BharatPe Credit Companion: Fixing the Trust Gap in Merchant Lending

01 · Context

Overview: The product works. The credit model works. What doesn't work is what happens when it says no.

BharatPe is India's largest merchant payment and lending platform, serving 1.7 crore merchants. Lending contributes 60% of revenue. The transaction infrastructure is trusted, the EDI repayment model is proven, and the credit model does what it is supposed to do.

What does not work is what happens when the credit model says no.

1.7Cr

Merchants on the BharatPe platform

60%

Revenue contribution from merchant lending

Rs 35.9Cr

Estimated annual revenue at stake from credit lock churn

The Core Tension

A merchant who repaid every single EDI on time — faithfully, without a single delay — finishes their loan and finds their credit line locked. The app shows one word: ineligible. No reason. No timeline. No next step.

02 · Problem Definition

The Problem: Ineligible

When a merchant's credit line gets locked after full loan repayment, the BharatPe app shows one word: ineligible. No reason, no timeline, no next step. The merchant who repaid every EDI on time, who transacted faithfully every day, is staring at a wall.

Customer care cannot explain it. The agent looking at the screen has the same information the merchant has: nothing actionable. The average reactivation time is 30 to 90 days. The industry benchmark is under 24 hours.

What the merchant experiences

Loan fully repaid — all EDIs cleared on time
Opens app expecting credit line to be available
Sees one word: ineligible
No reason given, no date, no action to take
Calls customer care — agent cannot explain either
Waits 30–90 days with no visibility on status

What the numbers show

10–12% of active borrowers experience a credit lock
20% churn post-lock — switching to a competitor or abandoning lending entirely
Industry benchmark for credit re-scoring after loan closure: under 24 hours
BharatPe actual: 30–90 days manual reactivation
Each lost borrower represents Rs 1,200–2,400 in annual interest income

The credit model works correctly. The lock is often legitimate — triggered by a UPI volume drop below the threshold for the next credit tier. The problem is not the decision. The problem is that nobody explains it. A rule engine can produce a generic error message. It cannot tell Ramesh why his credit line is locked, what he can do about it, or when it will be reviewed.

Why this is a trust problem, not just a UX problem

Trust in financial products is asymmetric. It takes months of consistent repayment behaviour to build, and a single unexplained rejection to break. Avaya's survey of Indian banking customers found that 37% switch after a bad experience. The switching cost for BharatPe is lower than most banks — a merchant's transaction history does not transfer to a competitor, but their frustration does.

I paid everything on time. Every single day, money was deducted. And then when it's over, they tell me I'm not eligible? For what? Nobody can tell me anything.

— Merchant sentiment, distilled from customer support patterns

03 · Product Solution

Credit Companion: An AI layer with three components

Credit Companion is not a redesign of BharatPe's credit model. It is an AI layer built on top of it — one that translates credit model outputs into merchant-facing explanations, automates re-scoring at the moment of loan closure, and gives the merchant a live window into their own credit standing.

01

Automated Credit Standing Refresh

Triggers re-scoring the moment a loan closes — eliminating the 30–90 day manual reactivation gap

How it works

At the moment the final EDI clears, an automated re-scoring pipeline is triggered
The merchant's UPI transaction data, repayment history, and transaction diversity are re-evaluated in real time
If eligibility is confirmed, the credit line is reactivated within 48 hours
If a review period is required (e.g., volume dropped below threshold), the merchant receives a clear timeline — not a blank ineligible state

Impact

Closes the gap between the industry benchmark (under 24 hours) and BharatPe's current performance (30–90 days). For merchants who clear re-scoring immediately, the credit line is available before they even notice it was gone.

02

Credit Lock Explanation

Plain-language reason for the lock, specific review date, and three ranked actions the merchant can take

What the merchant sees instead of "ineligible"

Plain-language reason: "Your UPI transaction volume in May and June was Rs 46,000 — lower than your usual Rs 68,000. Our system automatically reviews credit limits when volume drops below your tier threshold."
Specific review date: "Your credit line will be reviewed on 3 July 2025. I will notify you by SMS and app notification that day."
Three ranked actions: What the merchant can do between now and the review date, ordered by credit impact — e.g. keep transacting daily, accept card payments, continue using BharatPe QR
What is not affected: Explicit confirmation that payment acceptance continues normally — only new loan disbursements are on hold

For the agent (customer care)

Agent sees the same data the merchant sees, plus internal signals: lock trigger type, fraud flag status, whether an override is possible
AI-generated guidance script — not a generic template but specific to this merchant's situation
Explicit list of what the agent should NOT say (e.g., "I will escalate this" when there is nothing to escalate)

Impact

Transforms a frustrating dead-end into a transparent, actionable waiting period. The merchant understands what happened, knows exactly when it will be reviewed, and has specific things to do in the interim. The agent can confidently explain the lock without escalating unnecessarily.

