Tomorrow, 13 November 2026, the DPDP Rules turn one year old. That day, Rule 4 on Consent Managers comes into force. The duties that touch an online retailer's customers follow on 13 May 2027, which is 182 days from today.
E-commerce has a second date in between. The Consumer Protection (E-Commerce) (Amendment) Rules, 2026 were notified on 9 September and take effect on 1 January 2027. As a Bar & Bench analysis explains, new Rule 5(6) restricts two uses of consumer information by a marketplace. One is selling goods that carry the marketplace's own brand or name. The other is promoting a seller as associated with the marketplace.
Both regimes start with the same question. Where does your customer's data actually sit? Most platforms can answer for the order database. Few can answer for the rider app, the returns vendor, a seller's downloaded CSV or last festive season's staging copy. That gap is why data discovery has to come before every other DPDP control. The rest of this article explains why, and how to run it before May.
Why e-commerce data travels further than most
A bank keeps most customer data inside its own core systems. An online retailer hands pieces of every order to other businesses within hours. The payment gateway sees the transaction. A logistics aggregator passes the address and phone number to a courier, whose rider sees them on a handheld screen.
Cash on delivery stretches the chain further. Meesho's December 2025 prospectus put cash on delivery at 72% of its orders in the six months to September 2025. The company also noted that COD requires physical cash handling by last-mile partners. Each such order puts a name, address, phone number and amount in hands outside the platform.
Returns add another loop. A reverse pickup creates a second address record, and often photos of the item and its label. COD refunds often need a bank account or UPI ID that the original order never captured. Sellers on a marketplace download order reports to pack and ship. Support agents copy customer details into ticketing tools, call recordings and chat threads.
None of this is wrong. It is simply how online retail works. It does mean one customer's data may live in many systems, several of them run by other companies. Section 8(1) of the Act keeps the legal duty with the Data Fiduciary for processing done on its behalf by a Data Processor. That holds irrespective of any agreement to the contrary.
Every DPDP duty starts with a location question
No section of the Act is titled data discovery. Yet almost every duty in it assumes you know where personal data sits. The table shows the dependency for the duties that matter most to a retailer.
| Duty | Provision | What you must already know |
|---|---|---|
| Notice | Rule 3(b) | An itemised list of the data you collect |
| Consent withdrawal | Section 6(4), 6(6) | Which systems and processors must stop |
| Security safeguards | Section 8(5), Rule 6 | Which stores to encrypt, mask, log and back up |
| Breach intimation | Section 8(6), Rule 7 | Whose data sat in the affected system |
| Erasure | Section 8(7) | Every copy, including at processors |
| Processors | Section 8(2), Rule 6(1)(f) | Every vendor that touches customer data |
| Access requests | Section 11(1)(b) | Everyone you shared data with, and what |
| Children | Section 9 | Which records belong to children |
Breach intimation. Rule 7(2) requires an intimation to the Board without delay. Within 72 hours of becoming aware, you must add detailed information, the broad facts and a report on intimations to affected Data Principals. Rule 7(1) requires an intimation to each affected Data Principal. If a courier partner's server is compromised, you can only reach the right people if you know which fields you sent it.
Consent withdrawal. This is subtler than most teams assume. An illustration to Section 6(5) uses an online shopping app. A customer who withdraws consent can be stopped from placing new orders. Processing for goods she has already ordered and paid for continues. Section 6(6) then requires you to stop processing, and make your processors stop, within a reasonable time. That only works if each dataset is tied to a purpose and a system.
Erasure. Section 8(7) carries an illustration written for this sector. A woman sells her used car through an online marketplace. Once the sale concludes, the marketplace must not retain her personal data any longer. Section 8(7)(b) extends the duty to data given to processors. The duty gives way only where another law requires retention.
Access requests. Gaps here show up in writing to customers. Section 11(1)(b) entitles a customer to the identities of all other Data Fiduciaries and Data Processors you shared her data with. You must also describe what was shared. An incomplete vendor list produces a wrong answer to the person most likely to complain. Rule 14(3) sets ninety days as the outer limit for responding to grievances.
Children. Section 9(1) requires verifiable consent of a parent before you process a child's personal data. Section 9(3) bars tracking, behavioural monitoring and targeted advertising directed at children. A retailer selling toys, school supplies or teen fashion is likely to hold children's data in accounts, gift messages and reviews. Unless that data is found and flagged, the recommendation engine treats it like everyone else's.
The stakes are uneven across these duties. The Schedule to the Act sets its highest ceiling, ₹250 crore, for failing to take reasonable security safeguards under Section 8(5). Safeguards that leave a forgotten data store open are hard to defend as reasonable.
The three-year clock for large platforms
Rule 8 and the Third Schedule add a duty for a smaller group. It covers an e-commerce entity with not less than two crore registered users in India. The Schedule takes its meaning of e-commerce entity from the Consumer Protection Act, 2019. It excludes a seller offering goods or services on a marketplace.
For such a platform, personal data must be erased once the user goes three years without approaching it. The period runs from the latest of three events. These are her last approach for the specified purpose, her last exercise of rights, and the commencement of the Rules. Retention required by another law is preserved. Two purposes are also carved out: access to her user account, and access to virtual tokens usable for money, goods or services. For a retailer, wallet balances and gift cards are the obvious candidates.
Rule 8(2) adds a warning. At least 48 hours before the period ends, the platform must tell the user her data will be erased. She can stop it by logging in, initiating contact or exercising her rights.
