All articles
ComplianceEpisode 3 of 5· 16 min read
DPDP Rules 2025 for Fintech, BFSI & NBFC

DPDP Rules 2025 Episode 3: Data Retention & Deletion for BFSI

The field-level problem nobody solves. The Third Schedule does not bind your NBFC. Rule 8(3) is an erasure clock, not a retention licence.

AS
Ankita Sharma
Talk to an expert
0
Third Schedule
Does not bind banks / NBFCs
2
Five-year clocks
KYC vs transaction triggers
3
Buckets
Keep · May keep · Must delete
7+
Locations
Where deletion actually fails

In Episode 2 we built continuous governance. Now the part that breaks it: proving you deleted a specific customer's non-regulatory data from core, warehouse, support, backups, and the KYC vendor, with an evidence log.

Retention is not a policy problem. It is a field-level engineering problem with a legal overlay. Most BFSI teams are solving the wrong version of it.

Two corrections before you build anything

Most DPDP retention guidance is written for consumer platforms and then applied to finance. That produces two expensive errors.

Correction 1: The Third Schedule almost certainly does not apply to you

The 3-year inactivity erasure and 48-hour pre-erasure notice bind only three classes. A bank, NBFC, digital lender, PA, or broker is none of these.

E-commerce entities≥ 2 crore registered users in India
Social media intermediaries≥ 2 crore registered users in India
Online gaming intermediaries≥ 50 lakh registered users in India

Caveat: a group marketplace, BNPL-in-ecommerce, or gaming-adjacent product above threshold may catch that entity. Scope at entity level, not group level.

Correction 2: Rule 8(3) is an erasure obligation, not a retention licence

Retain personal data, traffic data and processing logs for a minimum of one year, then erase them unless another law requires further retention. It reaches processors too.

Floor, not ceiling

One year minimum for personal data, traffic data and processing logs.

Then erase

Keep for a year is half the rule. The other half is delete them unless another law extends.

Reaches processors

Data held by a Data Processor on your behalf is still your obligation.

DPDP does not just tell you what to keep. It tells you when to stop keeping it. Indefinite retention now needs a positive legal basis, per field.

The BFSI problem is over-retention, not under-retention

Pre-DPDP, uniform 10-year schedules were rational. Under DPDP that logic inverts: data without a lawful basis is unlawful processing, a larger breach blast radius, and discovery liability.

The largest gap is not a missing record. It is years of data nobody can justify: closed-account KYC packs, dormant rows, Drive CSVs, marketing DBs, call recordings, support transcripts, analytics copies. Fix with controlled retention: every field, mapped to a basis, with an expiry and a trigger.

Five years is not one clock

Same duration, different triggers. One job keyed to account_closed_date gets transaction records wrong. Keyed to transaction_date deletes KYC five years too early.

≥ 5 years
KYC / customer identification

Clock starts: After relationship ends / account closed

≥ 5 years
Transaction records

Clock starts: From the date of the transaction

≥ 5 years
PMLA records to FIU-IND

Clock starts: From transaction date (client ↔ reporting entity)

Until case closure
STR / investigation records

Clock starts: No fixed expiry. Legal hold.

Retention is a (field, basis, trigger event, duration, action) tuple. Anything less fails at the edges, and auditors sample the edges.

The retention register that actually matters

Stop writing a retention policy. Build a field-level register with the legal basis cited per row. Indicative for a lending NBFC. Verify against your primary circulars.

Basis
RBI KYC MD; PMLA
Retention
≥ 5 yrs
Trigger
Relationship end
On expiry
Erase + certificate

Exposure lives in the bottom half: marketing, support, warehouse copies, collections recordings, analytics derivatives. No regulatory retention mandate. Precisely what nobody deletes.

The three buckets

Resolve an erasure request per field, not per customer. A blanket "we are RBI-regulated" refusal is a finding, not a defence.

Statutory retention mandate
Example: KYC records, transaction records

Lawfully refuse. Cite the basis, document it, tell the customer.

On refusal: still erase MUST-DELETE fields, communicate the basis in plain language, and document the decision for DPB scrutiny.

