp: unnamed number

What Lakshmi's journey costs

Money spentNoneher visit is free: she arrives from a Google search
Money lostRs 26.1 lakha year across the two leaks on her path that we can price

If we fix itRs 26.1 lakh a year from the two priced leaks alone, without buying a single extra visit. Fixing Leak 1 first is what turns the other five on her path into numbers.

How we worked this out, step by step
  1. Lakshmi's walk touches seven leaks. Two of them can be priced today, and they are both about pages that cannot take an enquiry.
  2. The Gold Loan page, where the form sits nine screens down: Rs 3.4 lakh a year for each 1 in 100 visits that could enquire and does not.
  3. The rates page she leaves for, which has no form at all: Rs 22.7 lakh a year.
  4. Together that is Rs 26.1 lakh a year, at Rs 273 an enquiry, the price Manappuram pays to buy one through Google ads.
  5. The other five on her path cannot be priced yet: the missing price, the "instant" promise, the WhatsApp number, the form itself and the branch list. Three of them become measurable the moment Leak 1 is fixed.
  6. Her own visit costs nothing to win. She came from a free Google search, and 98 in 100 visits to the Gold Loan page arrive the same way.

Drawings, not screenshots: real screenshots are in each leak card. Every point shown was checked on the live site on 17 September 2026 at 360 pixels wide.

3C Her visit, timed on an ordinary phone

The live Gold Loan page, at 360 pixels wide, with the connection and processor slowed to the level Google's page testing tool uses for a mid-range phone on mobile data. Nothing was sent: the run stopped at Send OTP.

4.6 sbefore anything appears on screen
3.5 MBof mobile data used. Fully loaded after 19.9 seconds.
12 swipesof her thumb to reach the form
3.1 sfor the branch list to fill after typing pincode 505001. It offered 7 branches.
Every step of the simulated run
KeyboardThe pincode box does not ask the phone for the number keypad, so a letters keyboard opens.
Send OTPGreyed out after 5 digits, active only after all 10.
"Enquire Now" at the topJumps to the form in about 2.6 seconds. It works, if the visitor knows to tap it.
Rates pageFirst content after 3.3 seconds, 3.0 MB of data. The table is 522 pixels wide on a 360 pixel screen; every interest rate column is off the edge.
4

The leaks, one by one

Each with the journey today, the proof, the fix, and the journey after.

Each leak follows the same pattern: what Lakshmi goes through today, why people drop out, the proof, the change, her journey after, and how we will know it worked. Open "See all the evidence" for the full detail.

1

Nobody can see the enquiries, so no fix can be proved

Big leakAnalytics setup
8,692enquiries in Jan to Aug 2026 the reports never saw
What happens today
Enquire Now
Lakshmi
505001
Karimnagar Gandhi Road
98xxxxxx10
Submit
1. Lakshmi sends the Gold Loan formProofLive form on the Gold Loan page
→
✓
Thank you
2. Saved in Manappuram's enquiry systemProof8,692 Gold Loan enquiries in Manappuram's records, Jan to Aug 2026
→
Website reports
Gold Loan page visits, Jun to Aug 202649,932
Gold Loan page enquiries0
By page / branch / adNone
1
3. The reports never see it. Nobody can tell which page or ad workedProofNo Gold Loan form event among the 30 events recorded (Google Analytics, last 28 days). Oct 2025 to Sep 2026: 11,941 enquiries in Manappuram's records, 945 the reports can link to the Gold Loan pageHarmRs 3.52 crore of Google ads a year judged on visits, not enquiries

Why people drop out here

The website reports count visits, but not the moment someone sends the Gold Loan form. So they show zero Gold Loan enquiries, while Manappuram's own records show thousands.

Google, Facebook, WhatsApp and SMS cannot be compared on how many enquiries they bring, so ad money is judged on visits instead.

The reports are cluttered: they claim 99.74% of visits achieve something important, because simply arriving on the site is counted as an achievement.

The harm: 8,692 Gold Loan enquiries in January to August 2026 are invisible to the website reports. Rs 3.52 crore of Google ads in a year were judged on visits, not enquiries, and no change to any page can be proved to work.

What to change

Count every enquiry at the moment Manappuram's system saves it, with the page, branch and how it was sent. Clear out the clutter. Make every form record where the visitor came from. Check each week that the website count matches the enquiry records. Details for the analytics team are in chapter 6.

After the fix
Enquire Now
Lakshmi
505001
Karimnagar Gandhi Road
98xxxxxx10
Submit
1. Lakshmi sends the Gold Loan form
→
✓
Thank you
2. Saved, and the page tells the reports
→
Website reports
Gold Loan enquiries+1
PageGold Loan
BranchKarimnagar
Came fromGoogle search
3. Every page, branch and ad is judged on real enquiriesOutcomeEvery fix becomes measurable from its first weekHow we will knowThe website count of Gold Loan enquiries is within a few percent of Manappuram's records each week

How we will know

The weekly website count of Gold Loan enquiries is within a few percent of the enquiry records.

Is it safe? RBI, dark patterns and data rules
RBI rulesSafe. Nothing the customer sees or agrees to changes.
SEBI rulesSafe. SEBI covers investor information, not loan pages.
Dark patternsSafe. Nothing changes on screen.
Personal dataSafe if the event carries no name, phone or pincode. Send page and branch only.
2

The scheme rates page gets 208,385 visits and has no way to enquire

Big leakWebsite teamMarketing team
208,385visits in 3 months, 8 times the Gold Loan page
What happens today
Gold Loan
Learn more
1. Taps a Manappuram post inside FacebookProof78% of rates page visits opened inside another app, such as Facebook (Jun to Aug 2026)
→
Schemes and Rates
Scheme codes like TMA1A
2. Lands on a yellow table of scheme codes like TMA1AProofSeen on the live rates page, 17 Sep 2026People lost here829,532 people a year land on a page with no form and no button. What that is worth is priced once, in chapter 9.
→
Schemes and Rates
Rate columns cut off
Doubt
3. On her phone the interest rate columns are cut offProofLive test: table is 522 px wide on a 360 px phoneHarm99% of rates page visits are on phones
→
←
Leaves after 6 seconds
Lost here
4. No form, no button to ask. Leaves after about 6 secondsProofAverage visit 6 seconds; 91 of 829,532 visits in a year ended in an enquiryHarm829,532 visits a year with no way to enquire

Why people drop out here

The home page carries a top banner, "Gold Loan interest rate starting at 9.90%", and it links here. So the site itself sends interested people to its one gold loan page with no way to enquire.

The page is about rates, and on a phone the rates are the part she cannot see. 98% of visits are on phones, and nobody can tell which ad or message sent them.

This is not a one-off: over 12 months, September 2025 to August 2026, the page had 829,532 visits that started on it.

There is nothing to tap. Average time on the page is 6 seconds, and only 6,732 visits went on to the contact page.

We cannot tell who sends these visitors, because the links to this page are not labelled. 78% opened inside another app, such as Facebook.

Phone: rate columns cut off.
Phone: rate columns cut off.
Computer: the table fits, but 99% of visits here are on phones.
Computer: the table fits, but 99% of visits here are on phones.
The harm: 829,532 visits in a year landed on a page about rates with no way to enquire, and only 91 of them ended in an enquiry. These people were interested enough to look at the price.

What to change

1. Marketing: find which ad, post or message links here, and label the link so its visitors can be counted. 2. On phones, turn the table into one card per scheme, smallest loans first, rate in plain words. 3. Add the short enquiry and the bottom bar with Call, WhatsApp and Check my amount (same as Leak 3).

After the fix
Gold Loan
Learn more
1. Taps the same post
→
Gold Loan interest
13% to 24% a yearloans from Rs 10,000
From 9.90% a yearRs 5 lakh or more
2. Sees a card: "Loans from Rs 10,000: 13% to 24% a year, interest paid monthly"
→
Mobile number
Pincode
Send
CallWhatsAppMy amount
3. Right below: mobile number and pincode, or WhatsApp
→
✓
Enquiry sent
Success
4. Asks her question without leaving the appOutcomeFor scale only: if 1 in 100 enquired, that is about 695 enquiries a month. The Gold Loan page brings about 1,167 a month today.How we will knowEnquiries sent from the rates page (counted once Leak 1 is fixed). Time on page should rise from 6 seconds.

How we will know

Enquiries sent from the rates page (counted once Leak 1 is fixed). Time on page should rise from 6 seconds.

Is it safe? RBI, dark patterns and data rules
RBI rulesSafe. The same rates, shown readably. The RBI 2025 gold lending directions ask lenders to be clear about loan terms.
SEBI rulesSafe. SEBI covers investor information, not loan pages.
Dark patternsSafe. Nothing about the offer changes.
See all the evidence

The page called "Best Gold Loan Schemes and Rates" had 208,385 visits that started on it in June to August 2026, about eight times the Gold Loan page. 98% were on phones and arrived without any tracking tag, and 78% opened inside another app. A few carried a code Facebook adds to its links, so at least some come from Facebook or Instagram. On a computer the whole table fits. On a phone it shows scheme name, minimum and maximum loan; the rate columns are cut off. We asked for Manappuram's advertising data and have not received it, which is why we cannot name the source.

3

On a phone, the enquiry form is 10 screens down the Gold Loan page

Big leakWebsite team
12 swipeson a mid-range phone to reach the form
What happens today
gold loan
Manappuram Gold Loan
1. Taps Manappuram on GoogleProof21,197 visits started on this page from free Google search, Jun to Aug 2026Harm14,573 of them were on phones, and land on the screen shown next
→
Apply Online for Loan Against Gold
2. Sees a family photo and a heading. No form, no rateProof18,454 phone visits started on this page, Jun to Aug 2026. Live test: no form, amount or rate on screen 1Harm6,369 phone visits (35%) left within 10 seconds without opening another pagePeople lost here44,146 people a year leave within 10 seconds without opening another page. What that is worth is priced once, in chapter 9.
→
still no form
3. Scrolls past promises, offers, highlights and 10 questionsProofLive test on 29 September 2026: the form starts at 55% of the page height, on screen 11 of 20HarmOnly 4.6% of phone visits reached the bottom of the page (1,640 of 35,928). The average phone visit lasts 53 secondsPeople lost here72,471 people a year stay on the page but never scroll far enough to see the form. What that is worth is priced once, in chapter 9.
→
About 10 screens later, still no price. Goes back to Google
Lost here
4. About 10 screens later, still no price. Goes back to GoogleProofAt most 7.0% of visits that open this page end in an enquiry, and truly less: the enquiry records name no page, so every gold loan enquiry is counted against this one page. The campaign page, whose form starts on the opening screen, turns 12.6%, though its visitors come from adsHarm1,692 visits left for the calculator and 1,988 for the rates page, each one page further from the form

The real page on a phone, one screen at a time

Live Gold Loan page on a 360 by 800 phone, 17 September 2026. Each picture is one full screen. On that day the form started 6,673 pixels down, on screen 9, after 8 full swipes. The page has since grown: on 29 September the form sits 8,595 pixels down, on screen 11 of 20.

Opens hereGold Loan page on a phone, screen 1: Banner and
Screen 1Banner and "Apply Online" heading
→
After swipe 1Gold Loan page on a phone, screen 2:
Screen 2"Zero extra charges", loan per gram
→
After swipe 2Gold Loan page on a phone, screen 3: Special offerings
Screen 3Special offerings
→
After swipe 3Gold Loan page on a phone, screen 4: EMI calculator
Screen 4EMI calculator
→
After swipe 4Gold Loan page on a phone, screen 5: Highlights of the gold loan
Screen 5Highlights of the gold loan
→
After swipe 5Gold Loan page on a phone, screen 6: Links, then FAQ heading
Screen 6Links, then FAQ heading
→
After swipe 6Gold Loan page on a phone, screen 7: Questions and answers
Screen 7Questions and answers
→
After swipe 7Gold Loan page on a phone, screen 8: More questions
Screen 8More questions
→
Form here, after 8 swipesGold Loan page on a phone, screen 9: The enquiry form, at last
Screen 9The enquiry form, at last
The harm: only 4.6% of phone visits reach the bottom of this page (1,640 of 35,928, June to August 2026). Anyone who stops before it never sees the form unless they tap "Enquire Now" on screen 1.

Why people drop out here

On a phone the page is about 20 screens long. The form starts on screen 11.

People leave to find answers: visits that started here went on to the calculator (1,692), the rates page (1,988), the charges page (1,563) and the home page (1,995).

It gets worse for the people who engage most. Opening all ten questions near the bottom adds 4,443 pixels to the page, pushing the form a further five screens away. The more carefully someone reads, the further away the form moves.

The website reports note when someone scrolls to near the bottom of a page. The form sits about halfway down. In June to August 2026 only 4.6% of phone visits got that far (1,640 of 35,928), and 7.9% on computers. How many stopped at the form itself is not recorded.

Manappuram's own campaign page shows the first field of its form on the opening screen, 480 pixels down, and turns 12.6% of its visits into enquiries. IIFL Finance puts its form on the first screen.

The page makes its case for a gold loan once, in the special offerings block 1,769 pixels down: "Why go for Personal or Business loans when you have the power of gold with you!". That is a moment where a reader decides. Between that line and the form, 6,673 pixels down, there is not one button or link that leads to the form: the next things to tap are the calculator, then three boxes opening the rates, terms and charges pages.

The "Highlights of Manappuram Gold Loan" block, 3,524 pixels down, lists what to bring as a single unbroken sentence: a certified copy of passport, driving licence, Aadhaar, voter ID, NREGA job card, a recent photograph and "The Permanent Account Number or Form No. 60 as defined in Income-tax Rules,1962". The words "any one of" appear nowhere on the page, so a first time borrower cannot tell whether one of these is enough or all seven are needed, and reads the process as heavier than it is.

Anyone who scrolls past the form and still wants to ask meets a heading near the foot of the page, 11,014 pixels down, that reads "Contact Us for Gold Loan Enquiry". The button under it says "Write Your Feedback" and opens the contact page. The heading promises an enquiry and the button hands over a feedback form (measured on the live page, 17 September 2026).

In fairness: there is one "Enquire Now" text link near the top that does jump to the form. It is a red button, 164 by 42 pixels, on the first screen, and it is the only visible thing on the page that leads there. The point is not that it is hidden; it is that everything else on seventeen screens leads somewhere else. It sits on a yellow band near the top, and the scroll figures above are what people actually do with it.

Phone: first screen. No form, amount or rate.
Phone: first screen. No form, amount or rate.
Computer: first screen, also no form.
Computer: first screen, also no form.
The harm: 69% of Gold Loan page visitors are on phones, and only about 5 in 100 phone visits scroll to the bottom of the page. Anyone who stops before the middle never sees the form, and nobody counts how many that is.

What to change

Under the banner, add a two box enquiry: mobile number and pincode (the branch already fills itself in from the pincode). Keep all the page content below. Add a slim bar fixed to the bottom of the phone screen: Call free, WhatsApp, Check my amount.

After the fix
gold loan
Manappuram Gold Loan
1. Taps Manappuram on GoogleProofThe same 21,197 free Google visits a quarter, no extra ad money
→
Mobile number
Pincode
Send
CallWhatsAppMy amount
2. First screen: mobile number and pincodeProofThe campaign page already puts its form near the top: 63,529 of its 504,833 visits reached its thank you page, Jun to Aug 2026
→
Branch
505001
Karimnagar Gandhi Road
3. Karimnagar branch fills itself inProofLive test: the pincode box already fills the branch list, correctly for 13 of 19 cities tested (Leak 16 fixes the rest)
→
✓
Enquiry sent
Success
4. Sends it while the page is still open, or taps WhatsAppOutcomeMore of the people who arrive ready to ask will ask before they drift away. The campaign page shows Manappuram's visitors do use a form they can see, though its visitors come from ads, so it is not a like for like promise.How we will knowGold Loan page enquiry rate, today at most 7.0% of visits that opened the page, and lower once Leak 1 is fixed and each enquiry carries the page it came from. Show the new layout to half of visitors for the same weeks and compare.

