Scoring the businesses the bureau cannot see
Crux ResearchMore than six crore Udyam-registered MSMEs operate in India, and most of them carry a thin formal credit file or none at all. The estimated unmet credit gap sits somewhere between ₹25 and ₹30 lakh crore. That gap is not primarily a risk-appetite problem — it is an assessment problem. A lender cannot price what it cannot see.
Bureau scores were built on repayment history. Where there is no history, the score is silent, and a silent score is indistinguishable from a bad one. The result is that the enterprises that most need working capital are the ones least able to evidence that they deserve it.
What has changed
Three things converged that were not true five years ago. India built lawful, machine-readable rails — ULI, Account Aggregator, OCEN, ONDC, GSTN, NeSL, DigiLocker — that expose business data at scale with consent attached. AI matured to the point where LLMs, knowledge graphs, agentic monitoring and explainable scoring make assessment of a data-thin entity economically viable rather than a loss-making analyst exercise. And the DPDP Act 2023 drew a hard line: scraping-based models now carry penalty risk, while consent-architected platforms are rewarded. Compliance stopped being overhead and became the moat.
What an alternative-data assessment actually uses
- GST filing regularity, turnover trend and counterparty concentration
- Account Aggregator cash-flow patterns, with consent, rather than static balance sheets
- MCA21 corporate structure, charges and director history
- Court and tribunal records, tender and GeM participation, NeSL default data
- Entity-resolved supply-chain and geospatial signals from the knowledge graph
None of that is a substitute for judgement. It is the raw material that lets judgement operate at portfolio scale, and it is why every score we issue carries reason codes rather than a bare number.
Assessments feed monitoring. Monitoring feeds the graph. The graph sharpens every next assessment.
Where this leaves lenders
A Business Health Score built this way does not replace the credit committee; it gives the committee a defensible starting point for an applicant who would otherwise have been declined for lack of file. Combined with continuous monitoring, the same infrastructure surfaces stress early — from litigation, GST irregularity, supply-chain concentration or adverse media — instead of at default.
That is the whole thesis: one data spine, one knowledge graph, three engines, and an assessment that gets sharper every time the platform is used.