InfraVeritas360InfraVeritas360हिन्दीमराठीதமிழ்DPDPiq

DPDP Insights › Banking, financial services and insurance › Data and analytics head

DPDP for the Data and analytics head in BFSI

Reviewed by InfraVeritas360 ResearchFirst published Last reviewed

Sources: DPDP Act, 2023 and DPDP Rules, 2025 (MeitY) · RBI KYC Master Direction · RBI · RBI · RBI

Credit scoring, fraud models and cross-sell propensity all use customer data. DPDP asks whether each use fits the purpose the customer was told.

Written for: Data and analytics head, Banking, financial services and insurance. Open this seat in the interactive tool

What is different here

Fraud checks and credit decisions usually fit the purpose of the account or loan. Propensity models for partner products, and lists for insurers or fund houses, need specific consent. Credit bureau data also comes under the Credit Information Companies Act, and accuracy matters because models drive lending decisions.

The first four things to sort out

  1. List each model and the purpose it serves.
  2. Score for cross-sell only customers with consent for offers.
  3. Mask account and card numbers in analysis copies.
  4. Give each analysis copy a deletion date.

A worked example

Data and analytics head

A propensity model for a partner health policy

  1. Day 1The business asks for a model to pick customers for a partner insurer's health policy.
  2. Day 2The team checks the consent store. Customers without consent for insurance offers are left out of scoring and of any list.
  3. Week 1The model runs on masked IDs inside the warehouse. Only the final consented list leaves, through the CRM.
  4. AfterThe training copy gets a deletion date, and the model is registered with its purpose.

Evidence kept

  • Purpose check
  • Data specification
  • Model register entry
  • Deletion date

Score only the customers you may contact.

What others in the sector usually do. Many analytics teams start by tagging warehouse tables with the purposes they may be used for.

Where it usually goes wrong, by organisation type

Organisation typeHotspots
Scheduled commercial bankCross-selling insurance and mutual funds on account terms; Business correspondent devices and paper forms in villages; Old core banking archives with no deletion path
Co-operative bank (urban or rural)Vendor-run core banking with admin access from the vendor's office; Member and share registers kept on open shelves; Directors and staff who are also members and relatives of borrowers
NBFC and digital lenderApps asking for contacts, photos and call logs; Collection agents sharing borrower details with family or employers; Leads bought from aggregators with no consent record
Insurance company (life, general or health)Medical reports passed to TPAs and hospitals by email; Agent and broker access to policyholder data; Claims data kept long after the claim is closed
Broking, depository and wealthAuthorised persons with client lists on personal phones; Research-tip calls to people who never consented; Client KYC copies shared over email with partners
Payments and fintechCard numbers in logs and support tickets; Merchant onboarding documents in shared drives; Fraud models that use data beyond what users were told

10 questions the Data and analytics head asks, answered in full

How long must we keep KYC and transaction records, and what happens after?

Short answer: At least five years after the relationship ends; then erase

Under PMLA and RBI's KYC Master Direction, keep identity records for at least five years after the relationship ends, and transaction records for at least five years from the transaction. During that time, keep them only for that legal purpose. After it, erase or anonymise unless another law requires more.

What the law says

Section 8(7) allows retention where a law requires it. PMLA and RBI's KYC rules are such laws. Section 8(7) · Rule 8 · Section 7

Steps
  1. List KYC and transaction record types.
  2. Set the start date: end of relationship or date of transaction.
  3. Restrict access to closed-account records.
  4. Erase or anonymise after the period.
  5. Explain this in erasure replies.
Evidence to keep
  • Retention schedule
  • Access restrictions on closed accounts
  • Deletion logs
Common mistakes
  • Keeping everything for ever
  • Deleting before the legal period
  • Using closed-account data for marketing

What can our lending or banking app collect from a phone?

Short answer: Need-based only, with explicit consent

Only what you need, with explicit consent. RBI's Digital Lending Directions already bar access to contacts, files, media and call logs, and allow one-time access to camera, microphone or location for onboarding or KYC. DPDP adds a notice, the right to withdraw and the right to know what was collected.