How we will know

Gold Loan page enquiry rate, today at most 7.0% of visits that opened the page, and lower once Leak 1 is fixed and each enquiry carries the page it came from. Show the new layout to half of visitors for the same weeks and compare.

Is it safe? RBI, dark patterns and data rules
RBI rulesSafe. Moving a form does not change any disclosure.
SEBI rulesSafe. SEBI covers investor information, not loan pages.
Dark patternsSafe. It is easier to find, not harder to leave. No ticked boxes, no countdown timers.
See all the evidence
How far down a phone screen the enquiry form sits on three gold loan pagesManappuram Gold Loan pageform starts 8,595 px down, screen 12Manappuram campaign pageform starts 480 px down, screen 2IIFL Finance gold loan pageform starts 348 px down, screen 1Pale block is the first phone screen, 800 px tall. Red block is the enquiry form.

Before the form come a photo banner, four short promises, special offerings, a calculator link, highlights and ten questions. On a computer the form starts 5,957 pixels down. The "Enquire Now" button at the top does jump to the form. How far phone visitors really scroll is recorded in Microsoft Clarity, which is installed on this page. We now have that record, and chapter 12 sets it against the figures here.

4

The price was added on 29 September, in a table a phone cuts off

Big leakWebsite team
3of the four rate columns fit on a 360 pixel phone
What happens today
Wonders what it will cost. Now finds a rate table seven screens down
1. Wonders what it will cost. Now finds a rate table seven screens down
→
Schemes and Rates
Rates cut off, no way to enquire
Doubt
2. Opens the rates page. Codes and a table cut off on her phoneProof1,988 visits went from the Gold Loan page to the rates page, Jun to Aug 2026HarmShe leaves the page with the form to hunt for a price
→
Expects 9.90%
3. Remembers 9.90% from an ad and goes to the branch expecting itProofThe site banner says 9.90%; 210,781 ad visits were promised 82 paise
→
Offered a higher rate
Lost here
4. Is offered a higher rate. Feels misledProof20 customer messages about itHarmA customer lost at the branch after an enquiry was paid for

Why people drop out here

The two questions every borrower has are "how much will I get?" and "how much will I pay?". The page answers neither, so people leave to find out.

The banner's one supporting sentence is about the lender, not the loan: "India's first listed and highest credit-rated gold loan company, registered under the Reserve Bank of India and a trusted name for gold loans in India since 1949". It is a credential, not an amount, a rate or a speed, so the first screen gives a visitor nothing to weigh.

The rate they find elsewhere is often 9.90%, which needs a loan of Rs 5 lakh or more. 20 customer messages complain about that gap.

The rates page title people see on Google reads "Gold Loan Interest Rates 2026 @ 0.825% p.m.*". That rate is for loans of Rs 5 lakh or more only.

IIFL Finance shows its range on the page itself, "0.99% onward p.m. (11.88% to 27% p.a.)", next to a form asking only for name, mobile and pincode. Checked again on 18 September 2026.

The harm: In June to August 2026, visits that started on the Gold Loan page went off to the calculator (1,692), the rates page (1,988) and the charges page (1,563) to find the price. 20 customers wrote in about being offered a higher rate than the one they saw.

What to change

Add one plain box near the top. Suggested words, to be confirmed by MFL product and legal: "Interest 13% to 24% a year for loans from Rs 10,000, and from 9.90% for loans of Rs 5 lakh or more, when interest is paid every month. Base rate 24%. Annualised rate up to 26.82%." Add one worked example for a Rs 40,000 loan. Pull the numbers from the same source as the rates page so the two never disagree.

After the fix
Gold Loan interest
13% to 24% a yearloans from Rs 10,000
From 9.90% a yearRs 5 lakh or more
1. Sees: small loans 13% to 24% a year, paid monthly
→
Gold Loan interest
13% to 24% a yearloans from Rs 10,000
From 9.90% a yearRs 5 lakh or more
Rs 40,000: about Rs 400 a month
2. Sees what Rs 40,000 would cost her each month
→
✓
Enquiry sent
Success
3. Sends the enquiry knowing what to expect
→
Same rate as online
4. At the branch, the rate matches what she readOutcomeFewer people leave the page to hunt for a price, and fewer feel misled at the branch. Some will decide it is not for them before enquiring, which saves branch time.How we will knowShare of Gold Loan page visits that go on to the rates, charges or calculator pages should fall. Complaints about an advertised rate should fall.

How we will know

Share of Gold Loan page visits that go on to the rates, charges or calculator pages should fall. Complaints about an advertised rate should fall.

Is it safe? RBI, dark patterns and data rules
RBI rulesSafer than today. Showing the full range, not only the lowest rate, is what fair disclosure asks for.
SEBI rulesSafe. SEBI covers investor information, not loan pages.
Dark patternsFixes a risk. Showing a low starting rate first and the real cost later is close to "drip pricing", named in the 2023 dark pattern guidelines.
See all the evidence

This leak changed while the review was running. On 17 and 21 September 2026 the Gold Loan page carried no interest rate at all: no 9.90, no 0.825, no figure per year or per month. Only loan-to-value percentages and a Rs 10,000 to Rs 5 crore range, folded inside the ten questions near the bottom. When we re-walked the page on 29 September a new "Gold Loan Schemes" table had appeared 5,522 pixels down, on screen 7, showing MT + 5 at 9.90% a year for loans of at least Rs 5 lakh, SA + 1 at 11.50% from Rs 1.95 lakh, and DS at 12.00% from Rs 10,000.

Two things about that table. On a 360 pixel phone it is cut off at the right edge: the scheme name, the loan limit and the monthly rate fit, and the bi-monthly and quarterly columns run off the screen. That is the same fault this report describes on the rates page in Leak 2. And the rate a visitor sees first is still 9.90%, which needs a loan of at least Rs 5 lakh; someone borrowing Rs 30,000 pays 12.00%. The 20 customer messages about the advertised rate are about exactly that gap, and nothing in the new table closes it.

The table also made Leak 3 worse. It added about 1,900 pixels above the enquiry form, which moved from 6,673 pixels to 8,595, from 48% of the page to 55%. Microsoft Clarity recorded 16.1 in every 100 sessions reaching 48% of this page, and 11.6 reaching 55%. On how people scrolled before the change, moving the form that far costs roughly a quarter of the people who used to reach it.

5

"Apply Online, Instant Approval", but the first loan needs a branch visit

Medium leakWebsite teamMarketing team
2headings and the Google title promise an online loan
What happens today
gold loan
Gold Loan Up to Rs 5 Crore | Apply Online
1. Google shows: "Gold Loan Up to Rs 5 Crore | Apply Online"ProofLive page title, 17 Sep 2026
→
Get Gold Loan Online, Instant Approval
Enquire Now
2. Page says "Get Gold Loan Online, Instant Approval"ProofLive page heading, 17 Sep 2026
→
Bring your gold to a branch
Doubt
3. Learns she must bring her gold to a branchProofOnline Gold Loan page: online loans are "for the existing customers"HarmThe first promise she read turns out untrue
→
←
May not go
Lost here
4. Feels the promise was broken, may not goProofNot counted: nobody records enquiries that go coldHarmEnquiries that never become a branch visit

Why people drop out here

Manappuram's own Online Gold Loan page answers it plainly in its questions: "Do I need to visit a branch to apply for an online gold loan? Yes, a one-time branch visit is required to pledge your gold and complete registration." The same page says online gold loans are "for the existing customers", and step one is to pledge gold at a branch.

A first time borrower who believes the heading expects to finish on her phone. The first contact with Manappuram starts with a broken promise.

The Google title was changed on 17 September 2026 and still says "Apply Online".

The harm: Every one of the 49,932 visits to the Gold Loan page in June to August 2026 was told "Apply Online, Instant Approval". A first time borrower cannot get that. People who expect to finish online may give up when told to visit a branch; how many is not counted.

What to change

Change the heading to what happens: "Get a gold loan at your nearest branch. Book your visit here." Add a time such as "the same day" only if branches can keep it. Keep Online Gold Loan as a labelled option for existing customers: "Already have a gold loan with us? Borrow more online." Change the Google title the same way.

After the fix
Get a gold loan at your nearest branch
Enquire Now
1. Reads "Get a gold loan at your nearest branch. Book your visit here"
→
Bring your gold and Aadhaar
2. Knows from the start she will visit with her bangles and Aadhaar
→
✓
Enquiry sent
Success
3. Enquires and plans when to go
→
As promised
4. Arrives prepared. Nothing she read was untrueOutcomeEnquiries from people who will actually come to a branch.How we will knowShare of enquiries that become a branch visit, once branches report it. Complaints about misleading promises should fall.

How we will know

Share of enquiries that become a branch visit, once branches report it. Complaints about misleading promises should fall.

Is it safe? RBI, dark patterns and data rules
RBI rulesSafer than today. Describing the product accurately is what fair practice rules ask for.
SEBI rulesSafe. SEBI covers investor information, not loan pages.
Dark patternsFixes a risk. A promise the visitor cannot get is close to "bait and switch", named in the 2023 dark pattern guidelines.
See all the evidence

On the morning of 17 September 2026 the page title read "Gold Loan Online up to Rs 5 Crore". By that afternoon it had changed to "Gold Loan Up to ₹5 Crore | Apply Online | Manappuram Finance", and the description ends "apply online". The main heading says "Apply Online for Loan Against Gold"; the next says "Get Gold Loan Online, Instant Approval".

6

The WhatsApp button opens an unknown number, and phones have no call button

Medium leakWebsite team
0working ways to chat or call from a phone
What happens today
Gold Loan
1. Taps the WhatsApp iconProof172 taps on links leaving the Gold Loan page, Jun to Aug 2026; which link is not recorded
→
7594999984
hi
Unknown number
Doubt
2. Sees a chat with "7594999984", no name, no logoProofTested on a phone: link is missing the 91 codeHarmAn unnamed number looks like a scam
→
←
Closes the chat
Lost here
3. Worries it is a scam. Closes itProofNot counted: WhatsApp taps are not trackedHarmEvery WhatsApp enquiry from this button is at risk

Why people drop out here

The link is missing India's 91 code, so WhatsApp does not know it is Manappuram.

The only call link sits in the top bar, which is hidden on phones.

The reports recorded 172 taps on links leaving the Gold Loan page in three months, but not which link.

Today: unknown number.
Today: unknown number.
With 91 added: verified Manappuram account.
With 91 added: verified Manappuram account.
The harm: Every WhatsApp tap on the Gold Loan page opens a chat with an unnamed number, the kind fraud warnings tell people to avoid. The 69% of visitors on phones have no button to call. How many taps are lost is not counted.

What to change

Correct the WhatsApp link so it includes India's 91 code and opens the verified Manappuram account with "I want a gold loan" ready to send. Add "Call free 1800 420 22 33" to the phone bar. Count taps on both.

For the web team: the exact link
https://api.whatsapp.com/send?phone=917594999984&text=I%20want%20a%20gold%20loan
After the fix
Gold Loan
1. Taps WhatsApp
→
Manappuram Finance ✓
I want a gold loan
2. Sees "Manappuram Finance Limited" with the verified tick, message ready
→
✓
Enquiry sent
Success
3. Sends "I want a gold loan" and gets a replyOutcomeEvery tap reaches Manappuram. The link works the day it is changed.How we will knowWhatsApp and call taps from the Gold Loan page, and conversations started on the WhatsApp business account from the website.

How we will know

WhatsApp and call taps from the Gold Loan page, and conversations started on the WhatsApp business account from the website.

Is it safe? RBI, dark patterns and data rules
RBI rulesSafe.
SEBI rulesSafe. SEBI covers investor information, not loan pages.
Dark patternsSafe. It fixes a broken link.
See all the evidence

The link today is https://api.whatsapp.com/send?phone=7594999984&amp%3Btext=hi. Opened on a phone, WhatsApp shows "Chat on WhatsApp with 7594999984". The same number with 91 opens "Manappuram Finance Limited", a verified business account, with "hi" ready to send. There is a second fault in the same link: the text part is written twice over, so WhatsApp reads it as amp;text and the ready made "hi" never appears either. Both are the same one line fix. Checked again on the live page on 18 September 2026.

7

The form fights the phone, and asks people to give up Do Not Call protection

Medium leakWebsite team
4small blocks between typing and sending
What happens today
Mobile number
q w e r t y
a s d f g h
1. Types her pincode. A letters keyboard opensProofLive form code: the Pincode box is a plain text boxHarmSlower typing and wrong characters in a 6-digit box
→
Mobile number
Send OT
Side menu covers the button
Doubt
2. Taps Send OTP, but the side menu covers the buttonProofLive screenshot, 360 wide phone, 17 Sep 2026HarmMissed taps on the one button that moves her forward
→
Messages
Your OTP is 482913
Switching apps to copy it
3. Leaves to copy the OTP from SMS, then comes backProofOTP box has no auto-fill setting; 76 customer messages about OTP trouble
→
I authorise ... override any registration for DNC / NDNC
Submit
Lost here
4. Reads that ticking the box means giving up Do Not Call protection. Hesitates and leavesProofLive form wording; all 5,000 feedback records have it tickedHarmHow many leave here is not counted

The real form on a phone

Live Gold Loan page, 360 pixels wide phone, 17 September 2026. Red rings mark the problem. The phone keyboard and the SMS itself cannot be captured from the page, so the keyboard step is shown from the page code.

Pincode boxLive Gold Loan form on a phone: Pincode box. Plain text box. The phone opens a letters keyboard for a 6-digit number.
1. Pincode boxPlain text box. The phone opens a letters keyboard for a 6-digit number.
→
Send OTPLive Gold Loan form on a phone: Send OTP. The side menu sits on top of the right end of the button.
2. Send OTPThe side menu sits on top of the right end of the button.
→
The only tick boxLive Gold Loan form on a phone: The only tick box. Required, tiny, and it says ticking it
3. The only tick boxRequired, tiny, and it says ticking it "will override any registration for DNC / NDNC".
Same form on a 1366 wide computer, same day. The side menu sits outside the yellow form, so on computers only the tick box and the DNC sentence are problems.
Same form on a 1366 wide computer, same day. The side menu sits outside the yellow form, so on computers only the tick box and the DNC sentence are problems.
The harm: on phones, where most visitors are, every sender has to get past all three of these. A mistap on Send OTP means no code arrives, and the tick box asks people to give up Do Not Call protection just to ask a question. 76 customers wrote in about SMS code problems.

Why people drop out here

The Pincode and SMS code (OTP) boxes are plain text boxes, so a letters keyboard opens. The OTP box does not ask the phone to offer the code from the SMS. The Mobile box is set up correctly and opens the number keypad.

On a phone, the floating side menu covers a 6 pixel strip down the right edge of Send OTP. It is fixed to the screen, so how much of the button it covers depends on how far she has scrolled: at some positions none of it, at others the button's whole height. A tap that lands in that strip opens the menu instead of sending the code. The top 33 pixels of the 55 pixel button, and a tap test in that corner hits the menu, not the button. The tick box is 17 by 17 pixels, well under the 24 pixels a thumb needs. On a computer the menu ends 51 pixels clear of the button and does not cover it.

The form opens with the heading "Enquire Now" and goes straight into Name, Pincode, Branch, Phone and OTP. Nothing between the heading and the first box says why to send it, that the call is free, who rings, or when. This is the point where a visitor decides whether to hand over a phone number, and the page gives them nothing to decide on.

The last thing they tap is labelled "Submit". It names the mechanism, not the result, so at the moment of most hesitation nobody is told whether they are asking for a call, applying for a loan, or being approved. "Get my free call back" says what actually happens.