03

Live Credit Score

Built on the merchant's own UPI transaction data — with factor breakdown and indicative limit trajectory

What the merchant sees

BharatPe Business Score: A single number (e.g. 724) built entirely on BharatPe transaction and repayment data — not CIBIL, not a bureau score
Factor breakdown: Each contributing factor shown with its status — UPI volume consistency, repayment history, transaction frequency, transaction diversity — and what improving each factor would do to the credit limit
Limit trajectory: "3 months ago: Rs 80,000. Today: Rs 1,20,000. In 3 months at current pace: up to Rs 1,50,000." Shown as an indicative estimate, not a guaranteed offer
Specific improvement tips: Not generic advice but personalised actions — "Adding card payments could add Rs 8,000 to your limit"

Impact

Turns BharatPe's transaction data — which the merchant generates every day — into a visible asset the merchant can understand and act on. Increases merchant engagement, drives behaviour that improves creditworthiness, and strengthens the feedback loop between merchant behaviour and credit limit growth.

04 · AI Architecture

The AI Role: Why a rule engine cannot do this

A rule engine can produce a generic error message. It can tell every locked merchant the same thing. AI does something different: it personalises the explanation to the merchant's specific situation at scale.

NLP — Credit explanation layer

Translates credit model outputs into personalised merchant-facing explanations. Takes the structured output of the credit model — volume threshold breach, review date, tier classification — and renders it as natural language specific to this merchant's numbers and history.

ML — Transaction signal classification

Classifies UPI transaction patterns into credit-relevant signals across thousands of data points per merchant. Identifies volume trends, frequency patterns, transaction diversity, and seasonal behaviour. Feeds both the credit model and the merchant-facing factor breakdown.

Cohort matching — Limit trajectory

Projects indicative limit trajectory by matching the merchant against 3.5 lakh similar merchant profiles — same category, similar transaction volume, similar repayment history — and showing what limits merchants in that cohort reached over 3, 6, and 12 months.

Agent guidance generation

Generates a specific, ordered guidance script for the customer care agent based on the merchant's exact situation — not a template but a script tailored to this lock type, this merchant's history, and this call's context. Includes what to say, what not to say, and how to log the outcome.

Why AI, not rules

BharatPe has 1.7 crore merchants. No two credit lock situations are identical — the reason, the severity, the merchant's history, and the right set of actions differ across every case. A rule engine produces the same message for all of them. AI produces the right message for each of them.

05 · Business Impact

Impact: Rs 35.9 crore in recoverable revenue

The business case starts with a simple calculation. 10 to 12% of active borrowers experience a credit lock. Of those, 20% churn — they stop borrowing from BharatPe, either switching to a competitor or abandoning merchant lending entirely. At an average loan size of Rs 80,000 to Rs 1,00,000 and an effective interest rate of 18–24%, each lost borrower represents Rs 1,200–2,400 in annual interest income.

10–12%

Active borrowers who experience a credit lock in any given cycle

20%

Churn rate post-lock — anchored in Avaya's 37% banking switching rate, moderated for BharatPe's switching cost

Rs 35.9Cr

Estimated annual interest income at risk across the active borrower base

What Credit Companion recovers

Automated re-scoring eliminates the 30–90 day gap — merchants who clear re-scoring immediately stay in the lending cycle rather than churning during the wait
Credit lock explanation reduces irrational churn — merchants who understand the reason and have a clear timeline are significantly less likely to defect than merchants facing a blank ineligible state
Live Credit Score drives engagement and behaviour change — merchants who can see their credit trajectory and understand the levers have a direct incentive to transact more, improving their own creditworthiness and BharatPe's loan book quality simultaneously

The revenue at stake is not theoretical. It is currently being lost — one frustrated merchant at a time, through a combination of a 30–90 day reactivation gap and a single unexplained word on a screen.

06 · Deliverables

Deliverables: PRD, slide deck, and interactive prototype

This case study produced three artefacts: a full product requirements document, a presentation-ready slide deck, and an interactive HTML prototype demonstrating both the merchant flow and the agent guidance flow.

PRD (PDF) Slide Deck (PDF) Live Demo

AI Product Management Course | Hello PM | 2026

This case study was produced as part of the Hello PM AI Product Management programme. It covers problem discovery, solution design, AI architecture, business case modelling, and a fully interactive prototype demonstrating merchant and agent flows.