What the law says

Section 6 needs consent limited to what is necessary. RBI's Digital Lending Directions, 2025 set specific limits. Section 6 · Section 5 · Rule 3 · Section 8(5) · Rule 6

Steps
  1. List every permission the app asks for.
  2. Remove permissions not needed.
  3. Ask camera and location only at the KYC step.
  4. Show the notice in the app before sign-up.
  5. Check partner apps the same way.
Evidence to keep
  • Permission list with reasons
  • App store listing
  • Partner app review
Common mistakes
  • Contacts permission for 'reference checks'
  • Third-party analytics SDKs collecting device data
  • Partner apps not reviewed

Can we use customer data for analytics and models?

Short answer: Yes, within the purpose or with consent

Yes, where the use fits the purpose people were told, or you have consent for it. Improving the service people signed up for usually fits. Selling insights, profiling for third parties or targeting children does not. Mask or aggregate wherever the question allows, and give analysis copies their own deletion date.

From your seat: Data and analytics head. Check the purpose before each new analysis. Purpose decides whether you need consent.
What the law says

Section 4 needs a lawful basis for each purpose. Section 6 needs specific consent. Section 9 bars behavioural monitoring of children. Section 8(3) asks for accuracy when data drives decisions. Section 4 · Section 6 · Section 9 · Rules 10, 12 · Section 8(3)

Steps
  1. Write the question and the data needed.
  2. Check the purpose people were told.
  3. Mask or aggregate by default.
  4. Get consent for uses outside the purpose.
  5. Set a deletion date for the copy.
Evidence to keep
  • Purpose checks
  • Data specifications
  • Retention entries
Common mistakes
  • Using all data because it is there
  • Analysis copies with no end date
  • Profiling children

How do we build privacy into a new product or system?

Short answer: Four questions at design, then build the basics in

Ask four questions at the design stage: what personal data, why, for how long, and who sees it. Then build in what the Act assumes: a notice, consent where needed, role-based access, logs of record views, a way to delete, and a way to answer requests. Adding these after launch costs far more.

From your seat: Data and analytics head. Work with masked or aggregated data wherever the question allows.
What the law says

Sections 5 to 12 assume systems that can show notices, record consent, limit access, log, delete and answer requests. Section 5 · Rule 3 · Section 6 · Section 8(5) · Rule 6 · Section 8(7) · Rule 8 · Sections 11–14 · Rule 14

Steps
  1. Add the four questions to design templates.
  2. Drop data you do not need.
  3. Build access roles and record-view logs.
  4. Build deletion and request export.
  5. Register the system in the inventory at launch.
Evidence to keep
  • Design review records
  • Inventory entries
  • Deletion tests
Common mistakes
  • Privacy review only at launch
  • Collecting 'in case we need it'
  • No delete function

Do we need a Data Protection Impact Assessment?

Short answer: Required only for SDFs; good practice for high-risk changes

By law, only Significant Data Fiduciaries must do a yearly impact assessment. For everyone else it is good practice for high-risk changes: new tracking, profiling, large data sharing, children's data or moving data abroad. A short assessment before the change is cheaper than fixing it later.

From your seat: Data and analytics head. Profiling and automated decisions deserve an assessment.
What the law says

Section 10 and Rule 13 require a yearly impact assessment for Significant Data Fiduciaries. Section 10 · Rule 13 · Section 9 · Rules 10, 12 · Section 16 · Rule 15

Steps
  1. Define what counts as a high-risk change for you.
  2. Use a two-page template.
  3. Record the data, purpose, risks and safeguards.
  4. Get the DPO's view before approval.
  5. Keep it with the project record.
Evidence to keep
  • Assessment template
  • Completed assessments
  • DPO views
Common mistakes
  • Treating it as paperwork after launch
  • A 40-page template nobody completes
  • No trigger list

Do we process children's data, and what changes if we do?