Deletion engineering: where it actually breaks

Deleting from the primary database is not deletion. Personal data lives in at least seven places.

Core / LOS / LMSMonolithic tables, soft-delete flagsField-level purge; verify soft vs hard
Warehouse / analyticsCopies at ingestion; no lineageLineage + cascade keyed to source
Logs / SIEMPII leaks into logsMask at emit; Rule 8(3) erase job
Backups & snapshotsImmutable; cannot row-deleteCrypto-shred keys, or document rotation
Support & CRMDepends on vendor APIContractual SLA + verify via API
Exports (Drive, email)Invisible; never inventoriedDLP + prohibition + sweep
Vendors / LSPsYou deleted; they did notContract + certificate of destruction

Backups: crypto-shred per-record keys, or document rotation and no reintroduction on restore. Silence is not defensible. Cite NIST SP 800-88 and ISO/IEC 27040 in policy.

Legal hold must beat everything

An engine that deletes under litigation, an STR, or an RBI inspection creates spoliation. Hold is a first-class citizen with priority over every expiry rule.

Deletion pipeline
1Deletion trigger
2Legal hold check
3Statutory register
4Delete primary
5Cascade all targets
6Verify confirmations
7Write evidence

Vendors, LSPs and processors

Rule 8(3) extends to processing on your behalf. If your KYC vendor still holds the data, you still hold the liability. Contract for it, then verify it.

Deletion on instruction, with defined SLA
Certificate of destruction per event
Deletion on contract termination + exit
Sub-processor cascade
Audit rights to test deletion
LSPs: minimal basic data only
Overseas processing: delete and repatriate in 24h

The deletion evidence pack

Deletion you cannot prove is deletion that did not happen. Every event should write an immutable record.

Record / principal IDOr hashed reference
TriggerRights request, expiry, withdrawal
Requester auth methodHow identity was verified
Legal-basis decisionDeleted vs retained, citing register
Systems actionedPrimary, warehouse, logs, backups, vendors
MethodHard delete, crypto-shred, overwrite
Operator + timestampNTP-synced
Vendor certificatesDestruction proofs
Verification resultPer target

Measuring it

If you track only one number: over-retention debt. It converts abstract risk into a Board metric that moves.

% fields with a retention ruleCoverage. Below ~95% means guessing.
Mean time to fulfil erasureAgainst SLA; grievances cap at 90 days
Cascade completion rateHonest number. Often under 50%.
Vendor deletion SLA / certificatesSupply-chain exposure
Over-retention debt volumeBiggest DPDP liability, quantified
Log erase-after-1-year successRule 8(3), the half everyone forgets
Legal holds active / releasedSpoliation risk

Common mistakes in regulated finance

01

Building Third Schedule machinery you do not need, while missing Rule 8(3) erase-after-one-year

02

Uniform 10-year schedules inherited from pre-DPDP eras

03

One clock for all five-year records

04

Soft deletes that hide but do not remove

05

Deleting from primary only

06

No answer for backups

07

Blanket refusal citing RBI without deleting non-regulatory fields

08

No legal-hold override

09

Marketing and support treated as if a mandate covers them

10

No deletion evidence

The first 90 days

Build the register, engineer the pipeline, prove it. Over-retention debt compounds daily until 13 May 2027, and every month is unnecessary breach exposure.

0%

Next: Episode 4 on breach readiness and the four-clock incident runbook. This episode reflects DPDP Rule 8 and Schedules, RBI KYC MD, PML Rules, Digital Lending Directions 2025, CERT-In 2022, and the CIC Act. Retention periods are indicative. Verify against your entity class. Not legal advice.

See how Bugmetrics turns retention into evidence

Retention register as live state. Deletion evidence collected automatically. Over-retention debt quantified for the Board.

Book a demo

Or explore Bugmetrics for Fintech & BFSI

Keep reading

Episode 4 · 18 min

Breach readiness & the four-clock incident runbook

Episode 2 · 14 min

Continuous data governance for BFSI: why your regulator already requires it

Episode 1 · 12 min

DPDP Rules 2025: The Dual Compliance Playbook