Every loan DSA in India has had this month. You buy a fresh data pack, you sit down with a full battery and a fresh SIM, and you start dialling. By evening you have called 180 numbers. Forty did not connect. Sixty disconnected within five seconds. Twenty said they never applied for anything. Fifteen had already taken a loan last month. Six were interested but their salary was below policy. Two said “call me later” and never picked up again.
Zero files punched.
Then the same vendor messages you on WhatsApp: “Bhai fresh data aaya hai, 10K records, ₹8 per lead.” And because you have already spent the money and you need to recover it, you buy again.
This article is about breaking that loop. Not with motivation, but with arithmetic, a testing protocol you can run for under ₹2,000, and a realistic look at where loan lead data in India actually comes from.
Run the math before you run the dialler
Most DSAs never calculate their true cost per disbursal. They calculate cost per lead, which is the number the vendor wants them to look at.
Take a typical personal loan cycle:
- 5,000 leads purchased at ₹8 each = ₹40,000
- Connect rate on aged resold data: roughly 35% = 1,750 conversations
- Of those, genuinely interested: 8% = 140
- Of those, matching lender policy (age, salary band, CIBIL, serviceable pincode, employment type): 25% = 35
- Of those, documents complete and submitted: 60% = 21 files
- Of those, actually disbursed: 30% = 6 disbursals
Six disbursals at an average ticket of ₹2,00,000 with a 3% payout gives you ₹36,000. You spent ₹40,000 on data alone, before counting your SIM costs, your time, and the two months of follow-up.
You worked for free and paid ₹4,000 for the privilege.
Now change one variable. If the connect rate goes from 35% to 60% because the data is genuinely fresh and consented, and the policy-match rate goes from 25% to 45% because the data was filtered for your lender’s criteria, the same 5,000 leads produce closer to 20 disbursals. Same spend, same effort, triple the income.
Data quality is not a nice-to-have in this business. It is the business.
Where “data” actually comes from
Vendors rarely tell you the source. There are broadly five, and they are not equal.
1. Resold and recycled lists. The most common. A lead generated eight months ago is sold to one DSA, then resold, then bundled into a “fresh” pack and sold again. By the time it reaches you, twelve people have called that person. This is why the customer sounds irritated before you finish your first sentence.
2. Scraped data. Numbers lifted from classified sites, job portals, property listings, and public directories. These people never applied for a loan. There is no intent whatsoever, and there is no consent — which matters more than most DSAs realise.
3. Incentivised form fills. Traffic driven to a landing page that promises a free CIBIL check, a lucky draw, a recharge coupon, or a scholarship form. The person filled the form to get the freebie, not the loan. Volume is high and price is low, which is exactly why these packs look attractive.
4. Old CRM dumps. When a small fintech or NBFC shuts down or pivots, its database sometimes leaks into the market. These are real applicants with real intent — from 2022.
5. Genuine first-party leads. Someone ran a live campaign, captured the enquiry with consent, and is selling it within 24–48 hours, exclusively or to a capped number of buyers. This exists. It costs ₹150–₹600 per lead, not ₹8. When a vendor offers you first-party quality at scraped-data pricing, one of those two claims is false.
The five quality killers, in order of damage
Age and exclusivity
Lead value decays fast. A personal loan enquiry is most convertible in the first 6–12 hours and is close to dead after 15 days. Ask two questions of any vendor: how old is this lead and how many buyers is it sold to. If the answer to the second is “only you” at ₹8 a lead, the answer is not true.
Intent mismatch
A person who filled a form for a credit card is not a business loan prospect. A person checking their CIBIL score is often someone who already knows they have a problem. Data packs are frequently blended — credit card, personal loan, insurance and education enquiries mixed into one file — because a blended pack is easier to sell by volume.
Ask for the original form field. If the vendor cannot tell you what the person actually filled, treat the whole file as scraped.
Junk and defensive contact fields
This one is underrated, and it is the fastest quality test you have.
When a list contains email addresses, look at the domains. In any genuine, consented lead list you will see a normal spread of Gmail, Yahoo, Outlook and Rediff, with some company domains. What you should not see is a heavy concentration of throwaway domains.
There are two different things happening when disposable addresses show up, and both tell you something useful.
The first is defensive behaviour by real people. Indian consumers have learned the hard way that submitting an email to a finance-related form means permanent spam. So a meaningful share of genuinely interested applicants now use temporary email services when a form demands an address before showing them anything. That person may well be a real prospect — but their email is dead, which means your entire email follow-up sequence to that list is burning for nothing, and your “bounce rate” is not a deliverability problem, it is a data composition problem.
The second is fabrication. When a data vendor or an affiliate is paid per record, padding the file with auto-generated entries is trivial. Disposable and randomly generated domains are the cheapest way to make a row look complete.