The only consent box is required and says ticking it "will override any registration for DNC / NDNC", meaning the person gives up their Do Not Call protection. The same forced box is on the feedback and complaint forms: all 5,000 feedback records we received have it ticked.

The harm: Before sending, people must choose a branch, tap Send OTP, type the SMS code and tick a box that gives up their Do Not Call protection. 76 customers wrote in about SMS code problems. How many give up here is not counted, because we have not received the SMS code and form error figures we asked for.

What to change

Make the Pincode and OTP boxes open the number keypad, and let the phone fill in the SMS code with one tap. Split the consent box in two: a required "Call or message me about this enquiry", and an optional, unticked box for offers. Remove the sentence that overrides Do Not Call from all three forms. Make the tick box bigger, and on phones hide the side menu while the form is on screen.

For the web team: the exact settings
Pincode:  inputmode="numeric" maxlength="6" autocomplete="postal-code"
Mobile:   inputmode="numeric" autocomplete="tel-national"
OTP:      inputmode="numeric" autocomplete="one-time-code"
Name:     autocomplete="name"

Four more things the live code shows, all cheap to fix. The pincode box accepts 255 characters for a six digit number. The name box accepts English letters only, so a name in an Indian script is refused. The page has two elements sharing the id "edit-submit", the site search button and the enquiry button, which is invalid and makes the form hard to test or script reliably. And no error is ever written into the page: every message is a browser bubble that vanishes on the next tap and is always in English.

After the fix
Mobile number
1 2 3
4 5 6
7 8 9
1. Number keypad opens
→
Enquire Now
Mobile number
Pincode
Send OTP
2. Taps Send OTP cleanly
→
Enter OTP
From Messages: 482913
3. Phone offers the code from SMS in one tap
→
Call or message me about this enquiry
Send me offers (optional)
Send
Success
4. Ticks "Call or message me about this enquiry". SendsOutcomeFewer people who start the form give up before sending. The forms stop asking customers to give up a legal protection.How we will knowOf the people who start typing in the form, how many send it. The page already tells Google Ads when someone starts typing, so this can be added to the reports.

How we will know

Of the people who start typing in the form, how many send it. The page already tells Google Ads when someone starts typing, so this can be added to the reports.

Is it safe? RBI, dark patterns and data rules
RBI rulesSafer than today. Separate, specific consent is what RBI conduct rules ask for.
SEBI rulesSafe. SEBI covers investor information, not loan pages.
Dark patternsFixes a risk. A required box that bundles marketing consent is close to "forced action", named in the 2023 dark pattern guidelines.
Personal dataSafer than today, and closer to what the data protection rules require from May 2027.
See all the evidence

The form has a separate Send OTP button that stays a faded blue until a valid 10 digit number is typed, the only blue control on a page where every other button is red, and taps on it do nothing at all until then, with no message saying why, a code box, then a required checkbox. The DNC wording was raised in the June 2026 audit and is still live. Customer records show how easily fields are misread: 132 feedback forms have an amount typed into the Loan Number box.

8

Customers who want to borrow more get stopped without warning

Big leakApp and product team
951customer messages about borrowing more on pledged gold
What happens today
My gold loans
Top up
1. A year later Lakshmi needs Rs 5,000 for medicine. Taps top up in the appProof951 customer messages about borrowing more
→
My gold loans
Top up
OVD update. Contact nearest branch
Lost here
2. App says "OVD update" or "contact nearest branch"Proof87 messages describe the app sending them to a branchHarm72 of them mention an emergency
→
Come back later
3. Goes to the branch. Waits. Told to come backProof81 messages mention waiting for head office
→
←
Borrows elsewhere
4. Complains. Borrows from someone elseProofMuthoot is named in 38 messagesHarmA loan lost to a competitor

Why people drop out here

These people already trust Manappuram with their gold. They are the easiest loans to give.

87 messages describe the app sending them to a branch to update identity documents (KYC and OVD, such as PAN or Aadhaar). 44 in 2025 mention a limit of 10 repledges a month; 9 from 2026 describe a limit of 6.

Complaints about it rose from 3 in June 2025 to 151 in October 2025. Muthoot is named in 38 messages.

The harm: 951 messages from existing customers who tried to borrow more on their gold and got stuck. 72 of them mention an emergency. Muthoot, a competitor, is named in 38 messages. These are the easiest loans Manappuram could give.

What to change

For MFL's app and product teams: 1. Check each customer's identity documents monthly and message the customer before it blocks a top up, naming the document and branch. 2. Show "repledges used this month: 4 of 6" (or the current limit) in the app, and state the limit on the Online Gold Loan page. 3. When a top up fails, give the reason and a button to book a branch visit.

After the fix
Manappuram Finance ✓
Your PAN needs updating at Karimnagar branch
1. Two weeks earlier, WhatsApp: "Your PAN needs updating at Karimnagar branch before your next top up"
→
PAN updated
2. Updates it on a normal day, without pressure
→
My gold loans
Top up done
Success
3. When she needs money, the top up goes throughOutcomeMore top ups completed in the app, fewer customers turned away at the moment they need money.How we will knowTop ups blocked by document checks each month, from the app. Complaints about borrowing more, which peaked at 151 in October 2025, should fall.

How we will know

Top ups blocked by document checks each month, from the app. Complaints about borrowing more, which peaked at 151 in October 2025, should fall.

Is it safe? RBI, dark patterns and data rules
RBI rulesSafe. KYC checks stay as strict; customers are warned earlier.
SEBI rulesSafe. SEBI covers investor information, not loan pages.
Dark patternsFixes a risk. Customers learn about the monthly limit only when it blocks them, often in an emergency. That is not a named dark pattern, but it is close in effect. The KYC blocks are a process failure, not a dark pattern.
Personal dataMessages must be about the customer's own loan. Offers need the separate optional consent from Leak 7.
See all the evidence

See Journey A in chapter 5 for the full chain, the exact app messages customers quote, and what we have not been able to see ourselves.

9

The Gold Loan page gets no credit for enquiries it wins

Smaller leakWebsite teamAnalytics setup
282enquiries in 3 months credited to the wrong page
What happens today
gold loan
Manappuram Gold Loan
1. Arrives from Google, taps "Instant Gold Loan"ProofLive link on the Gold Loan page redirects to the campaign page
→
Enquire Now
Mobile number
Pincode
Submit
2. Redirected to the paid campaign page and sends the form thereProof1,676 visits moved there, Jun to Aug 2026
→
Website reportsCredited to a paid ad
Lost here
3. Counted as an ad resultProof282 enquiries credited to the campaign pageHarmAds look better, the Gold Loan page looks worse

Why people drop out here

The Gold Loan page looks weaker than it is, and paid campaigns look stronger.

1,676 visits that started on the Gold Loan page moved to the campaign page in June to August 2026.

The harm: 282 enquiries in three months that were won by the Gold Loan page are credited to paid ads. Ads look better than they are, and the Gold Loan page looks worse, so money can be moved the wrong way.

What to change

Point the link at the Gold Loan page's own short enquiry. If it must stay, label it so enquiries are credited to the Gold Loan page.

For the analytics team: the label
?utm_source=manappuram&utm_medium=internal&utm_campaign=gold-loan-page
After the fix
gold loan
Manappuram Gold Loan
1. Arrives from Google
→
✓
Enquiry sent
Success
2. Sends the short enquiry on the Gold Loan page
→
Website reportsCredited to the Gold Loan page
3. Counted where it was wonOutcomeNo change for the visitor. Budget decisions are made on true numbers.How we will knowCampaign page enquiries from visits that started on the Gold Loan page should drop near zero, and Gold Loan page enquiries rise by the same.

How we will know

Campaign page enquiries from visits that started on the Gold Loan page should drop near zero, and Gold Loan page enquiries rise by the same.

Is it safe? RBI, dark patterns and data rules
RBI rulesSafe.
SEBI rulesSafe. SEBI covers investor information, not loan pages.
Dark patternsSafe.
See all the evidence

The link goes to /campaigns/instant-gold-loan-3, which redirects to /campaigns/instant-state-wise-en.

10

"Zero extra charges" and two different loan limits on the same site

Smaller leakWebsite team
2claims the rest of the site contradicts
What happens today
Zero extra charges
1. Reads "Zero extra charges"ProofLive Gold Loan page text, 17 Sep 2026
→
Processing charge Rs 30
Lost here
2. At closing, pays a processing chargeProofCharges page: Rs 30 gold processing charge at settlementHarm"Zero extra charges" is shown to be untrue
→
Tells family: they hide charges
3. Tells her sister Manappuram hides chargesProof298 messages about interest or charges being higher than expectedHarmWord of mouth turns negative

Why people drop out here

The charges page lists auction charges of Rs 540, a Rs 30 gold processing charge at settlement, penal charges and stamp duty in three states.

The per gram page says "up to 75%" and, lower down, "up to 85%". The RBI 2025 directions allow 85% up to Rs 2.5 lakh, 80% up to Rs 5 lakh, 75% above.

The harm: A customer who reads "Zero extra charges" and then pays a Rs 30 processing charge, or Rs 540 at auction, feels cheated. 298 messages are about interest or charges being higher than expected.

What to change

Replace "Zero extra charges" with what is true, for example "No processing fee to take the loan", if so, and link the full list. MFL product should confirm the loan to value limits, and the per gram page should state them once, in the three bands. Being able to borrow up to 85% on a small loan, if offered, is a strong honest reason to choose Manappuram.

After the fix
No processing fee to take the loan
1. Reads "No processing fee to take the loan", with a link to all charges
→
Rs 30, as shown online
2. At closing, the Rs 30 charge is one she already saw
→
Tells family: as described
Success
3. Tells her sister it was as describedOutcomeTrust that holds up at the branch, and less risk of the claim being treated as misleading.How we will knowComplaints about hidden or extra charges should fall.

How we will know

Complaints about hidden or extra charges should fall.

Is it safe? RBI, dark patterns and data rules
RBI rulesNeeds sign off from MFL legal and product on exact words. The change moves the page closer to the rules.
SEBI rulesSafe. SEBI covers investor information, not loan pages.
Dark patternsFixes a risk. An unqualified "zero charges" claim next to a charges page is the kind of claim the guidelines treat as misleading.
11

The language campaign pages ask for the details in English

Big leakWebsite teamMarketing team
470,178visits to the 12 regional language pages since December 2025
What happens today
தங்க கடன்
Learn more
1. Taps a Tamil ad and reads the page in TamilProof47,461 visits to the Tamil page, Dec 2025 to 28 Sep 2026
→
மணப்புரம் தங்க கடன்
பெயர்
Choose A Branch
Send OTP
Doubt
2. Starts the form. "Choose A Branch" and "Send OTP" are in EnglishProofLive page code, 17 Sep 2026HarmThe language promise breaks at the form
→
மணப்புரம் தங்க கடன்
பெயர்
தேவையான தொகை
தேவையான தொகை
Lost here
3. The phone number box is labelled "Required amount" in TamilProofLive page code: the phone box label and its error textHarmShe cannot tell where to type her phone number
→
மணப்புரம் தங்க கடன்
பெயர்
Please enter your name
Branch field is required
4. Error messages appear in English. Gives upProofThe error we could trigger without sending the form came back in English on Tamil, Hindi and Malayalam (live test, 21 Sep 2026)HarmThe Tamil page: 5.7% enquired, against 11.0% on other regional pages

Why people drop out here

On the Tamil page, the box for the phone number is labelled "தேவையான தொகை", which means "Required amount". The real amount box has the same label, so there are two boxes that both ask for an amount.

On all thirteen pages, "Choose A Branch", "Send OTP", "Enter OTP" and the four loan type choices are in English from the moment the page opens. So is the error a visitor sees after typing a wrong number: "Please enter a valid 10-digit mobile number."

Spelling mistakes a careful reader will notice: "Over 2 Core" on the Tamil, Kannada and Bengali pages, and a customer review starting "Test old loan" on the Gujarati page.

The reason is one setting, not a translation job half done. Every one of these pages is built as an English page, so every word the website itself supplies, the buttons, the branch list, the errors, comes out in English however much of the page is translated. Checked on the live pages on 21 September 2026.

What this does not explain. All thirteen pages carry the same English words, and they do not perform alike: the Assamese page turns 12.6 visits in every 100 into an enquiry, the Hindi page 12.2, and the Malayalam page 3.2. A fault every page shares cannot be the reason one page does four times better than another. So fix the English form because it is wrong, not because we have shown it is what costs these pages their enquiries. We have not.

What does differ between them. The ads. Ads bought for one state landed people on another state's page, and those visits enquire at about a third of the rate of visits that landed on the right one. That is measured, and chapter 2 sets it out. It is the strongest explanation we can evidence for why the worst performing pages perform worst.

There is a fourteenth page, in Haryanvi, whose headline is in Haryanvi and whose form is in English from top to bottom: Name, Pincode, Branch, Loan Type, Required Amount, Phone, Submit. The reports record no visits to it, so it is not in the numbers above, but it is running.

Tamil page: the last box is the phone box, labelled "Required amount". "Choose A Branch" is in English.
Tamil page: the last box is the phone box, labelled "Required amount". "Choose A Branch" is in English.
All language pages after pressing send with empty boxes: most errors appear in English.
All language pages after pressing send with empty boxes: most errors appear in English.
The harm: 470,178 visits to pages that promise a customer their own language, then ask for their details in English. On the Tamil page, a visitor looking for where to type their phone number sees two boxes asking for an amount. These pages include the four that convert at half the rate of the others.

What to change

First change the page language setting: each page is marked as English, which is why the website's own words stay English no matter what is translated. Then translate every word on the form, including buttons, the branch list message and every error, for all 13 languages. Fix the Tamil phone box label to "மொபைல் எண்" (mobile number). Correct the spelling mistakes and remove the test review. Have a native speaker check each page before it goes live.

After the fix
தங்க கடன்
Learn more
1. Reads the page in Tamil
→
மணப்புரம் தங்க கடன்
பெயர்
கிளையைத் தேர்ந்தெடுக்கவும்
OTP அனுப்பு
2. Every box, button and message is in Tamil
→
மணப்புரம் தங்க கடன்
பெயர்
மொபைல் எண்
3. The phone box says "Mobile number"
→
✓
Enquiry sent
Success
4. Sends the enquiryOutcomePeople who chose to read in their own language can finish the form in it.How we will knowEnquiry rate on each language page, compared before and after, starting with Tamil, Telugu, Malayalam and Kannada.

How we will know

Enquiry rate on each language page, compared before and after, starting with Tamil, Telugu, Malayalam and Kannada.

Is it safe? RBI, dark patterns and data rules
RBI rulesSafer than today. The RBI 2025 gold lending directions ask for communication in the language of the region or of the borrower's choice.
SEBI rulesSafe. SEBI covers investor information, not loan pages.
Dark patternsSafe.
See all the evidence

The label was read from the live page code on 17 September 2026: the box named field_phone carries the label and the error text "தேவையான தொகை field is required." Every language page is also marked as English in its code (lang="en"), so phones and Google treat it as English.

12

"Doorstep enquiry" on every inner page just opens the home page

Medium leakWebsite team
2,514visits in a year to the doorstep address
What happens today
BranchCalculatorDoorstep enquiry
1. Sees "Doorstep enquiry" in the side menu on any pageProofOn all 60 inner pages we checked
→
BranchCalculatorDoorstep enquiry
2. Taps it, hoping for gold loan at homeProof2,514 visits to the doorstep address in 12 months
→
One of India's Most Trusted Gold Loan Providers
Nothing about doorstep
Lost here
3. Lands on the home page. Nothing about doorstepProofLive test: /doorstep redirects to the home pageHarmA service promised on every page leads nowhere

Why people drop out here

The link sits in the floating side menu on all 60 inner pages we checked. It points to /doorstep, which redirects to the home page.