Short answer: Check every channel; children often appear where you least expect

Anyone under 18 is a child under the Act. For a child's data you need verifiable consent from a parent or lawful guardian, and you must not track, behaviourally monitor or show targeted ads to children. Some classes and purposes are exempt under Rule 12 and the Fourth Schedule, for example healthcare to the extent needed to protect the child's health, and educational institutions for their educational work.

From your seat: Data and analytics head. No behavioural monitoring of under-18s.
Example in BFSI

Minor savings accounts and student loans involve children's data. Parent consent and no targeted offers apply.

What the law says

Section 9 sets the duties. Rule 10 explains how to verify the parent. Rule 12 and the Fourth Schedule list the exemptions. Section 9 · Rules 10, 12 · Section 6

Steps
  1. Find where children's data enters: customers, dependants, interns, visitors, scholarships, app sign-ups.
  2. Decide whether an exemption in the Fourth Schedule applies to that purpose.
  3. Where none applies, add an age question and a parent-consent step.
  4. Switch off tracking and targeted ads for under-18 users.
  5. Record the decision for each channel.
Evidence to keep
  • Channel-by-channel note on children's data
  • Parent-consent records
  • Ad and tracking settings
Common mistakes
  • Assuming 'we are B2B, so no children'
  • Using the age 13 or 16 from foreign laws
  • Treating a tick-box from the child as parental consent

How long can we keep personal data?

Short answer: For the legal or business period, then erase

Keep data for as long as its purpose needs, or as long as a law requires, and then erase it. Every organisation must keep personal data and logs for at least one year under Rule 8(3). Write a retention schedule by record type, with the law or reason against each period.

From your seat: Data and analytics head. Analytics copies often outlive the source. Give them their own period.
Example in BFSI

PMLA and RBI KYC rules set the floor at five years for identity and transaction records.

What the law says

Section 8(7) asks for erasure when the purpose is over, unless a law requires retention. Rule 8(3) sets a one-year minimum for personal data, traffic data and logs. Section 8(7) · Rule 8 · Section 8(5) · Rule 6

Steps
  1. List the record types you hold.
  2. Write the period for each, with the law, regulator rule or business reason.
  3. Set a trigger for the period to start: end of relationship, date of transaction, exit date.
  4. Automate deletion where you can; for paper, schedule shredding.
  5. Keep a deletion log.
Evidence to keep
  • Retention schedule approved by Legal
  • Deletion log
  • Shredding or disposal certificates
Common mistakes
  • 'Keep everything forever' because storage is cheap
  • Deleting before the legal minimum
  • Forgetting email, shared drives and backups

Where does personal data live in our organisation?

Short answer: Start with one row per system

Usually in more places than anyone expects: core systems, email, shared drives, laptops, vendor systems, backups, test copies, spreadsheets and paper. A simple inventory, one row per system, is the base for every other duty.

From your seat: Data and analytics head. Register every data set and its source.
Example in BFSI

Include data warehouses, CRM, call centre tools and partner platforms, not only core banking.

What the law says

Every duty in Sections 5 to 12 assumes you know where the data is. Rule 6 needs safeguards for each system, and Section 11 needs you to find the data when someone asks. Section 8(5) · Rule 6 · Sections 11–14 · Rule 14 · Section 8(7) · Rule 8 · Section 3

Steps
  1. List systems, then shared drives, email, spreadsheets and paper stores.
  2. For each one, note whose data, which items, purpose, owner, hosting location and vendors.
  3. Add copies: backups, test, analytics.
  4. Get each owner to confirm their rows.
  5. Update it whenever a system is bought or retired.
Evidence to keep
  • Data inventory
  • Owner confirmations
  • Change log
Common mistakes
  • A 200-column spreadsheet nobody finishes
  • Leaving out SaaS tools bought by departments
  • No owner for each row

What must a vendor contract say about personal data?

Short answer: Yes, every vendor that touches personal data

You stay responsible for what your vendors do with personal data. The contract should say what data they get, for what purpose, the security they must keep, how fast they must tell you about an incident, that sub-contractors need your approval, and how data is returned or deleted at the end.