Either way, the practical move is the same: before you buy 5,000 records, run the email column of the sample through a verification check and look at the domain distribution. If more than 10–15% of addresses sit on disposable or nonexistent domains, the file has been padded or aggressively recycled. You have learned this for free, in twenty minutes, before spending ₹40,000.
The same logic applies to phone fields. Sequential numbers, repeated digits, and numbers that fail a basic operator lookup are all padding signals.
Policy mismatch
This is the killer that feels like your fault but isn’t. You get a genuinely interested, genuinely fresh lead — and then the person turns out to be 19 years old, or self-employed when your lender only takes salaried, or earning ₹12,000 in a Tier 1 city, or living in a pincode the NBFC does not service.
Serious data buyers filter before purchase, not after. Give your vendor your lender’s basic policy and ask for the file to be pre-filtered on the fields they actually hold: age band, employment type, declared income band, city or pincode. If the vendor holds none of these fields, they are selling you a phone list, not loan leads, and it should be priced like a phone list.
Compliance exposure
This is the part most DSAs ignore until it becomes expensive.
Unsolicited commercial calling in India is governed by TRAI’s telecom commercial communications regulations. Calling numbers registered on the DND / customer preference registry, from an unregistered telemarketing channel, is not a grey area — it can get your number blocked by the operator, and repeat complaints escalate. Losing your primary calling number mid-pipeline costs you far more than a data pack.
Separately, the Digital Personal Data Protection Act, 2023 establishes a consent-based framework for handling personal data in India, with obligations and penalties that are being phased in as the rules come into force. The practical implication for a DSA is simple: you are handling other people’s identity and financial information, and “I bought it from someone” is not a strong position. Ask your vendor where the data originated and whether consent was collected for this purpose. If they cannot produce an answer, that is information about them.
None of this is legal advice, and the specifics of your obligations depend on your structure and volume — worth a short conversation with a CA or lawyer if data buying is core to how you operate.
The ₹1,600 vendor test
Never make your first purchase from a vendor a bulk purchase. Run this instead.
Step 1 — Buy a 200-record sample. At ₹8, that is ₹1,600. Most vendors will resist. A vendor who refuses a paid sample is telling you the bulk file will not survive inspection.
Step 2 — Deduplicate against your own history. Load it against every list you have bought in the last year. If more than 5% overlaps with data you already own, the vendor is recycling from the same pool as everyone else.
Step 3 — Check field completeness. What percentage of rows have all four of name, number, city, and income or employment indicator? Below 70% completeness, the file is scraped or merged from partial sources.
Step 4 — Audit the email domains. As above. Distribution tells you more than any vendor claim.
Step 5 — Dial 50 records in the first 24 hours and log four numbers:
- Connect rate (target: above 55%)
- “I never applied” rate (target: below 20%)
- Recall of the specific offer they filled for (target: above 40% remember something)
- Policy match rate (target: above 30%)
Step 6 — Decide with the numbers, not the relationship. If the sample clears, buy in tranches of 1,000, not 10,000, and re-run step 5 on each tranche. Vendors have a documented habit of sending a clean sample and a dirty bulk file.
Two hours of discipline here protects two months of income.
The exit: leads you own
Every DSA who builds a stable income eventually stops buying cold data, because purchased leads have no compounding. Owned lead sources do.
The realistic ones, roughly in order of return per rupee:
Existing disbursed customers. Your single most valuable asset and the one almost everyone neglects. A personal loan customer becomes eligible for a top-up or balance transfer around month 11–14. Maintain a simple calendar and call them. Conversion rates here are several times anything you will buy.
Referral partners in adjacent trades. Chartered accountants, property brokers, insurance advisors, used-car dealers, computer and mobile shop owners on EMI, local builders. These people sit next to a loan requirement every single week and currently do nothing with it. A structured referral share turns each of them into a small standing pipeline.
Google Business Profile. Deeply underused in the loan agent segment. Someone searching “personal loan agent near me” in your city has intent that no purchased lead can match, and the profile is free.
WhatsApp and community presence. Local trade groups, RWA groups, and your own broadcast list. Slow to build, near-zero cost, and it converts because you are a known face rather than an unknown number.
One honest landing page. A single page that states which lenders you work with, the documents required, and realistic timelines will outperform a hundred generic ones, because the enquiries that arrive are pre-qualified by the content.
Purchased data can be a bridge while you build these. It is a terrible destination.
The one-line version
Cost per lead is a vanity number. Cost per disbursal is the only figure that pays your rent — and the single fastest way to improve it is to spend ₹1,600 and two hours testing a vendor before you spend ₹40,000 trusting one.