The home page says nothing about doorstep service, so the visitor cannot tell whether it exists.

Where "Doorstep enquiry" leads: the home page.
Where "Doorstep enquiry" leads: the home page.
The harm: The reports recorded 2,514 visits to the doorstep page address from September 2025 to August 2026. Today that address sends people to the home page, so a service promised on every page leads nowhere.

What to change

Either link it to a working doorstep enquiry page, or remove the button until the service is offered.

After the fix
BranchCalculatorDoorstep enquiry
1. Taps "Doorstep enquiry"
→
Enquire Now
Mobile number
Pincode
Book doorstep visit
Success
2. Sees a short doorstep form, or a clear note if it is not offered in her areaOutcomeEvery tap leads to a real enquiry, or the false promise is gone.How we will knowTaps on the doorstep button and enquiries from the doorstep page.

How we will know

Taps on the doorstep button and enquiries from the doorstep page.

Is it safe? RBI, dark patterns and data rules
RBI rulesSafer than today. A service shown but not offered is a misleading claim.
SEBI rulesSafe. SEBI covers investor information, not loan pages.
Dark patternsFixes a risk. Offering something that is not there is close to a "bait and switch".
13

Branch pages have no way to enquire, and their phone numbers cannot be tapped

Big leakWebsite team
577,212visits in a year to the branch locator pages
What happens today
branch near me
Branch Network | Manappuram
1. Searches "Manappuram branch near me" and lands on the branch listProof577,212 visits to the branch pages in 12 months, of which about 151,447 are on a phoneHarmAbout three quarters of these visits are on a computer, which is the opposite of the rest of the site, so some of this is Manappuram staff rather than customers
→
Branch list
Karimnagar0878 22xx
Kothi Rampur0878 23x
Numbers cannot be tapped
Doubt
2. Finds a phone number, but it cannot be tapped; on a phone it is cut off mid numberProofAbout 8,254 branch numbers are plain text (live check)HarmCalling the branch takes copying a cut off number
→
Samastipur
No results
Lost here
3. Searches her village or pincode: no resultsProof"Samastipur" and "848101" returned nothing (live test)HarmRural visitors cannot find their branch
→
←
Leaves
4. No enquiry form, no button. LeavesProof100 of 577,212 visits ended in an enquiryHarmPeople ready to visit a branch cannot enquire

Why people drop out here

About 8,254 branch phone numbers are plain text, not tap to call. On a phone the table cuts them off.

Searching "Samastipur" or the pincode 848101 returns nothing.

Neither branch page has an enquiry form or button. The only enquiry link, "Doorstep enquiry", is broken (Leak 12).

Most of these visits are on computers, so some may be staff. The reports cannot separate staff from customers.

Branch list on a phone: numbers are plain text and cut off.
Branch list on a phone: numbers are plain text and cut off.
The harm: 577,212 visits a year to pages built for people looking for a branch, with no way to enquire and no number they can tap.

What to change

Make every branch phone number tap to call. Add "Enquire at this branch" to each branch, opening the short form with the branch filled in. Let the search find villages and pincodes. On phones, show branches as cards, not a wide table.

After the fix
Branch list
Karimnagar0878 22xx
Kothi Rampur0878 23x
Enquire at this branch
1. Finds her nearest branch
→
✆
Calling Karimnagar branch
2. Taps the number to call, or taps "Enquire at this branch"
→
✓
Enquiry sent
Success
3. Sends an enquiry with the branch already chosenOutcomePeople who are ready to visit a branch can enquire or call in one tap.How we will knowCalls and enquiries started from the branch pages.

How we will know

Calls and enquiries started from the branch pages.

Is it safe? RBI, dark patterns and data rules
RBI rulesSafe.
SEBI rulesSafe. SEBI covers investor information, not loan pages.
Dark patternsSafe.
14

Many gold loan pages have no way to enquire, and articles bury the form at the bottom

Medium leakWebsite team
161,300visits in a year to these pages
What happens today
Charges on Gold Loans
No way to enquire
1. Reads the charges, FAQ or terms page to check before borrowingProof161,300 visits to these pages in 12 months
→
No form and no enquire button on the page. The only way on is an unlabelled icon that leads to the complaints page
Lost here
2. No form and no enquire button on the page. The only way on is an unlabelled icon that leads to the complaints pageProofMeasured across 75 gold loan pages on 17 Sep 2026: 13 carry no form at all, and 32 show no enquire button on the first phone screenHarm32 of 75 pages give a phone visitor nothing to tap without scrolling

Why people drop out here

No form and no button: charges (22,534 visits), general terms (16,115), online gold loan help (10,733), FAQ (2,865), products (19,871).

Articles and guides (89,182 visits across 297 articles) have the form only at the bottom. We measured sixteen of them: the form sits between 4 and 16 phone screens down, most often around 9, and not one has a button near the top.

The charges page opens with the auction charges table (total Rs 540) as the first thing a visitor sees.

There is one route off these pages: a small icon in the floating menu, labelled "Get in touch" in the code but showing no words on a phone, which leads to the contact page. That page is headed "Contact Us" and surrounded by complaint and feedback forms, so it reads as the place to complain, not the place to borrow. It is a route, but not one a borrower would take.

Manappuram has already built what we are asking for: the city pages, such as the Thrissur and Mumbai gold loan pages, carry the enquiry form in the page. This is a matter of putting an existing block on more pages, not building something new.

The harm: 161,300 visits a year to pages people read just before deciding, with no way to act on that decision.

What to change

Add the same short enquiry and bottom bar (Call, WhatsApp, Check my amount) to every gold loan page and article. On the charges page, lead with "No processing fee to take the loan" (if true) before the auction table.

After the fix
Charges on Gold Loans
1. Reads the same page
→
Mobile number
Pincode
Send
CallWhatsAppMy amount
Success
2. A short enquiry sits at the top and the bottomOutcomePeople can enquire from the page where they made up their mind.How we will knowEnquiries sent from these pages, counted once Leak 1 is fixed.

How we will know

Enquiries sent from these pages, counted once Leak 1 is fixed.

Is it safe? RBI, dark patterns and data rules
RBI rulesSafe. The charges stay fully shown.
SEBI rulesSafe. SEBI covers investor information, not loan pages.
Dark patternsSafe.
See all the evidence

Visit counts are from Google Analytics for the year to 31 August 2026: charges 22,534, general terms 16,115, online payment help 10,733, questions 2,865, products 19,871, and 89,182 across 297 articles. Whether each page has a form or an enquire button, and how far down the articles put theirs, was checked on the live site on 17 September 2026 at 360 pixels wide.

15

Four pages open already scrolled past their headline and Apply button

Smaller leakWebsite team
4pages affected
What happens today
gold loan
Gold Loan Transfer | Manappuram
1. Opens the gold loan transfer page from GoogleProof1,955 visits to this page in 12 months
→
Headline and Apply button are above, out of view
Lost here
2. The page jumps down on its own, past the headline and the Apply buttonProofLive test: page moves 873 px down on loadHarmThe Apply button is out of view from the start
→
Headline and Apply button are above, out of view
3. Lands in the middle of a paragraphProofScreenshot a few seconds after opening, 17 Sep 2026HarmVisitors start past the Apply button

Why people drop out here

Agriculture gold loan, gold loan transfer, Gurgaon and Indore pages jump 873 to 2,380 pixels down on load, because a tab section on the page moves itself into view.

Gold loan transfer page on a phone, a few seconds after opening: already past the headline.
Gold loan transfer page on a phone, a few seconds after opening: already past the headline.
The harm: 2,540 visits a year to these four pages start in the middle of the page, past the button that leads to the form. Gold loan transfer, one of them, is a product the report suggests growing (Idea E).

What to change

Stop the tab section from scrolling itself into view when the page loads.

After the fix
gold loan
Gold Loan Transfer | Manappuram
1. Opens the page
→
Gold Loan Transfer
Enquire Now
Success
2. Sees the headline and Apply button firstOutcomeVisitors see the headline and Apply button first.How we will knowShare of visits to these pages that tap Apply.

How we will know

Share of visits to these pages that tap Apply.

Is it safe? RBI, dark patterns and data rules
RBI rulesSafe.
SEBI rulesSafe. SEBI covers investor information, not loan pages.
Dark patternsSafe.
16

The branch list sends some cities to the wrong branches, and it changes without warning

Big leakWebsite team
2 of 19city pincodes return a wrong branch list or none, checked 21 September 2026
What happens today
Types her Bhubaneswar pincode, 751001
1. Types her Bhubaneswar pincode, 751001ProofOne of the 19 city pincodes we sent to the form's own branch lookup, 17 and 21 Sep 2026
→
The required Branch list shows Angul, Talcher, Dhenkanal, Nayagarh
Lost here
2. The required Branch list shows Angul, Talcher, Dhenkanal, NayagarhProofLive lookup, 21 Sep 2026: all four are 100 to 150 km away, and three branches inside Bhubaneswar are skippedHarmBranch is a required box, so the form cannot be sent without choosing one of them
→
Enquire Now
Mobile number
Pincode
Branch is required
Send OTP
3. None is near her. She cannot send the form without picking oneProofBranch is a required box (live form)
→
←
Gives up
4. Picks one at random, or gives upProofNot countedHarm2 of the 19 pincodes we tested still get a wrong list or none, on 21 September 2026

Why people drop out here

Branch is a required box and the list is filled from the pincode. On 17 September, five of the nineteen pincodes we tested returned lists made up mostly of branches in other states: Mumbai was offered Kazipet and Ameerpet, Delhi was offered Chhapra City and Madhubani.

The lookup changes without notice. Nothing on the site records when it changes, and nothing tests it, so a wrong list can appear and disappear without anyone knowing.

Bhubaneswar still fails. Pincode 751001 returns Angul, Talcher and Dhenkanal, all 100 to 150 kilometres away, while three branches inside Bhubaneswar are skipped. The neighbouring pincode 751007 returns those same three correctly, so the fault is per pincode and invisible unless someone tests that exact one.

For Kohima the list says "No Branches Found", so the form cannot be sent at all, and there is no other way offered. Manappuram has no branch in Nagaland, so the honest answer would be to say so and offer a call back instead of a dead end.

The branch search on those pages only matches a pincode a branch actually sits on. Typing 400001, 110001, 226001 or 751001, or the word Bhubaneswar, returns "The Branch you have searched for is not found", even where branches exist a few streets away.

If a visitor picks a far away branch just to get past the box, that branch may be the one asked to call back.

Pincode 600001, Chennai

  1. Broadway
  2. Parrys
  3. Parrys Corner(FOREX)
  4. Thambu Chetty Street
  5. GPO junction Mount Road
  6. Chinthadripet
  7. Thousand Light
  8. Mylapore
  9. Triplicane
  10. New Secretariate Triplicane

Pincode 751001, Bhubaneswar

  1. Sanker Cinema Road Angul
  2. NALCO Chowk Angul
  3. Banarapal
  4. Talcher
  5. Athamalik
  6. Dhenkanal
  7. Nayagarh
  8. Itamati
  9. Kanchan Bazar
  10. Kamakhya Nagar
The harm: Nothing on the site records when this list changes and nothing tests it, so a wrong list can sit there unnoticed. In the last six months, visits that started on the Gold Loan page came from Mumbai (2,037), Delhi (1,736), Bhubaneswar (1,860) and Lucknow (1,111), four of the cities where the list was wrong. How many of them gave up at the branch box is not counted. We tested one pincode per city, so other pincodes in these cities may behave differently.

What to change

Test the lookup against the top 100 visitor pincodes every month and after every release, because this is a fault that comes and goes silently. Fix the lookup so it always returns the nearest branches, sorted by distance. If no branch is within reach, allow sending without a branch and route the enquiry to the regional office. Check the lookup for the top 100 visitor pincodes every month.

After the fix
Branch
400001
1. Types her pincode
→
Branch
400001
Nearest branch
Next nearest
2. Sees the branches nearest to her, closest first
→
✓
Enquiry sent
Success
3. If none is near, sees "We will call you to find your nearest branch" and can still sendOutcomeEvery visitor can pick a nearby branch, or send the enquiry without one.How we will knowEnquiries by state, and complaints or call backs where the chosen branch is far from the customer's pincode.

How we will know

Enquiries by state, and complaints or call backs where the chosen branch is far from the customer's pincode.

Is it safe? RBI, dark patterns and data rules
RBI rulesSafe.
SEBI rulesSafe. SEBI covers investor information, not loan pages.
Dark patternsSafe. It removes a block, and nothing is chosen for the customer.
See all the evidence
PincodePlaceBranch list offered
680567Thrissur, KeralaNearby branches
505001Karimnagar, TelanganaNearby branches
600001ChennaiNearby branches
560001BengaluruNearby branches
500001HyderabadNearby branches
700001KolkataNearby branches
411001PuneNearby branches
380001AhmedabadNearby branches
302001JaipurNearby branches
452001IndoreNearby branches
462001BhopalNearby branches
800001PatnaNearby branches
122001GurugramNearby branches
400001MumbaiNearby branches
110001New DelhiNearby branches
226001LucknowNearby branches
751001BhubaneswarMostly far away: Sanker Cinema Road Angul, NALCO Chowk Angul, Banarapal, Talcher
848101Samastipur, BiharNearby branches
797001Kohima, NagalandNo branch offered

Checked on 17 September 2026 by sending each pincode to the Gold Loan form's own branch lookup, exactly as the form does when a pincode is typed. No enquiry was sent. Kohima is shown as "No Branches Found", even though the form still says "Branches loaded".

4A Smaller leaks from the full website check

We checked 65 pages and all 13 language campaign pages, on a phone and a computer. These are real but smaller, or need checking.

What we foundDetailSize
The home page has an "Apply Now" button that shows nothingHalf way down the home page, "Apply Now" jumps the visitor back to the top of the page, where the enquiry form is still folded shut. The visitor is returned to the start and shown nothing. Live test, 21 September 2026.Medium
There are two rates pages, and only one of them can take an enquiryThe scheme rates page has no form at all. A second page, the gold loan interest rates page, does carry a form and an "Apply Gold Loan" button. The two pages also frame the price differently, one as 9.90% a year and the other as 0.825% a month, so a visitor who sees both gets two answers to one question.Medium
No map and no "near me" anywhere on the siteThe branch page is headed "Gold Loan Branch Near Me", but there is no map and the site never asks for the visitor's location. Finding a branch means typing a pincode a branch already sits on.Medium
On a phone there is no number to tap on the pages built for finding a branchThe toll free number is hidden at phone width on the branch pages, the branch locator, the products page, the charges page and the terms page. Only the home page shows a tappable "Call Now".Medium
The form rejects names written in Indian scriptsThe name box accepts English letters only. On the language campaign pages, a visitor who types her name in Tamil, Hindi or Malayalam is refused, by a message that is itself in English.High
The form never shows an error in the pageEvery error is a small grey bubble from the phone browser itself. It disappears on the next tap, it is always in English whatever language the page is in, and a visitor who has scrolled never sees it.Medium
A page you can reach in Haryanvi is answered in EnglishA fourteenth campaign page, in Haryanvi, carries a Haryanvi headline and a form that is English from top to bottom. A fifteenth, labelled Rajasthani, is word for word the Hindi page. Neither has any recorded visits, so neither is in the numbers, but both are live.Medium
No call number on gold loan pages on a phoneThe call link in the top bar is hidden on phones and the side menu has no call icon. Only the home page shows "Call Now".Medium
Tamil and Kannada pages put the phone box below the first screenThe accessibility button also covers a form box on every campaign page.Medium
Campaign and home forms ask two extra required questions"Type of service" (including business and property loans) and "Required amount". The Gold Loan page form does not.Medium
Pages are heavy on a mid-range phoneHome page 4.1 MB and about 7 seconds before anything shows; main gold loan pages 3.1 to 3.5 MB and 4 to 6 seconds. The English campaign page shows its form in 2.4 seconds. Every page also makes a failing request to an outside service.Medium
Saved campaign pages carry an earlier visitor's ad codesThe saved page itself carries another visitor's ad click code, and every language page reports itself to the reports as "instant-state-wise-en". If enquiry records take their source from these, the language comparison is distorted. Not confirmed: needs access to the enquiry records.Needs checking
Home "Apply Now" on a phone jumps to the top with the form still closedSmall
404 page and site search do not lead back to a gold loan enquiryThe 404 page only offers "Back to Home". Searching "gold loan" does not show the Gold Loan page in the top 5; Hindi or "gold lone" returns nothing.Small
No language choice on the main websiteLanguage versions exist only for campaign pages; the main pages do not link to them.Small
Search result titles and descriptions have errorsOne description is page code, one title is just "IGGL", language pages read "Instant Gold Loan Hi", the contact title is "Get Touch With US".Small
Calculator accepts impossible weights100,000 grams shows about Rs 104 crore, above the Rs 5 crore maximum. It does lead to the form.Small
Gold Loan page form errors are browser pop-upsIn English, one at a time, with nothing marking the box that is wrong.Small

What we checked and found working: the home page "Enquire Now" opens the form; Apply and Enquire buttons near the top of inner pages scroll to the form once the page has finished loading, though on a cold load the scroll can stop part way and leave the visitor stranded; both calculators work and lead to the form; no gold loan link in the menus or footer returns an error page, although the "Doorstep enquiry" link quietly redirects to the home page (Leak 12); the contact page has a form on its second phone screen.

