Industry playbook

DPDP for fintech (India): KYC, risk & records playbook

Audience: risk, compliance-adjacent product, ops, partnerships · Last reviewed: March 2026

Fintech teams already live under proportionality and auditability pressure from regulators and partners. DPDP adds a parallel track: data principal-facing clarity and operational proof for access, correction, deletion, and grievance handling. This playbook is written for commercial-intent search—implementation, not theory—alongside the checklist and complaint preparation.

If you cannot reconstruct KYC lineage (source, vendor, timestamp, human override) for a field shown to a user, you will struggle on both access requests and internal trust reviews.

30-day playbook

  1. Week 1 — Lineage v1: For each major attribute (ID, address, income proxy, device risk), map source system, vendor API, manual review queue, and downstream models.
  2. Week 2 — Retention truth: Align credit/fraud retention with retention checklist outputs; include backup and dispute archives.
  3. Week 3 — Frontline: Call-center scripts, chat exports, WhatsApp business use—tie to access and suppression expectations.
  4. Week 4 — Diligence pack: Canonical answers for security questionnaires; refresh subprocessor list.

Lifecycle (typical retail fintech)

  1. Pre-account: Attribution, waitlists, referral incentives.
  2. Onboarding: Document/liveness, bureau pulls, device risk, manual QC.
  3. Active: Transactions, collections notes, cross-sell journeys.
  4. Exit: Closure, dormant accounts, partner caches—often where notices overpromise.

Partner & stack grid

High-risk layers (customize to your stack)
Layer What to govern Review questions
KYC / fraud SaaS Verification APIs, watchlists, device intelligence Data residency, subprocessors, model change logs, human override audit trail
Core banking / ledger Identifiers, beneficiaries, settlement records Who can query PII? break-glass? export logs for diligence?
Collections & comms Voice logs, SMS, email, WhatsApp Retention of recordings; opt-out vs legal obligation conflicts
Growth & CRM Audiences, referrals, partner campaigns Consent vs legitimate-use framing per counsel; hashed lists still need governance

Disclosure: Vendor names are examples of categories only. We do not accept pay-to-rank placement. See editorial policy for how we handle referrals or affiliates.

Failure modes

Illustrative hypothetical (fiction, not factual): “RupeeArc Lending” retires a risk vendor but keeps a read-only warehouse slice “for backtesting.” A user’s access request shows current fields only; a marketing email references an older score that no longer appears in-app. The issue is inconsistent surfaces of truth. Fix: attribute inventory, aligned notices, and engineering deletion of derived artifacts—not longer policy PDFs.