Finds bank and card charges that break RBI rules and helps you claim them back
TL;DR Feefinder reads your bank and credit card statements, flags charges that may break RBI conduct rules and drafts the complaint for you. It goes to the bank first, and then to the RBI Ombudsman if the bank says no. Customers pay a small fee per dispute pack or a share of whatever they get back.
Small bank charges are easy to ignore and a pain to fight. People are complaining more anyway, and the RBI Ombudsman received 13,34,244 complaints in 2024-25, up 13.55% in a year[1]. Credit card complaints alone rose 20.04%[1].
A lot of people also complain in the wrong order. Of the complaints the Ombudsman closed as not maintainable, 36,289 had gone straight to the Ombudsman without first going to the bank[1].
Here's the thing most customers don't know - the rules are on their side more often than they think. A minimum-balance penalty needs prior notice, at least a month to restore the balance, and a charge in proportion to the shortfall[2]. Card late fees apply only to the overdue amount, and only after three days past due[3].
First, the regulatory bit - Feefinder does statement analysis and complaint drafting only. It does not lend, hold money or give investment advice, and it doesn't pull bank data through Account Aggregators, which only regulated entities may use as data users[4]. Customers upload their own statements, and data protection law covers the rest.
The customer uploads a PDF statement and Feefinder reads every debit, picking out minimum-balance penalties, card late fees, mandate bounce charges, penal charges on loans and annual fees on cards that were sold as free. Each one is checked against the relevant RBI rule, such as the 2024 ban on capitalising loan penal charges[5].
A human reviewer confirms every flag before the customer sees it. Then the customer gets a ready-to-send complaint to the bank with dates and rule references, plus a reminder to escalate after 30 days if needed.
I reframed the seed, which pitched AI nudges against small spending leaks. Without regulated access to bank data, nudges rely on manual entry and rarely change anyone's habits. Charges that break written rules are a narrower leak with a clear fix.
The account holder pays. My test pricing assumption is ₹99–299 per dispute pack or 20% of any amount refunded, whichever the customer picks. A free scan shows what was found. Later on, employers and credit-card communities could buy scans in bulk.
What's the alternative today? Free and slow - writing to the bank yourself or filing with the Ombudsman.
Credit card complaints grew 20.04% in 2024-25[1], and the penal-charges rules that kicked in from January 2024[5] gave borrowers a new, specific ground to push back on. The Ombudsman's own numbers show tens of thousands of customers still file in the wrong place[1], and that's a gap a guided product can close.
Nobody publishes a figure for how much Indian customers recover in disputed bank charges. The firmest proxy is complaint volume, with 49,833 complaints about deposit accounts and 41,457 about credit cards reaching the Ombudsman against banks in 2024-25[1].
That's demand already acting, before any product helps it along.
A large US money app finds subscriptions, cancels them and negotiates bills, and claims more than 10 million members[6]. Debt-relief platforms in India help borrowers with loan settlement and harassment, not everyday charges[7]. Banks' own apps show charges but don't tell you which ones might be wrong.
In my limited checks I found no Indian service that audits charges against RBI rules. Full web search was unavailable though, so verify this before you build.
| Difficulty | Moderate (38/100) |
| Starting capital | ₹35–70 L |
| Time to launch | 4–6 months |
Reading statements from 15–20 banks reliably is fiddly but solvable. The harder parts are judgement and trust, because many charges are lawful so your flags have to be precise, and people have to be willing to upload full statements to a brand they've never heard of.
No financial licence, as long as you only read statements and draft complaints. The limits matter though.
Building this outside India? The categories above usually have a local equivalent, so check with your own regulator.
Don't let AI send a flag or complaint without human review. A wrong claim wastes the customer's time and damages trust, and parsing errors on odd statement formats are common.
Want to work out the AI layer for Feefinder? Talk to Yash →
Here's what can go south -
I'd say this is worth a shot for a founder with bank operations or grievance-handling experience and a careful approach to data. If you're expecting a nudge app to grow on downloads alone then look elsewhere.
Before you build much, prove two things - that one in four scans finds a charge worth disputing, and that banks reverse at least half of those.
Plan for roughly ₹35–70 L over 4–6 months, mostly on statement parsing for 15–20 banks, human reviewers and building trust.
Not a financial licence if you only analyse uploaded statements and draft complaints. You can't use Account Aggregators, and you need DPDP compliance for statement data.
Yes, the model works wherever banks have written conduct rules and an ombudsman. Check local rules on charging for complaint help and handling bank data.
The free Feefinder validation kit has the line-by-line capital breakdown, the effort scoring, the RBI rule and data-protection checklist, a 14-day validation sprint with pass/fail targets, a 90-day launch plan, 10 interview questions for account holders, and 5 assumptions to test. Download the kit →
Thinking about building Feefinder or something close to it? The make-or-break work is proving that enough charges are actually wrong and earning enough trust for people to upload statements, and I'm happy to think that through with you. If you need a hand, you know where to find me. Talk to Yash →
For illustration only. Names are fictional and ideas hypothetical; sourced facts are real but may be out of date. Nothing here is legal, financial or professional advice, and Yash Khatri is not personally liable for any use of it. Read the full disclaimer.