5

What customers are already telling you

Complaints and feedback show the leaks from the other side.

6,622 gold loan complaints and feedback messages, January 2025 to September 2026, sorted by what they are about. We read a random sample of each group to check the sorting.

2,697complaints
3,925feedback messages
2,000complaints missing from the file we received (14 Nov 2025 to 11 Apr 2026)
5,000feedback records: the file stops at exactly this number, on 2 August 2026
What Gold Loan customers wrote about, share of 6,622 complaints and feedback messages, January 2025 to September 2026The app or website is not working1,165 (18%)Borrowing more on pledged gold (top up, repledge)951 (14%)Money not reaching the account, or payment not showing944 (14%)Stuck updating KYC, bank account or mobile number939 (14%)Bad service, with no detail given823 (12%)Getting gold back or closing the loan526 (8%)Staff behaviour505 (8%)Long waits at the branch419 (6%)Nobody answers the phone or calls back334 (5%)Interest or charges higher than expected298 (5%)Gold damaged, or weight and purity disputes182 (3%)Credit score (CIBIL) wrong or not updated179 (3%)Branch system or server not working153 (2%)Auction of pledged gold116 (2%)One message can mention more than one problem, so the shares add up to more than 100%.

5A Six journeys where customers get stuck

Complaints rarely describe one problem. We traced the chains. Counts are messages mentioning each problem; one message can mention several.

Journey A: borrowing more on gold already pledged

An existing customer wants more money on the same gold, through a top up or repledge in the app.

What the customer goes through
My gold loans
Top up Rs 5,000
1. Sees an eligible extra amount in the appProof951 customer messages mention this
→
My gold loans
Top up
2. Taps top up or repledgeProof228 customer messages mention this
→
My gold loans
Top up
Contact nearest branch
Lost here
3. App blocks it and sends them elsewhereProof87 customer messages mention thisHarm72 mention an emergency
→
Branch
4. Visits or calls the branchProof149 customer messages mention this
→
Waiting for head office
Lost here
5. Branch waits on head officeProof81 customer messages mention thisHarmThe loan is delayed past the day it was needed
→
Bank account
Rs 0 received
6. Money does not arrive, or they borrow elsewhereProof225 customer messages mention thisHarmLoan lost on the day it was needed

What is going wrong underneath

1
The ruleThe app sends identity document checks (KYC, such as PAN) to the branch. More than 10 repledges a month needs a business head in 2025 messages; from January 2026, 9 messages describe a limit of 6.
44 messages
2
The handoverThe branch often cannot approve it and passes it to head office.
81 messages
3
The warningCustomers find out only when blocked. None of the six gold loan pages we checked mentions a repledge limit.
4
The paymentEven a successful top up can fail to reach the bank account.
225 messages
The harm: A loan the customer asked for, on the day they needed it. 149 messages describe contacting more than once; 12 say they will move lender. 951 messages. 72 mention an emergency, medical need or urgency.

I am not able to take top up loan through App as it asks Update PAN. Branch staff unable to resolve the issue when i visited branch

it's showing this is your 11 th repledge in this month approval needed from business head and operational head

I have repladge 6 times in this month, I cant repladge anymore in the app showing

The app messages customers quote, and what we have not seen

Customers quote the app as (spelling tidied): "Transaction not possible without complete KYC", "OVD update", "Update your PAN and photo", "Please contact nearest branch", "Technical error, contact branch manager", "This is your 11th repledge in this month, approval needed from business head". We have not seen the screen ourselves. Reaching it needs a customer account with gold already pledged, so there is no screenshot; we have asked MFL's app team for them.

What the website and app can change: branch staffing and head office approvals are outside this programme. The Gold Loan page promise (Leak 5), explaining the limit and documents in advance (Leak 8), and app messages that name the branch, document and time (Leak 8) are within reach.

Gold Loan complaints each month, with the two biggest problems highlighted0250500Jan2025FebMarAprMayJunJulAugSepOctNovDecJan2026FebMarAprMayJunJulAugSep

All Gold Loan complaints Borrowing more Money not reaching the account. Shaded: 2,000 complaints missing from the file, 14 November 2025 to 11 April 2026.

Journey B: paying interest or repaying online

The customer pays through the app, UPI or net banking to avoid a branch visit.

What the customer goes through
My gold loans
Pay interest
1. Pays in the app or by bank transferProof345 customer messages mention this
→
Bank account
- Rs 236
2. Money leaves their accountProof739 customer messages mention this
→
My gold loans
Pay interest
Interest due, unpaid
Lost here
3. Loan still shows unpaid, interest keeps addingProof719 customer messages mention thisHarmTrust in paying online is lost
→
My gold loans
Pay interest
Late fee added
4. Charged a late feeProof68 customer messages mention this
→
✆
Customer care: cannot trace
5. Nobody can trace the paymentProof43 customer messages mention this
→
Calls again and again
6. Contacts again and againProof196 customer messages mention this

What is going wrong underneath

1
The payment linkPayments leave the account without showing on the loan.
719 messages
2
The systemBranch systems described as down or slow.
43 messages
3
The fixNobody reachable can trace it quickly, so penalties keep running.
68 messages
The harm: Customers stop trusting online payment and go back to the branch. Payments made on time show as late. 719 messages; 196 describe contacting more than once.

I have paid my interest through ogl app and the amount was deducted from my account but it's hasn't shown paid of interest....and the interest is getting increased day by day

I have pay interest rs. 236 but your system not show

Journey C: closing the loan and getting the gold back

The customer has repaid in full and comes to collect their gold.

What the customer goes through
My gold loans
Close loan
1. Pays the full amountProof526 customer messages mention this
→
Collect gold
2. Goes to collect the goldProof94 customer messages mention this
→
Server down, come tomorrow
Lost here
3. Told the server is down, come tomorrowProof41 customer messages mention thisHarmRBI requires release within 7 working days
→
Gold returned late
4. Gold returned days later, sometimes damagedProof34 customer messages mention this
→
Credit report: loan still open
Lost here
5. Credit report still shows the loan openProof179 customer messages mention thisHarmThe next loan is lost

What is going wrong underneath

1
SystemsRelease stops when the branch system is down.
41 messages
2
Branch capacityLong waits.
94 messages
3
HandlingDamaged or short gold reported at release.
34 messages
4
ReportingClosed loans not updated with credit bureaus.
179 messages
The harm: The next loan: someone who struggled to get their gold back rarely pledges again. RBI 2025 directions: release on the day of repayment, within 7 working days at most (para 35), Rs 5,000 a day after that (para 46), lender pays for damage (para 44). The records do not show how long each delay lasted. 526 messages about getting gold back, 179 about credit reports.

I visited your branch today to release my pledged gold, but was unable to complete the process as I was told that the server is busy.

Iam closed my loan showing active in cibil pls update cibil bearue immedeltly very need

Journey D: updating KYC, bank account or mobile number

The customer needs to change a detail, often because the app asks for it.

What the customer goes through
My gold loans
Upload PAN
1. Tries to upload PAN or change bank detailsProof939 customer messages mention this
→
My gold loans
Upload PAN
Branch must approve
Lost here
2. App says the branch must approve itProof182 customer messages mention thisHarmA task the app offers cannot be finished in the app
→
Branch
3. Visits the branchProof68 customer messages mention this
→
Head office will do it
4. Branch says head office will do itProof63 customer messages mention this
→
My gold loans
Top up
Blocked
Lost here
5. Nothing changes; blocked from top upsProof182 customer messages mention thisHarmTop ups lost

What is going wrong underneath

1
The appTakes the document, then says the branch must approve it.
2
ApprovalBranches wait on head office, or send the customer away.
131 messages
3
Follow upNobody answers or calls back.
64 messages
4
The knock onThe pending detail blocks borrowing more.
182 messages
The harm: The same lost top ups as Journey A, plus several branch visits for one change. 939 messages; 160 describe contacting more than once.

I want to change my bank account details.i visited branch 3 times but no use

I need top up on my loan last time I already submitted my pan card bank details all KYC but every time when I approached for top up loan need to visit branch again for KYC submission

Journey E: a first time borrower

Someone not yet a customer, often coming from the website or an ad.

What the customer goes through
Low rate, transfer
Learn more
1. Sees an ad: low rate, loan transfer, instant loanProof20 messages mention an advertised rate
→
Branch
2. Enquires or walks into a branchProof165 customer messages mention this
→
Waits
3. WaitsProof419 customer messages mention this
→
Higher rate, or no transfers
Lost here
4. Offered a higher rate, a lower amount, or told transfers are not doneProof298 customer messages mention thisHarmA new customer lost
→
←
Goes to another lender
5. Leaves, sometimes to another lenderProof82 customer messages mention this

What is going wrong underneath

1
The promiseThe 9.90% rate needs Rs 5 lakh or more; the campaign page promotes transfers; one visitor got less than the calculator showed.
20 messages
2
The branchBranches refusing transfers or not explaining schemes.
40 messages
3
MeasurementNobody records which ad brought each enquiry, so nobody sees which promise brings people who then leave.
The harm: New customers who were ready to borrow. Only 165 messages came from non customers, so most lost first time borrowers never write in; this is the visible edge. 165 messages from non customers.

I saw your ad about transfer existing loan but they denied about this helpfull service

In your gold loan calculator it's showing today's loan to value is Rs ... But when I visited they are giving Rs ...

Rate ishu es liye apka customer nahi ban paye

Journey F: getting help when something goes wrong

Every journey above ends here.

What the customer goes through
✆
Calling branch
1. Calls the branch, manager or customer careProof334 customer messages mention this
→
✆
No answer
Lost here
2. No answer, or told to call backProof334 customer messages mention thisHarmThe problem grows while nobody responds
→
Enquire Now
Mobile number
Pincode
Overrides DNC / NDNC
Submit complaint
3. Writes a complaint on the websiteProof2,697 gold loan complaints in the records
→
No reply
Lost here
4. Hears nothing backProof334 customer messages mention thisHarmThe customer complains again or leaves
→
←
Warns others
5. Complains again, or gives up and warns othersProof669 customer messages mention thisHarmA fixable problem becomes a lost customer

What is going wrong underneath

1
ReachabilityPhones not answered, call backs not made.
334 messages
2
The complaint form224 people sent the same complaint two or more times, which usually means no clear sign it was received. We did not send a test complaint.
3
ConsentThe form will not send unless the customer gives up Do Not Call protection.
The harm: A fixable problem becomes a lost customer and a warning to family and friends. 82 messages say they will move or not recommend Manappuram. 669 messages describe contacting more than once or waiting with no response.

I am trying for Topup loan from last five days But unable to do, I am intimate this issue to branch and customer care but still not resolve

My relationship Manager not pickup my call

5B What else the messages show

The website's own forms confuse people

  • 1,088 feedback messages rated "Excellent" or "Very Good"; about half (589) describe a problem. Of 20 read at random, 17 were real problems.
  • 132 forms have an amount like 50000 in the Loan Number box. 91 messages contain only an email address.
  • 224 people sent two or more complaints (518 in all).

Many do not write in English

Devanagari63
Telugu10
Tamil8
Kannada7
Malayalam7
Gujarati5
Bengali0
Odia0
Gurmukhi0
Hindi typed in English letters251

The forms and the Gold Loan page are English only.

Customers compare, and some leave

Muthoot is named in 38 messages, banks such as SBI in 47. 49 say they will move their loan or not recommend Manappuram.

I have visited branch then they said that we don't transfer loans from other banks or nbfcs but in your ad you saying we transfer

February 2025

The advertised rate is not what small borrowers get

20 messages mention an advertised rate such as 9.90% or 82 paise against what they were charged.

Branch staff told me the rate of intrest as I saw on Internet will 82 paise ... they involve me n my gold in 15% rate of intrest plan

April 2026
6

Why none of this showed up in the reports

The analytics setup counts the wrong things and misses the right one.

The website reports are the tool that should tell Manappuram which pages, ads and messages bring enquiries. Today they cannot, because of how they were set up over the years. In plain words:

They miss the one thing that matters

They do not count Gold Loan enquiries at all. Manappuram received 8,692 from January to August 2026; the reports show none.

They count the wrong things

Simply opening the website is counted as an achievement, so the reports say 99.74% of visits achieve something. Some campaign leads are counted twice when a page is reloaded.

They cannot say who sent a visitor

37% of visits arrive with no record of which ad, message or site sent them. Facebook ads are recorded as free traffic.

They include visits that are not customers

Test websites and staff visits are counted alongside real customers.

6A The same four claims, with the numbers behind them

Nothing above is an opinion about the setup. Each claim is one of these four measurements, all from Google Analytics property 293901177 for the twelve months to 31 August 2026 unless a row says otherwise.

1. The reports count no Gold Loan enquiries at all

What Manappuram received, Jan to Aug 20268,692
What the reports counted0

The second bar is zero, so it does not draw. No event records a sent Gold Loan form. The only enquiries the reports see are campaign thank you pages.

2. Almost every visit is counted as a success

Visits the reports call an achievement99.74%
Visits that sent a Gold Loan enquiry, on the page itself0.77%

Arriving on the site is itself counted as an achievement, so the rate is near 100% and tells nobody anything. Landing page report, Total row, last 28 days to 16 September 2026, against thank you page views from Gold Loan page visits over twelve months.

3. More than a third of visits have no source, and Facebook ads are filed as free traffic

Visits with a source recorded1,127,931
Visits with no source recorded674,881
Paid social, as the reports show it6,692
Visits from Facebook and Instagram ad style links397,751

Google only treats a link as paid when it is labelled cpc, paid or retargeting. Manappuram's links say "targeted", "broad-targeted" and "remarketing", so the ads they brought are filed as free traffic. Marketing can confirm the real figure against the Meta ad account.

4. Some of the traffic is not customers, and the device split shows it

Contact page99.8% on phones
Scheme rates page99.4% on phones
English campaign page96.3% on phones
Gold Loan page78.9% on phones
Branch network28.0% on phones
Branch locator24.4% on phones

Customers use this site on a phone: every customer facing page runs between 78.9 and 99.8 per cent. The two branch pages invert that. People looking up branches on a computer, in office hours, are far more likely to be Manappuram's own staff than borrowers, and the filter that would separate them is switched off. That is 577,212 visits a year whose make up nobody can currently split.