From your seat: Data and analytics head. Analytics and cloud vendors receive large copies. Check their terms.
Example in BFSI

Collection agencies, BCs, DSAs, KYC vendors, card processors and the core banking vendor all need data schedules aligned with RBI outsourcing rules.

What the law says

Section 8(1) keeps responsibility with you. Section 8(2) allows a processor only under a valid contract. Rule 6 asks for security terms in that contract. Section 8(1)–(2) · Section 8(5) · Rule 6 · Section 8(6) · Rule 7 · Section 8(7) · Rule 8

Steps
  1. List vendors who receive or can see personal data.
  2. Rank them by how much and how sensitive.
  3. Add a data-protection schedule to each contract, starting with the top ten.
  4. Ask for evidence: certificates, test results, deletion confirmations.
  5. Review the top vendors every year.
Evidence to keep
  • Vendor register
  • Signed data-protection schedules
  • Annual review notes
Common mistakes
  • Relying on the vendor's standard terms
  • No incident-notice time
  • No exit and deletion clause

The sections you will use most

Other rules that sit alongside DPDP

RuleWhat it saysWhat it means alongside DPDPSource
Prevention of Money-laundering Act, 2002 and RBI KYC Master Direction, 2016Keep transaction records for at least five years from the transaction, and identity records for at least five years after the relationship ends.These periods override an erasure request. Explain the retention to the customer and stop every other use.RBI KYC Master Direction
RBI Master Direction on IT Governance, Risk, Controls and Assurance Practices, 2023In force from 1 April 2024 for commercial banks, larger NBFCs, credit information companies and all-India financial institutions. Requires IT governance under the board, audit trails, logging and incident reporting to CERT-In and RBI.Most of the DPDP security duty is already here. Map controls once and use the evidence for both.RBI
RBI Master Direction on Outsourcing of IT Services, 2023The regulated entity stays responsible for outsourced IT, with contracts, audit rights and exit plans.Line up DPDP processor contracts with this direction, so one schedule meets both.RBI
RBI direction on storage of payment system data, 2018All data relating to payment systems must be stored only in India.This is stricter than DPDP Section 16, and it continues to apply.RBI
RBI rules on card storage and tokenisation (from 1 October 2022)Only card issuers and card networks may store actual card data. Others use tokens, created with the cardholder's explicit consent.Check logs, call recordings and support tickets for card numbers.RBI
RBI (Digital Lending) Directions, 2025Collect only need-based data with prior explicit consent and an audit trail. Apps should not access contacts, files, media or call logs; one-time access to camera, microphone or location is allowed for onboarding or KYC with consent.Your app permissions and lending partner contracts are where DPDP and RBI meet.RBI
IRDAI Information and Cyber Security Guidelines, 2023Report cyber incidents to CERT-In within six hours, and to IRDAI within 24 hours of the CERT-In report.One incident plan should run the CERT-In, IRDAI and Data Protection Board steps together.IRDAI
SEBI Cybersecurity and Cyber Resilience Framework (CSCRF), 2024Security, logging and incident-reporting duties for SEBI-regulated entities. Stock brokers and depository participants report cyber incidents within six hours.Use CSCRF evidence for DPDP security, then add notices, consent and rights.SEBI
CERT-In Directions, 28 April 2022Report specified cyber incidents within six hours. Keep ICT logs for 180 days within India. Sync clocks to Indian time sources.Applies to every BFSI entity in addition to the regulator's own clock.CERT-In
Credit Information Companies (Regulation) Act, 2005Governs what credit information is shared with credit bureaus and how errors are corrected.Credit bureau sharing has its own law; DPDP rights requests about bureau data should point to that process too.Act
RBI Integrated Ombudsman Scheme, 2021Customers can escalate unresolved complaints to the RBI Ombudsman.Privacy complaints may reach both the Ombudsman and the Data Protection Board. One complaint log helps.RBI
Explore our research-built assessment platformsDPDP Insights helps you understand the law. DPDPiq is where an organisation's evidence is assessed. Each platform comes out of the same Foundation Layer research. No AI, API or outside service is used when you use this site. Everything runs from InfraVeritas360's own server.