The document was never the whole identity.
Praman Labs evaluates the part of application-time identity risk that valid credentials alone cannot answer.
We kept seeing the same missing question.
Onboarding checks establish that records exist and agree. They do not establish that the named person independently controls the application.
The gap matters because modern identity fraud often borrows valid material. The document belongs to someone. The mobile number works. The face on the video can match. Yet the surrounding economic history, device network and submission behaviour point to a different operator.
Transaction monitoring can identify suspicious movement after an account opens. Credit models can estimate repayment risk from the file that exists. We think a separate decision belongs at application time: is there a real, independent economic actor behind this request?
That is a narrower claim than replacing KYC, a bureau or fraud operations. It is also testable. A retrospective backtest should show where the signal would have changed a decision and where it would have been wrong.
What we believe.
These are product positions that affect thresholds, deployment and what we are willing to claim.
Validity is not control.
A PAN, Aadhaar record and face match can all be valid while another person controls the application and the account that follows.
Thin is not false.
A first-time borrower with little bureau history should not be treated like an identity assembled for fraud. Contradiction and absence are different signals.
The useful test happens before procurement.
A risk team should be able to test the claim on its own historical applications before committing engineering time or changing a credit policy.
Privacy architecture is part of model quality.
Institutions will not pool customer identity records, and they should not have to. Cross-member learning must work without building a central customer database.
How we work with institutions.
Evidence, authority and accountability stay visible from the first backtest through any later operational use.
Evidence before authority.
Backtests and shadow mode come before a score changes an application outcome. The institution reviews the results and decides when, where and whether to grant decision authority.
Institutional control.
Each institution owns its policy thresholds, review capacity, exception paths and the legal basis for its final decision. Praman Labs supplies evidence and versioned recommendations.
Claims with context.
Any performance figure must identify the cohort, outcome label, decision threshold and evaluation date. A result from one institution is not presented as a universal benchmark.
Assurance by scope.
Security and governance evidence is considered with its scope and date. A public description does not replace an institution's diligence or contractual controls.
Judge the argument on your own applications.
A backtest shows where the signal helps, where it misses and how much review capacity it requires before your team considers operational use.
Book a backtest