The harm: Rs 3.52 crore was spent on Google ads in a year, plus Facebook ads we cannot see, without knowing which ones bring enquiries. No change to the website can be proved to work.

The fix is a clean setup with one agreed rule book, described for the analytics team below.

Who set it up: the reports keep a log of every settings change. Of 59 changes in two years, 54 were made by one agency account (sfcclients), 3 by Manappuram's digital team, 1 by another agency (iprospectsouth) and 1 by Zyxware (a scroll measure, February 2025).

6B What exactly is wrong in the Google Analytics setup

Each card shows what the reports say, what is really true, and the harm.

1. The most important action is not tracked

Reports say

0 Gold Loan enquiries

What is true

8,692 enquiries, Jan to Aug 2026

  • No event records a sent enquiry form except campaign thank you pages. An event is the name Google Analytics gives an action it counts; a key event is one the business has marked as a result that matters.
  • At the limit of 30 custom key events. A "purchase" key event exists on a site that sells nothing.
Harm: nobody can tell if a change works, what the real enquiry rate is, or which ads bring enquiries.

2. Lead events are duplicated and overcounted

Reports say

81,392 English campaign leads

What is true

52,800 people. Reloads count again.

  • 13 thank you events, one per language, created 5 and 8 December 2025. Totals need 13 numbers added by hand.
  • Two "thank you" events fire mostly on the branch locator page.
  • No naming rule: Gold_Mortgage_Loan_Interest_Rates, applygold_calculator_click, bs-programmatic-pageview, and two spellings of april_doorstep_lp3_thankyou.
  • On 9 December 2025 key event settings were changed 18 times; next day "engaged" visits fell from near 100% to about half. Likely linked, not proven.
Harm: cost per lead looks lower than it is, and branch page visits can be added up as leads.

3. Paid ads reported as free traffic

Reports say

Paid social: 6,692 visits

What is likely true

397,751 visits from Facebook and Instagram ad style links

Google only reads words like cpc, paid or retargeting as paid. Manappuram's links say "targeted", "broad-targeted", "remarketing". Marketing can confirm against the Meta ad account.

Harm: ad money judged on the wrong number; unpaid posts get credit for what ads brought.

4. Campaign labels broken or missing

  • No source: 616,817 visits (34%), including almost all rates page visits.
  • Could not be grouped: 58,064, e.g. "(not set) / targeted" (4,307).
  • Same thing, many spellings: goldloan / Goldloan, facebook / fb / m.facebook.com.
  • Affiliate placeholders never filled: 21,854 visits named like "iggl-en_5424_{aff_sub}_icubeswire".
  • Meaningless labels: "aggregattor", QR_Code / search (3,172), campaign IDs like 23928751102.
Harm: more than a third of visits cannot be credited; affiliate claims cannot be checked; reports need hand cleaning.

5. Traffic that is not customers

  • 28,419 visits in Jan to Aug 2026 from other addresses: stage.manappuram.com (24,388), a developer copy (mafl-revamp.lndo.site), localhost, and unrelated sites such as www.zanbank.co.nz.
  • The internal traffic filter is set to "Testing", so it removes nothing.
  • 25,802 visits have no starting page recorded.
  • A second, empty property invites people to read the wrong one.
Harm: 0.6% of visits are test traffic under real page names; staff visits count as customers.

6. Ads and tags send data to the wrong places

  • Ads pointed at the wrong language page. From April to June 2026, Facebook ads meant for Maharashtra, Telangana and Uttar Pradesh sent visitors to the Malayalam campaign page. Across the language pages, 28,215 visits came from ads meant for another state or language (chapter 2, section 2D).
  • Every language page reports itself as English. The saved campaign pages carry the campaign name "instant-state-wise-en" in the form's hidden fields, whatever the language, and some carry an earlier visitor's ad click code. If enquiry records take their source from these, results by language and by ad are wrong. Needs checking in the enquiry records.
  • Staff and customers mixed. The branch locator pages are mostly visited on computers, which suggests staff use, but the filter that would separate staff is switched off.
Harm: ad and language results can be credited to the wrong campaign, and staff visits distort the trends Manappuram uses to plan.
For the analytics team: what a correct setup looks like

6C What a correct setup looks like

Events to track

EventWhenKey event
generate_leadSystem confirms an enquiry was saved. Attach form, product, page, branch, language, methodYes
appointment_bookedA branch visit is booked. Branch, date, productYes
form_startFirst tap in any enquiry formNo
otp_requested, otp_verifiedSend OTP tapped; code acceptedNo
form_errorA field is rejected. Field and reasonNo
branch_lookupBranch list returned for a pincodeNo
whatsapp_click, call_clickButton tapped. Page and positionNo
calculator_usedEstimate shown. Type, amount bandNo
language_switchVisitor changes languageNo

Google's recommended names, lowercase with underscores, built in Tag Manager from the form's own confirmation, not from page views. Register form, product, method, language and branch as custom dimensions.

Campaign link rules, for every agency

LabelRuleExample
utm_sourceWhere the link was placedgoogle, facebook, whatsapp, sms, qr
utm_mediumFixed list Google understandscpc, paid_social, display, affiliate, sms, whatsapp, email, qr
utm_campaignproduct_goal_language_yearmonthgoldloan_leads_te_202610
utm_contentWhich ad or messagevideo_30s_v2
utm_termSearch keyword, from the ad platformgold loan near me

Account hygiene

  • One property; close the empty one.
  • Keep only www.manappuram.com; switch the internal filter to Active.
  • Remove tracking from test sites.
  • Keep Google Ads auto tagging on; set Meta link parameters at account level; check affiliate links before launch.
  • Custom channel groups for SMS, WhatsApp and affiliates.
  • Link BigQuery (free tier) to keep data beyond 14 months.
  • Check online.manappuram.com payments are measured.
  • A written tracking plan reviewed before every campaign.
7

Bigger ideas worth testing

Beyond fixing what is broken.

These go further than fixing what is broken. Each uses something Manappuram already has, or removes a step.

A. Missed call instead of OTP

Offer "Give a missed call and your branch calls you back" next to the form. The call proves the number, like an OTP does.

76 gold loan messages mention OTP trouble, e.g. "otp not coming for mobile number, 10-15 time try".

Today
  1. Lost hereTypes number, waits for SMS, types code, ticks consent
With the idea
  1. Enquiry sentTaps "Give a missed call". A branch calls back

Measure: enquiries by missed call each month, and branch visits from them.

Is it safe?
RBI rulesSafe, if the call back says who is calling and why, and consent is recorded then.
SEBI rulesSafe. SEBI covers investor information, not loan pages.
Dark patternsSafe. The customer starts the contact.
Personal dataRecord consent and purpose when calling back.

B. Book a valuation slot

Let the visitor pick a branch and time, with a WhatsApp confirmation, map link and "bring Aadhaar and your gold".

419 messages are about waiting at branches. Insider, a messaging tool, already loads on the page.

Today
  1. Lost hereSends an enquiry, waits for a call, waits an hour at the branch
With the idea
  1. Enquiry sentBooks Saturday 11 am, gets a WhatsApp with the map, seen on time

Measure: bookings made and bookings attended.

Is it safe?
RBI rulesSafe.
SEBI rulesSafe. SEBI covers investor information, not loan pages.
Dark patternsSafe, if no false scarcity. Never "only 2 slots left" unless true.
Personal dataMessages only about the booking; offers need separate consent.

C. Answer "how much will I get" first

Three questions at the top: grams, carat, pincode. "About Rs X today. Final amount after your gold is checked at the branch."

One visitor wrote that the branch valued their gold lower than the calculator, so the condition must be shown with the number.

Today
  1. Lost hereNot sure her bangles are worth Rs 40,000. Enquires anyway, or not
With the idea
  1. Enquiry sentEnters 20 grams, 22 carat, sees about how much, books a slot

Measure: estimates run, and enquiries from people who ran one.

Is it safe?
RBI rulesSafe if the estimate uses the loan to value band for the amount and states branch valuation decides.
SEBI rulesSafe. SEBI covers investor information, not loan pages.
Dark patternsSafe. Estimate and conditions shown together.

D. Speak the visitor's language

Offer "Read this in Telugu / Tamil / Kannada / Marathi / Hindi" on the Gold Loan page, suggested by location. The campaign page already has 13 languages.

351 customer messages are written in an Indian language.

Today
  1. Lost hereReads the English page slowly, leaves it for later
With the idea
  1. Enquiry sentTaps Telugu, reads the same page, enquires

Measure: enquiry rate for visitors who switch language.

Is it safe?
RBI rulesSafer than today. RBI 2025 directions ask for communication in the language of the region or of the borrower's choice.
SEBI rulesSafe. SEBI covers investor information, not loan pages.
Dark patternsSafe, if the visitor chooses and can switch back.

E. Make moving a loan to Manappuram a first choice

Rupeek opens with "New Gold Loan" and "Shift Your Loan". Manappuram's transfer page sits low in a list of links. Give it a top button with a monthly saving comparison, once branches confirm they handle transfers.

Customer records show people being turned away when they tried.

Today
  1. Lost hereNeighbour with another lender finds no clear option, stays
With the idea
  1. Enquiry sentTaps "Move my gold loan", sees his saving, books a visit

Measure: transfer enquiries and completed transfers.

Is it safe?
RBI rulesSafe if it uses the customer's own numbers and the full rate range.
SEBI rulesSafe. SEBI covers investor information, not loan pages.
Dark patternsSafe. Never compare against a competitor's highest rate.

F. Watch real visits in Microsoft Clarity

Microsoft Clarity, a free tool that shows how people use a page, is already on the Gold Loan page. It shows how far phone visitors scroll and where they stop in the form, with no new software.

Decisions on the layout in Leak 3 can then be made on real behaviour.

Today
  1. Lost hereLayout decided by opinion
With the idea
  1. Enquiry sentLayout decided by what visitors actually do

Measure: scroll depth and form drop off before and after each change.

Is it safe?
RBI rulesSafe.
SEBI rulesSafe. SEBI covers investor information, not loan pages.
Dark patternsSafe.
Personal dataConfirm Clarity masks everything typed into forms.
8

How we will prove each fix worked

Every change gets a number. Some numbers need data only Manappuram has.

8A How the results will be proved

MeasureWhat it meansToday
Gold Loan page enquiry rateEnquiries from the page divided by visits that opened it7.0%, June to August 2026, counting all enquiries
Rates page enquiriesEnquiries sent from the scheme rates pageNo way to enquire
Language page enquiry rateAd visits that send an enquiry, per languageTamil, Telugu, Malayalam, Kannada 5.6%; other regional 11.0% (December 2025 to 28 September 2026)
Form completionEnquiries sent divided by forms startedNot measured yet
WhatsApp and call tapsTaps per buttonNot measured; WhatsApp link broken
Top ups blocked by document checksFrom the app87 complaints describe it

Wherever possible, show a change to half of visitors and the old page to the other half in the same weeks. That separates the change from seasons, festivals and ad spend.

8B What we need from Manappuram

  1. Needed first
    Enquiry records split by the page they came fromEvery enquiry already carries the page address. Gives the exact enquiry rate per page.
  2. Needed first
    Where the rates page visits come fromWhich ad, post, SMS or app links to it.
  3. Needed first
    Google and Meta advertising dataAsked for, not received. Needed to explain the language page gap.
  4. Needed first
    The ad creatives behind the campaign labelsEspecially ads labelled 82 paise and low interest, to check whether the Rs 5 lakh condition was shown.
  5. Needed first
    OTP sent and delivered counts, and form error recordsAsked for, not received.
  6. Needed soon
    Google Ads report of form startsThe Gold Loan page already sends a "started typing" signal to Google Ads. That count shows how many people start the form today.
  7. Needed soon
    Enquiries split by phone and computerAsked for, not received.
  8. Received
    Access to Microsoft ClarityReceived. Chapter 12 is what it showed.
  9. Needed soon
    The complete complaint records2,000 entries missing, numbered 1,839 to 3,838, 14 November 2025 to 11 April 2026.
  10. Needed soon
    The complete feedback recordsThe file stops at exactly 5,000 entries, on 2 August 2026.
  11. Needed soon
    Screenshots of the app messages that block a top upKYC, OVD, PAN and repledge limit screens.
  12. Needed soon
    What changed in online top ups, July to October 2025And whether the repledge limit changed from 10 to 6 in 2026.
  13. Helpful
    SMS, WhatsApp and app messages sent July to September 2025With dates and audiences, to explain the visit spike.
  14. Helpful
    January 2025 enquiriesOne record shows 372 and another 752.
  15. Helpful
    Sign off from product and legalOn rates, loan to value and charges wording before anything goes live.
  16. Helpful
    Branch operations confirmationThat loan transfers are handled, before Idea E.
9

Where more enquiries will come from

Every leak in this report, ranked by how many people it touches and how directly it stops an enquiry.

The largest gains are in visitors Manappuram already gets, not in buying more. Tap any row to jump to the evidence.

1No enquiry is countedThe foundation: without it no fix can be proved8,692 enquiries unseen (Jan to Aug 2026)
Big leakAnalytics setup
2Branch pages lead nowhereNo form, no button, phone numbers cannot be tapped577,212 visits in 12 months
Big leakWebsite team
3Language pages ask for the details in EnglishTamil phone box is labelled "Required amount"470,178 visits since December 2025
Big leakWebsite team
4Rates page has no way to enquireBiggest single page of interested visitors208,385 visits in 3 months
Big leakWebsite team
5South Indian language campaign pages convert at halfOver nine months, and with the same ad label202,812 visits since December 2025
Big leakMarketing team
6Form is 10 phone screens downGold Loan page, most visitors on phones49,932 visits opened the page
Big leakWebsite team
7No price on the Gold Loan pagePeople leave to find it, some feel misled later49,932 visits opened the page
Big leakWebsite team
8Existing customers blocked from borrowing moreThe easiest loans to give951 customer messages
Big leakApp and product team
9The form offers far away branches for some big citiesMumbai, Delhi, Lucknow, Bhubaneswar; Kohima gets none2 of 19 pincodes tested
Big leakWebsite team
10Charges, FAQ, terms and articles have no way to enquirePeople decide there, and cannot act161,300 visits in 12 months
Medium leakWebsite team
11"Apply Online" promise a new borrower cannot getBroken promise at first contact49,932 visits opened the page
Medium leakWebsite team
12WhatsApp opens an unknown number, no call buttonCloses the route many prefer49,932 visits opened the page
Medium leakWebsite team
13Form fights the phone and makes people give up Do Not Call protectionThe last step before sending49,932 visits opened the page
Medium leakWebsite team
14Ads sent to the wrong language pageMalayalam page, April to June 2026; stopped in July, but nothing checks it28,215 visits since December 2025
Medium leakMarketing team
15"Doorstep enquiry" opens the home pageA promise on every page that leads nowhere2,514 visits in 12 months
Medium leakWebsite team
16Four pages open scrolled past the Apply buttonVisitors start mid page2,540 visits in 12 months
Smaller leakWebsite team
17Gold Loan page enquiries credited to adsWrong budget decisions282 enquiries misattributed
Smaller leakAnalytics setup
18Claims the rest of the site contradictsTrust breaks at the branch2 contradictions
Smaller leakWebsite team

9A What the leaks add up to in money

Rs 3.52 croreMoney spentGoogle ads in the year to August 2026. Facebook and Instagram are not linked to the reports, so the real figure is higher.
Rs 1.12 croreMoney lostRs 10.5 lakh of it is cash already burnt. The rest, Rs 1.02 crore, is 37,263 enquiries a year the site does not send, priced at what Manappuram pays to buy one.

An enquiry is priced at Rs 273, Manappuram's own average cost of buying one through Google ads in the year to August 2026. What an enquiry is worth in loans and interest needs branch conversion and average ticket size, which we asked for and have not received, so every figure here is a floor, not a forecast. The counting gap in Leak 1 means today's enquiry numbers are understated.

