Pre-Loan Credit Decisions: By the time a borrower fills out a loan application, the most useful information about them has already been generated.
It sits on the platform where they sell their goods, the logistics network where they earn their living, and the school that admitted their child. Behavioral data like order frequency, earnings consistency, and dispute history are accumulated across months and years of observable activity. None of it appears on a standard credit application. Most of it never reaches a bureau.
Traditional lending ignores this. It asks the borrower to describe themselves through a form, supplements that with a bureau score, and calls the result underwriting. Selection at source starts somewhere different — inside ecosystems where the borrower’s behaviour is already visible, and the underwriting begins with a fundamentally better picture.
The Thin File Problem
According to TransUnion CIBIL’s 2022 global study, Empowering Credit Inclusion, only 179[1] million Indian adults were considered “credit served”. This means they had sufficient bureau data for traditional underwriting. Another 164 [2] million were “credit underserved,” and 408[3] million were entirely “credit unserved.” More recent analyses suggest the gap has not closed; a 2025 estimate places the combined total of unserved and underserved Indians at nearly 850 million[4] . These hundreds of millions are not uncreditworthy. They are unobserved by the formal systems that traditional underwriting depends on.
- A delivery partner who has completed 4,000 orders across two years with a platform has demonstrated discipline, consistency, and earning capacity in ways that a bureau score cannot capture.
- A small D2C merchant who has processed ₹80 lakh in annual revenue through a payment gateway has generated a repayment capacity signal that no self-declared income figure can match.
- A student admitted to a premier engineering institution has passed a selection process that tells a lender more about their future earning trajectory than a credit score ever could.
Hence, traditional underwriting cannot see any of this. It is looking in the wrong place, which is at formal credit history, for information that already exists in a richer, more reliable form somewhere else.
What Ecosystem Data Actually Tells You
The value of ecosystem data in credit decisioning is not simply that it is more data. It is more relevant data, generated closer to the actual behaviour the lender is trying to predict. A CGAP study[5] found that transactional data from gig platform workers and micro and small enterprises can predict creditworthiness as effectively as formal credit history — and that combining both improves predictive accuracy further without increasing portfolio risk.
Capacity
A gig platform that has twelve months of a driver’s earnings history knows more about their income stability than any payslip can confirm. Moreover, a supplier network that has recorded three years of a merchant’s order fulfilment, returns, and dispute resolution knows more about their operating discipline than a reference check would reveal.
India’s gig workforce reached 12 million[6] in FY2025, up 55% from FY2021, according to the Economic Survey 2025-26. The Survey notes that these workers have “thin-file” credit access and face income volatility that traditional underwriting struggles to assess. Classified as freelancers and independent contractors, the vast majority carry no formal income documentation that conventional lenders can use. Their earnings data exists. It simply lives on the platform, not in a bureau file.
Consistency
A single month of strong earnings tells a lender very little. Eighteen months of stable earnings, with visible patterns across seasons, demand cycles, and platform changes, tell them considerably more.
India’s Account Aggregator Framework facilitated loan disbursements of ₹1.47 lakh crore[7] in the first half of FY2026 alone, primarily using bank statement and transactional data to assess creditworthiness, according to Sahamati. Ecosystem data is longitudinal in a way that bureau scores are not. It captures not just what a borrower can do, but how reliably they have done it across varying conditions.
Intent
A borrower who has been active on a platform for two years, maintained a high completion rate, and managed platform-level credit responsibly has demonstrated something beyond capacity and consistency. Behavioural signals such as dispute resolution patterns, response to platform incentives, voluntary repayment of platform advances, tell a lender something about character that no form can elicit and no bureau can record.
According to the DataIntelo Alternative Credit Scoring Market report published in March 2026, incorporating alternative data into underwriting models improves predictive accuracy by 20 to 35%[8] [A9] without expanding portfolio risk. Intent is the component of that improvement that no traditional data source was ever designed to capture.
Why This is a Structural Edge
Platform-based lenders operating with ecosystem data do not simply have better information at the point of underwriting. They have a structural advantage that compounds over time.
Pavitra Pradip Walvekar, a Pune-based entrepreneur and investor whose work spans Indian fintech, credit, and capital allocation, has described this as the difference between lending against a photograph and lending against a film. A bureau score is a photograph, a static representation of a borrower at a single point in time. Ecosystem data is a film. It is a continuous record of behaviour that reveals patterns, trends, and consistency in ways a snapshot cannot.
This distinction matters most at the margins, which is where most credit decisions live. For the borrower with a strong formal credit history, the bureau score is adequate. For the borrower with a thin file, the bureau score is nearly useless, and the ecosystem data is the only reliable picture available.
Selection Before Application
The deepest expression of selection at source is not just using ecosystem data to underwrite better. It is using ecosystem data to identify creditworthy borrowers before they have applied, and in some cases before they know they want to borrow.
A platform that monitors earnings consistency, order completion rates, and behavioural signals can identify the subset of its participants who meet a lending threshold before any application is made. Pre-approved offers based on observed behaviour remove the friction of a formal application process, reduce adverse selection, since the borrower population is defined by the platform’s data rather than self-selected by the willingness to apply, and create a lending relationship that feels like a natural extension of the platform rather than an external financial product.
This is the structural edge that generic consumer finance cannot replicate. A lender without platform access cannot acquire this data. A fintech without embedded distribution cannot reach these borrowers at the moment their borrowing need is legible. Both the information and the timing are only available from inside the ecosystem.
The Underwriting Happened Already
Traditional lending asks: Who is this borrower, and can they repay?
Ecosystem lending asks a different question: what do we already know about this borrower, and how long have we known it?
In most cases, the answer is: enough, and long enough. The underwriting happened in the months and years before the loan application was ever opened. The lender’s job is to read it correctly — not to reconstruct from a thin file what the ecosystem has already made visible.