What one lost enquiry is worth in loans

Every rupee in this chapter is also shown as loans not written. Two of the four numbers below are Manappuram's own published figures. Two are assumptions, because Manappuram has not shared them. Change them and every figure in this chapter changes with them.

Rs 1.3 lakhlent out, per loan
Rs 11,046interest earned on it, gross
Rs 5,427after the 8.8% cost of borrowing
Rs 2,762what one enquiry is worth, at the conversion rate above

Sources: Manappuram Finance Investor Presentation Q4 FY2026, average loan ticket size Rs 127,700, gold loan yield 17.3% and cost of borrowing 8.8%. The conversion rate and the holding period are ours, clearly marked, and both are adjustable above. For comparison, buying one enquiry through Google ads costs Manappuram Rs 273, which is the floor used everywhere else in this report.

9B Leak by leak, in loans

LeakEnquiries a yearWhat buying them costsWhat they would have been worth in loans
The rates page has no way to enquire8,295Rs 22.7 lakh2,074 loans, Rs 26.48 crore lent, Rs 2.29 crore interest
The language pages convert below the English one20,182Rs 55.2 lakh5,046 loans, Rs 64.44 crore lent, Rs 5.57 crore interest
The branch pages cannot take an enquiry5,772Rs 15.8 lakh1,443 loans, Rs 18.43 crore lent, Rs 1.59 crore interest
Charges, terms, help and articles have no route1,613Rs 4.4 lakh403 loans, Rs 5.15 crore lent, Rs 44.5 lakh interest
The form sits nine screens down on a phone1,166Rs 3.2 lakh292 loans, Rs 3.73 crore lent, Rs 32.3 lakh interest
The branch list is wrong for five big cities210Rs 57,31653 loans, Rs 67.7 lakh lent, Rs 5.9 lakh interest
The doorstep link opens the home page25Rs 6,8746 loans, Rs 7.7 lakh lent, Rs 66,276 interest
All of them together37,263Rs 1.02 crore9,317 loans, Rs 118.98 crore lent, Rs 10.29 crore interest

9C The same loss, in loans

The figures above price a lost enquiry at Rs 273, what Manappuram pays to buy one through Google ads. That is the floor. What the business actually loses is the loan that enquiry would have become. Move the two assumptions above and this moves with them.

In loans9,316 loans a year, Rs 118.96 crore lent out, Rs 10.29 crore of interest earned, Rs 5.06 crore of it after the cost of borrowing.

Based on 37,263 enquiries a year at a cautious 1 in 100, Manappuram's published average gold loan of Rs 127,700 and gold loan yield of 17.3% a year (Investor Presentation Q4 FY2026), and two assumptions we have marked: the share of enquiries that become a loan, and the months a loan is held. Nothing here counts the enquiries the ad chapters describe, so it is not double counted.

Do first

  • Count every enquiry (Leak 1), so each fix below can be proved.
  • Fix the WhatsApp link (Leak 6). One line, works the same day.
  • Give the rates page a way to enquire (Leak 2), the biggest single page.

Then

  • Short enquiry on the first phone screen, with price and honest heading (Leaks 3, 4, 5).
  • Find out why the Tamil, Telugu, Malayalam and Kannada pages convert at half (chapter 2).
  • Lighter form, no forced Do Not Call override (Leak 7).
  • Warn existing customers before a top up is blocked (Leak 8).

Repeat what already worked

August 2025 had the most Gold Loan enquiries in the records, 3,455, when bulk message bursts sent visits to the page. Find out what was sent and send it again on purpose, with a tracking tag on every link (chapter 10).

We cannot honestly promise a multiple of today's enquiries until the counting in Leak 1 is in place. Numbers shown for scale are illustrations, not forecasts.

10

All the numbers behind this report

For anyone who wants to check our working.

10A Gold Loan enquiries, month by month

Jan to Aug 2026: 8,692 enquiries and 147,807 visits that opened the Gold Loan page, so about 5.9%. Counting only visits that started on the page, 11.4%. The true rate is between the two, because enquiries and visits are counted in different systems.

Monthly visits landing on the Gold Loan page and monthly Gold Loan enquiries, January 2025 to August 20260k15k30k45k01,0002,0003,0004,000Jan2025FebMarAprMayJunJulAugSepOctNovDecJan2026FebMarAprMayJunJulAug

Visits that started on the Gold Loan page Gold Loan enquiries

Month by month table
MonthStarted on the pageOpened the pageEnquiriesEnquiry rate
Jan 2025 *8,92517,8483722.1%
Feb 20256,16014,4874012.8%
Mar 202520,72630,8074781.6%
Apr 20259,66120,6737143.5%
May 20256,91017,9507354.1%
Jun 202511,73721,6831,3816.4%
Jul 202521,37733,1251,9445.9%
Aug 202525,68538,1983,4559.0%
Sep 202539,95750,7882,5485.0%
Oct 202514,22824,5299744.0%
Nov 202510,09417,2848094.7%
Dec 202510,10819,5769034.6%
Jan 202611,24021,4969994.6%
Feb 20269,09417,8219835.5%
Mar 20268,46818,9079354.9%
Apr 202610,80020,3241,0495.2%
May 20269,71818,9091,2256.5%
Jun 20267,67915,6851,2097.7%
Jul 20269,93617,9371,2907.2%
Aug 20269,08816,7281,0026.0%

* January 2025 is 372 in one record and 752 in another; 372 is used. Rate is enquiries divided by visits that opened the page. A visit that "started on the page" had the Gold Loan page as its first page; "opened the page" also includes visits that arrived from another page. About half reach it from another page.

10B Why visits and enquiries jumped from July to September 2025

Most likely: bulk SMS, WhatsApp or app messages with a link to the Gold Loan page. Search visits did not change; the whole rise came from visits with no source, arriving in single day bursts across many states, biggest on 1 August 2025.

Not certain: the links carried no tag, and we have not received the advertising data or a list of messages sent.

3,455enquiries in August 2025, the most in the records, more than double June

If messages caused it, sending again on purpose with a tagged link could repeat the lift. In September visits kept rising but enquiries fell, so who gets the message matters.

10C Where visitors to the Gold Loan page come from

Free search results (Google, Bing)21,197
No record of where they came from4,687
Search ads544
Free social media posts172
Could not be sorted159
AI tools such as ChatGPT88
Google ads across several places60
Links on other websites54

Visits that started on the Gold Loan page, June to August 2026. By how people first found Manappuram: 72% of those starting on the page and 55% of everyone opening it first came from search; 32% had no source, 8% came from search ads, 5% from everything else. The monthly review traffic matches these reports exactly for February to October 2025, the months it fills in.

What each source means
Organic SearchUnpaid Google, Bing and other search results
DirectTyped the address, used a bookmark, or tapped a link that carried no source, for example from WhatsApp, SMS or an app
Paid SearchPaid search ads, such as Google Ads
Organic SocialUnpaid posts on Facebook, Instagram and similar
UnassignedSome source details arrived, but not enough to put the visit in any group, usually because a link carried an incomplete or non-standard tracking tag
AI AssistantLinks from AI tools such as ChatGPT
Cross-networkGoogle ads that run across several Google networks at once
ReferralLinks on other websites

10D Why the 3.56% target baseline should not be used

The baseline

3.56% = 4,627 enquiries divided by 130,055 visitors

What the 130,055 is

People who viewed the page in Jan 2024 to Jan 2025 and first came from search. Everyone else (about 41%) left out.

It matches the reports within about 1% each month (January 2024: 7,269 in the review, 7,268 in the reports). The 191,134 total traffic figure could not be traced. Measured on all visits, the page is already between 5.9% and 11.4% in Jan to Aug 2026. Reset the target once enquiries are counted.

10E Engagement figures before 10 December 2025 cannot be trusted

Until 9 December 2025, 95% to 101% of visits were counted as engaged, which does not happen on a real site. On 10 December it fell to about 52% and stayed, the day after the agency account changed the reports' settings 18 times.

Share of visits counted as engaged, every day from 15 November to 31 December 20250%50%100%15 Nov01 Dec10 Dec15 Dec31 Dec

Up to 9 December 2025 From 10 December 2025

10F How this was checked

Website findings were measured on the live site on 17 September 2026 as a 360 pixel phone and a 1440 pixel computer. No enquiry was sent and no OTP requested. Figures come from the live Google Analytics property (G-N79NWH6G0D), read only; they can differ by one or two percent depending on how a report is split. Customer messages were sorted using word patterns in English, Hindi and common misspellings, then sampled and read; counts are close, not exact. Quotes are as written, with names, phone and account numbers removed. We did not test what happens after an enquiry is sent, the branch call back, or the Online Gold Loan app. Competitor pages we could load and measure: IIFL Finance and Rupeek. Muthoot Finance returned "405 method not allowed" and Bajaj Finance returned "403 forbidden" to our checks, so nothing in this report claims what is on their pages.

Where the enquiry numbers come from, and what we do not know about them. Every enquiry count in this report that is not a website report figure comes from one table in the file Manappuram sent us, headed "Gold Loan, Monthly Backend Sub". It gives a month and a count, nothing else: no page, no source, no form name. The other table in the same file covers only visits that started on the Gold Loan page, which suggests the counts belong to that page rather than to the whole site, but nothing in the file says so. We have used it as the number of Gold Loan enquiries received, because that is how it was given to us, and every rate built on it is marked where it appears. Before these figures are used to make a decision, somebody at Manappuram should confirm which forms that count covers. If it covers only one form, the leaks in this report are larger than we have said, not smaller.

12

What the session recordings add

A second reading of the same visitors, from the tool that watches the page itself.

12A What the session recordings add

Everything before this chapter comes from Google Analytics and from Manappuram's own records. Microsoft Clarity has been running on the site all along, and it records something neither of those can: how far people actually scroll, and what they tap. It is a second, independent reading of the same visitors.

Project ngcwpik1zw, the same tag that loads on the Gold Loan page. Window 1 February to 28 September 2026, read on 29 September 2026. Clarity leaves bots out: 600,516 sessions were excluded as bots over this window, which is part of why its totals sit below the Analytics ones.

4,132,985 sessions recorded

Across the whole site over the eight months, after bots were removed.

291,161 of them include a dead tap

A dead tap is a tap on something that is not a link or a button. Nothing happens.

12,348 include a rage tap

The same spot tapped over and over in a few seconds, which is what people do when a page stops answering them.

12B It confirms the biggest leak, using a different tool

Chapter 4 says the enquiry form starts at 48% of the page height, and that the website reports show 4.6% of phone visits reaching the bottom. Clarity measures the same page a different way, by recording how far each visitor scrolled.

Reached 10% of the page65.3%
Reached 25%35.4%
Reached 48%, where the form starts16.1%
Reached 75%5.7%
Reached 90%3.5%
Reached the very bottom1.2%

Share of sessions on the Gold Loan page that scrolled at least this far. Microsoft Clarity scroll map, 98,943 sessions, 1 February to 28 September 2026.

The two tools agree

Clarity puts 3.5% of sessions at 90% of the page. Google Analytics, whose scroll event fires at that same depth, puts phone visits at 4.6%. Different tools, different samples, the same answer.

It also answers what we could not

Chapter 4 says how many people stop at the form itself is not recorded. It is now: 16.1 in every 100 sessions get as far as the form. The other 84 never see it.

12C New: the page people reach after paying does not exist

This one sits outside the enquiry journey, and it is the largest single thing this review found.

202,017sessions landed on /app/success.php1 February to 28 September 2026
404is what that address returnsChecked live on 29 September 2026. The page is titled "Page not found"
2 secondsaverage active time on itPeople arrive, find nothing, and leave

Where they arrive from settles what the page is for. 87% of them, 175,717 sessions, arrive straight from secure.payu.in, a payment gateway. Nearly all the rest come from other gateways.

Arrived fromSessions
secure.payu.in175,717
merchant.now.hdfc.bank.in26,742
api.payu.in1,516
netportal.hdfcbank.com712
test.payu.in288
www.manappuram.com281
The harm: these are not people browsing. They are customers who have just paid money, and the screen they get says the page cannot be found. Nothing tells them whether the payment went through. Every one of them is a reason to ring the branch or the helpline.

Two smaller things in the same table. test.payu.in sent 288 live sessions, so a test gateway is pointing at the live site. And this address appears nowhere in the website reports we were given, which is how eight months of it went unnoticed.

What to change

Point the gateway return address at a page that exists and that states the outcome: paid or not paid, the amount, the reference number, and what happens next. Keep that address in the reports so it can be counted. Then take the test gateway off the live site.

12D New: people are tapping things that are not buttons

Clarity marks a tap on something that is not a link or a button as a dead tap. On the Gold Loan page it happens in more than one session in five.

PageSessions with a dead tapSessionsShare
Gold Loan page20,61193,37322.1%
Branch network page27,279142,45619.1%
Calculator14,91284,79317.6%
Contact page4,86330,10616.2%
Branch list38,905259,33615.0%
Home page74,652554,62113.5%
Rates page24,563441,8815.6%
The harm: each one is a person who expected something to happen and got nothing. It matches what chapters 4 and 6 already describe: a phone number printed as plain text, four floating icons with no words on them, and a rates table that looks tappable and is not.

What we can say is how often this happens and on which pages. Which element is being tapped needs the recordings watched one at a time, which we have not done. We tried to read it from the tap coordinates and found they do not line up with the page height Clarity reports, so that reading was dropped rather than published.

12E The screens these came off

So nobody has to take our word for it. These are the Clarity screens as they looked on 29 September 2026. The numbers on them move as the window rolls forward; where to look does not.

Clarity dashboard for the whole site. Sessions, scroll depth, dead taps and rage taps all sit on this one screen
Clarity dashboard for the whole site. Sessions, scroll depth, dead taps and rage taps all sit on this one screen
Heatmaps. Paste a page address here, or pick it from the list, to get the scroll band used in 12B
Heatmaps. Paste a page address here, or pick it from the list, to get the scroll band used in 12B
The same dashboard with the dead click filter applied, which is how the table in 12D was counted
The same dashboard with the dead click filter applied, which is how the table in 12D was counted

The dashboard card also carries two signals this report does not use: quick backs, where someone opens a page and goes straight back, and excessive scrolling. Both are worth watching once the fixes in chapter 8 are underway.

11

Where every number came from

Each figure paired with the report that produces it, so anyone can check it themselves.

11A Why the dates are not all the same

One window across the whole report would be tidier, and we could not use one. Three of the things measured here did not exist for the same period, two come from files whose span we did not choose, and some are a live check on a day rather than a period at all. Forcing one window would mean either dropping the newer pages or dividing their results by months in which they did not exist. Each figure carries its own dates for that reason, and here is why each window is what it is.

WindowWhat it coversWhy this one
29 Sep 2025 to 28 Sep 2026The backbone. Visits to every page, the year's ad spendA rolling twelve months ending the day before we read the data, so it is a full year and it is current. It was moved forward from the earlier window for two reasons: everything now reproduces on the day you read it, and the old window carried a traffic spike in September 2025 that made the Gold Loan page look busier than it is.
1 Feb to 28 Sep 2026Everything in chapter 12, from the session recordingsEight months, the longest window where Clarity holds solid data for this site. Its record thins out before roughly April, so we did not reach further back and pretend otherwise.
1 Jun to 31 Aug 2026Phone and computer split, scroll depth, movement from one page to the nextThe most recent complete quarter. These describe how the site behaves now. Averaged across a year they would blend in changes made to the pages during it.
Dec 2025 to 28 Sep 2026The 13 language campaign pagesForced twice over. Those pages took their first visit in December 2025, and the events that record their enquiries were created on 5 and 8 December 2025 (property change history). Before that there is no counting to read. A twelve month window would divide nine months of enquiries by twelve months of calendar and make the pages look worse than they are.
Nov 2025 to Aug 2026The English campaign pageSame reason. Its first traffic is November 2025, one month before the language pages. This is the whole life of the page.
Jan to Aug 2026The 8,692 enquiries set against what the reports countedChapter 6 compares what Manappuram received with what the website reports recorded, and that comparison runs on calendar 2026. The enquiry records hold 21 months; we stop at August, again because September was unfinished.
Jan 2025 to Sep 2026What customers wrote inThe span of the two files we were given, and they do not end together: the feedback file stops on 2 August 2026, the complaints run into September. We print the wider span rather than trim one to match the other.
Last 28 days, 20 Aug to 16 Sep 2026The list of events recorded, and the 99.74% achievement rateThis is how the property is set up today, not a trend, so the newest window is the right one. 28 days is the window Google Analytics itself uses for that view.
17 and 21 September 2026, all re-walked 29 SeptemberEverything measured on the live pages: form position, page weight, the 75 page census, the branch lookupNot a range. A page is whatever it is on the day you open it, so each of these carries the date we opened it. It matters: the Gold Loan page title changed between the morning and the afternoon of 17 September.
Q4 FY2026Average gold loan of Rs 127,700 and the 17.3% yieldManappuram's own published investor figures, latest published quarter. We do not choose this window; the company does.

One outer limit applies to all of them. Analytics keeps 14 months, so nothing before roughly August 2025 opens in the exploration reports. Clarity holds its own separate record, and for this site that thins out before about April 2026.

11B Numbers that come from Google Analytics

Property 293901177. A visit means a session. Open that property, choose the report named, set the dates, and you should land on the same number. Where a figure is a sum of rows or is worked out from two numbers, it says so, because no single cell in Google Analytics will show it.

NumberWhat it backsWhere in Google AnalyticsDates
829,532Rates page visits in a yearLanding page, sessions1 Sep 2025 to 31 Aug 2026
91Rates page visits that ended in an enquiry anywhereworked out, not one cellSessions from that landing page that reached a thank you page1 Sep 2025 to 31 Aug 2026
123,313Gold Loan page visits in a yearLanding page, sessions1 Sep 2025 to 31 Aug 2026
945Thank you page views from visits that started on the Gold Loan pageworked out, not one cellSessions from that landing page that reached a thank you page1 Sep 2025 to 31 Aug 2026
1,614,789English campaign page visitsLanding page, sessionsNov 2025 to Aug 2026, the life of the page
209,434Thank you page views on the English campaign pageIts own thank you pageNov 2025 to Aug 2026, the life of the page
577,212Branch page and branch locator visitsworked out, not one cellLanding page, sessions, two pages added together1 Sep 2025 to 31 Aug 2026
161,300Charges, terms, help, questions and article visitsworked out, not one cellA hand built list of pages added together, including 297 articles. No single report will show this total, and we did not keep the query behind it1 Sep 2025 to 31 Aug 2026
1,640 of 35,928Phone visits that reached the bottom of the Gold Loan pageworked out, not one cellTwo reads set against each other, both filtered to page /gold-loan and to phones: sessions with the scroll event, against sessions with a page view. The scroll event fires at 90% of page depth1 Jun to 31 Aug 2026
28,215Paid visits sent to a page in a language the ad was not meant forworked out, not one cellCampaign and ad content against landing pageDec 2025 to 28 Sep 2026
210,781Visits from ads promising interest of 82 paiseworked out, not one cellAd contentDec 2025 to 28 Sep 2026
48,332 and 12,042Paid taps on the banner line, against visits recorded from themworked out, not one cellGoogle Ads clicks, against Google Analytics sessions for the same campaigns1 Sep 2025 to 31 Aug 2026
Rs 3.52 croreGoogle Ads spend in the yearGoogle Ads campaigns, Cost1 Sep 2025 to 31 Aug 2026
99.74%Share of visits counted as achieving something importantLanding page report, Total row, session key event rateLast 28 days, 20 Aug to 16 Sep 2026
80%Share of all site visits on phones (1,436,146 of 1,800,363)worked out, not one cellDevice category, sessions. Read the two numbers and divide1 Jun to 31 Aug 2026
69%Share of Gold Loan page visits on phones (18,454 of 26,871)worked out, not one cellDevice category, landing page /gold-loan. Read the two numbers and divide1 Jun to 31 Aug 2026

11C Numbers that do not come from Google Analytics

No Google Analytics screenshot can show any of these, because Google Analytics never held them. That is the whole of Leak 1.

NumberWhat it backsWhere it comes fromDates
11,941Gold loan enquiries receivedManappuram's own enquiry records. The file gives a month and a count, with no page, source or form name1 Sep 2025 to 31 Aug 2026
8,692Gold loan enquiries receivedManappuram's own enquiry recordsJan to Aug 2026
6,622Gold loan customer messages readManappuram's complaint and feedback records. The feedback file stops on 2 Aug 2026; the complaints run to SeptemberJan 2025 to Sep 2026
8,595 pxHow far down the Gold Loan page the enquiry form startsMeasured on the live page, 360 by 800 phone29 Sep 2026, re-walked
4.6 s, 3.5 MB, 19.9 sSpeed and weight of the Gold Loan pageMeasured live at the speed Google's own tool uses for a mid range phone17 Sep 2026
2 of 19City pincodes returning a wrong branch list or noneEach pincode sent to the form's own branch lookup17 and 21 Sep 2026
60Inner pages carrying the Doorstep enquiry link that opens the home pageChecked live, page by page17 Sep 2026
127,700Average gold loan, used to turn enquiries into lendingManappuram's own published investor figures, Q4 FY2026Q4 FY2026

11D Four things to know before checking these

Every rupee figure rests on one assumption

The cost field Google Analytics returns is recorded in its own notes as "likely micro units", and we divide by a million on that basis. The size looks right: the alternative would be a yearly ad spend larger than Manappuram's whole loan book. Confirming the Google Ads cost total in rupees settles it for good.

The internal traffic filter is switched off

It is set to "Testing", so it removes nothing. Every visit figure here includes Manappuram's own staff and test traffic, which is one reason the branch pages look busier than they are. Switching it on would not change any figure here, because the filter only applies to data collected after it is switched on. It would mean a later comparison is not like for like.

Older figures may not open in every report

The property keeps 14 months of data. Anything before roughly July 2025 may not appear in the exploration reports, though the standard reports are built differently and often still show it.

Two counters, never mixed

Manappuram's enquiry records and Google Analytics thank you page views count different things, and nothing in this chapter divides one by the other. Two figures earlier in this report do: the 7.0% in chapter 8A and the 5.9% and 11.4% in chapter 10A all put Manappuram's enquiry count over a Google Analytics visit count. Read each of them as a ceiling, not a rate. The enquiry records carry no page or form name, so every gold loan enquiry from every page sits in the top of the sum while only one page's visits sit underneath. The true figure for any single page is lower, and cannot be worked out until Leak 1 is fixed.

11E What we measured across every gold loan page

On 17 September 2026 we opened all 75 gold loan pages on a 360 pixel phone and recorded, for each one, whether it has an enquiry form, how far down the form starts, whether anything on the first screen leads to it, and whether a phone number can be tapped. This is a count, not an impression, and the record is kept. It is the evidence behind Leaks 3, 6 and 14.

13 of 75 pages have no form at all

A visitor who lands on one of these has nothing to fill in, wherever they scroll.

32 of 75 show nothing to tap on the first screen

No form and no enquire button in the first phone screen, so the page opens with no way forward in view.

The typical form starts on phone screen 9

Across the 62 pages that do carry a form, the middle one starts on screen 9. 34 start on screen 9 or lower, and only 13 start on screen 1. The Gold Loan page is not an exception; it is the rule.

60 of 75 show nothing a phone can dial

On some pages the number is printed as plain text. On others, the Gold Loan page among them, a working dialling link sits in the page code but nothing visible on a phone leads to it. Either way the visitor has nothing to tap. Of the rest, 5 do show one and 10 could not be read.

11F The counted number behind each leak

Every leak in chapter 4 is something we saw on the live site. This table gives the counted figure that says how big each one is, and whether anyone can reproduce that figure today. Three cannot be, and they are marked. The leak is still real in those three; what we cannot currently show you is our working for the size of it.

LeakWhat it isThe counted figure behind itWhere that figure comes fromCan it be reproduced
1Nobody can see the enquiries8,692 enquiries the reports never sawManappuram's enquiry records against Google AnalyticsYes
2The rates page has no way to enquire208,385 visits in 3 months, 829,532 in a yearGoogle Analytics, landing pageYes
3The form is 8 screens down1,640 of 35,928 phone visits reached the bottom, and across all 75 pages the typical form starts on screen 9Google Analytics, plus our own measured census of every gold loan pageYes
4The price was added on 29 September, in a table a phone cuts off1,988 visits left the page for the rates page in 3 months, before the table existedGoogle Analytics, page to page, plus our own re-walk of 29 SeptemberYes
5"Apply Online, Instant Approval"49,932 visits in 3 months were shown that promiseGoogle Analytics, page views of the Gold Loan pageYes
6WhatsApp opens an unknown number60 of 75 gold loan pages show no number a phone can tapOur own measured census of every gold loan page, 17 Sep 2026. Google Analytics adds 172 taps on links leaving the page, but not which linkYes
7The form fights the phoneAll 5,000 feedback records carry the consent box ticked; 76 messages about the codeManappuram's feedback and complaint recordsYes
8Borrowing more gets stopped951 customer messagesManappuram's complaint and feedback recordsYes
9The page gets no credit for enquiries it wins282 enquiries in 3 months credited to another pageGoogle Analytics, thank you page by landing pageYes
10"Zero extra charges" is contradicted298 messages about charges being higher than expectedManappuram's complaint and feedback recordsYes
11The language pages ask in English470,178 visits since the pages went live. Conversion runs from 3.2% to 12.6% across pages that share the same English formGoogle Analytics, landing page. The English form is a defect we measured, not a cause we provedYes
12"Doorstep enquiry" opens the home page2,514 visits in a year to that addressGoogle Analytics. We did not keep the query, so this one cannot be reproducedNo
13Branch pages cannot take an enquiry577,212 visits in a year, of which 151,447 on a phoneGoogle Analytics, landing page, two pages added. Three quarters of this traffic is on a computer, so treat the phone figure as the customer facing floorYes
14Articles bury the form13 of 75 pages carry no form, and the typical form starts on phone screen 9Our own measured census of every gold loan page, 17 Sep 2026. The visit count of 161,300 is a hand built list whose query we did not keepYes
15Four pages open past their own button2,540 visits a year to the four pagesGoogle Analytics. We did not keep the queryNo
16The branch list sends cities to the wrong branches2 of 19 city pincodes return a wrong branch list or noneEach pincode sent to the form's own branch lookup, 21 September 2026Yes

11G What has changed since we measured

This site changes while it is being audited. On 17 September the Gold Loan page title changed between the morning and the afternoon. So every page in this report was walked again on 29 September 2026, and this is what had moved.

74 pages walked again

The same pages, the same 360 by 800 phone, the same measurements.

62 still carry an enquiry form

The same count as on 17 September. No page gained one and none lost one, so Leak 14 stands unchanged.

59 still show no number a phone can tap

Unchanged. Leak 6 stands.

One page changed in a way that matters

A Gold Loan Schemes rate table was added to the Gold Loan page. It answers the complaint in Leak 4, which said the page never gave a price. It also pushed the enquiry form about 1,900 pixels further down, which makes Leak 3 worse. Both are written up in those two leaks.

PageForm was atForm is now atMoved
/gold-loan6,673 px8,595 px+1922 px
/gold-loan/online-gold-loan6,743 px6,364 px-379 px
/10,114 px10,383 px+269 px
/doorstep10,114 px10,383 px+269 px
/gold-loan-calculator9,736 px9,887 px+151 px

Everything else held. The forms that moved under 400 pixels are ordinary page edits and change nothing in this report.

What this means for the counted figures

Every figure from Google Analytics and from the session recordings in this report describes the pages as they were between February and September. They measured a Gold Loan page whose form sat at 48% of its height. It now sits at 55%. Give it a few weeks and both tools will measure the page as it is now.

Two things we checked and did not publish

A first pass suggested 32 pages had gained an enquire button on the opening screen, and three branch pages had gained an enquiry form. Neither survived checking. The first was the floating "Doorstep enquiry" icon, which is Leak 12 and opens the home page. The second was the branch lookup filter, not a form anyone can enquire on.

11H How to check any of these yourself

Everything in this report was read on 29 September 2026. Analytics keeps reprocessing data for a while after it arrives, so a figure read months later can move by a fraction of a percent. What should not move is the method, so the method is written out here. You will need access to Analytics property 293901177 and to Clarity project ngcwpik1zw.

In Google Analytics

To checkWhere to goDimension and metric
Visits to a page, and visits that ended in an enquiryReports, then Engagement, then Landing page. Set the dates. Type the page address into the search box above the table. The Sessions column is what this report calls visits.For the enquiry side, build a segment of sessions where Page path and screen class contains thankyou and apply it to the same table. That keeps every enquiry attached to the page the visitor actually arrived on.Landing page, Sessions
How many people opened a page, rather than arrived on itReports, then Engagement, then Pages and screens. Set the dates and search for the page. This counts everyone who opened it, whether they landed there or reached it later.The two are different and the report always says which one it means. On the Gold Loan page the gap is close to double.Page path and screen class, Sessions
Phone against computer, for the site or for one pageExplore, new blank exploration. Dimension Device category, metric Sessions. For one page, add Landing page and filter it.Read the two numbers and divide. The report never quotes a share without the pair behind it.Device category, Sessions
How far people scrolledExplore. Dimensions Event name and Device category, metric Sessions, filtered to the page. Compare the scroll row against the page_view row.The scroll event fires once, at 90% of page depth. It tells you who reached the bottom, not who reached the form.Event name, Sessions
Google ads money and tapsExplore. Dimension Session Google Ads campaign, metrics Advertiser ad cost, Advertiser ad clicks and Sessions. Set the dates.The interface shows cost already in rupees. The figure the reporting interface hands back is in micro units, which is why chapter 11C explains the division by a million.Session Google Ads campaign, Advertiser ad cost
Which ad sent someone to which pageExplore. Dimensions Session manual ad content and Landing page, metric Sessions.Session manual ad content is the label the advertiser writes into the link. Do not use the campaign name here: every one of these campaigns reports the same campaign name whatever language page it points at, so a check on campaign names shows nothing wrong.Session manual ad content, Sessions

In Microsoft Clarity

To checkWhere to go
How far people scrolled on a pageHeatmaps. Paste the page address, choose Scroll, set the dates. The band down the side gives the share of sessions that reached each depth.This is the one that answers how many people ever see the form.
Taps that did nothingFilters, then User actions, then Insights, then Dead clicks. Apply it, then read Sessions. Add a page filter for one page.Rage clicks sits beside it and works the same way.
Where the traffic on a page came fromSet a page filter, then read the Referrers card on the dashboard.This is how the payment gateway behind /app/success.php was found.
Watching it happenRecordings, with the same filters applied. Clarity replays the session.Useful for anything in this report you would rather see than read.
One trap worth knowing

Adding twelve monthly figures does not give the same answer as asking for the twelve months in one go. Analytics counts a session once inside whatever window you ask for, so month by month totals come out higher. Every visit and enquiry figure taken from Analytics in this report is a single window query for that reason, so ask for the year in one go. The one place a sum of months is correct is Manappuram's own enquiry records, which arrive as a month and a count and have no finer grain to query.

Two that still cannot be reproduced

The visit counts behind Leaks 12 and 15 were typed into the working file without the query being kept. They are marked in the table above. The leaks themselves were seen on the live pages; it is the size of them we cannot currently show our working for.