Each practical consequence is a discovery problem.
First, "last approached" has to be worked out for every user from every system that records contact. Section 8(11) counts contact she initiates in person or by electronic or physical communication. App logins, orders and support tickets usually sit in different stores, often under different customer IDs.
Second, erasure has to reach every copy. A record purged from the order database but alive in the warehouse, the CRM and a seller's export hasn't been erased.
Third, the Rules also pull the other way. Rule 8(3) requires every Data Fiduciary to keep personal data, associated traffic data and processing logs for at least one year. Its illustration has an e-book platform keeping order, payment and delivery logs for a year, even after the buyer deletes her account. Your retention engine must apply both rules at once, store by store.
The Rules don't say which commencement date starts the three-year period. On the earliest reading, 13 November 2025, no account reaches three years before 13 November 2028. That isn't a reason to wait. The last-contact dates you rely on in 2028 must be captured correctly from 2027.
Discovery, classification and mapping are three different jobs
Teams often use the three words as one. They are separate jobs with separate outputs. Programmes stall when the jobs blur. We have set out the distinction in our piece on discovery, classification and mapping. In short:
Discovery answers "where". It finds every system, table, file, bucket, log and vendor store holding personal data, including stores nobody listed.
Classification answers "what". It labels each finding by data type, purpose, owner and sensitivity. It also flags children's data and any law that requires the record to be kept.
Mapping answers "how it moves". It traces each flow from collection to processors, sellers and other recipients, with retention at every stop. The result feeds your record of processing activities.
Order matters. You can't classify a store you haven't found. A map drawn from interviews, without discovery beneath it, records what people believe happens.
Where personal data hides in an online retail stack
Architecture diagrams and interviews find the systems people remember. In the programmes we've run, the exposure usually sits in places neither captures.
Free-text fields. Delivery instructions, gift messages, reviews and chat transcripts collect what no form asked for. Customers type alternate numbers, landmarks, family names and, on pharmacy orders, medical details.
Logs. Application, gateway and CDN logs often capture phone numbers or emails in request parameters. Rule 6(1)(e) requires logs to be kept for a year, so they need the same protection as the order database.
Backups and snapshots. Rule 6(1)(d) expects backups for continuity. Backups also preserve data you've erased from production. Your erasure design needs a stated position on every backup set.
Test and staging databases. Production copies taken to debug a sale-day failure tend to outlive the bug. They rarely carry the access control and masking that Rule 6(1)(a) and (b) expect.
Seller exports. Marketplace sellers download order reports and shipping labels to laptops and inboxes outside your systems. Whether a seller acts as your processor or as a separate fiduciary depends on the arrangement. Either way, Section 11(1)(b) puts the seller on your map.
Vendor systems. Logistics aggregators, couriers, payment gateways, call centres, CRM tools, messaging providers and advertising pixels each hold a slice. Your accounts payable ledger is often the most complete list of them.
The data warehouse. Customer tables get joined, copied and exported into dashboards and spreadsheets. Each copy counts as a store for erasure and breach purposes.
A discovery programme in seven steps
This is the sequence we use with retailers and marketplaces. Each step produces something the next step needs.
- Fix scope and ownership. List every legal entity, brand, app and website that collects customer data. Give one accountable owner, usually the DPO or CISO, authority across engineering, operations and marketing.
- Build the inventory from three directions. Start with the architecture register and the list of cloud accounts. Add the accounts payable vendor list and SaaS tools paid on corporate cards. Then ask each team which exports and spreadsheets they use every week.
- Define what you are looking for. Write detection patterns for Indian mobile numbers, email addresses, PIN codes, UPI IDs, order IDs and customer IDs. Rule 14(5) offers a starting list of identifiers, such as customer identification numbers and application reference numbers.
- Scan structured and unstructured stores. Run automated scans across databases, object storage, logs, backups and non-production environments. Sample free-text fields and shared drives instead of assuming they are clean.
- Classify every finding. Record data type, purpose, owner, retention rule and whether it concerns a child. Flag the stores that the Third Schedule clock or Rule 8(3) will touch.
- Map flows to processors and sellers. For each outbound flow, record the recipient, the fields sent and the contract required by Section 8(2). Check that the contract carries the security provision Rule 6(1)(f) expects. Build the result into your data map.
- Keep it current. Tie discovery to change management, so a new vendor, table or campaign tool triggers a scan. Re-run full scans on a fixed cycle and before each major sale event.
If you do one thing this month, start step 2 with your finance team. A vendor you pay but never listed is the gap we find most often. It is also the cheapest to close early.
A structured DPDP gap assessment at the start of the programme shows how far your current inventory is from complete. It also tells you which of the seven steps will take longest in your organisation.
What to finish before 13 May 2027
Twenty-six weeks is less time than it sounds. Sale events, code freezes and vendor contract renewals all fall inside it. Work back from the deadline.
| Date | What happens | Where you should be |
|---|---|---|
| 13 Nov 2026 | Rule 4 on Consent Managers in force | Scope, owner and inventory under way |
| 1 Jan 2027 | E-Commerce Amendment Rules in force | Rule 5(6) data uses identified |
| 31 Jan 2027 | Internal milestone | Discovery and classification complete |
| 31 Mar 2027 | Internal milestone | Notices, consent, rights and breach runbook built on the map |
| 13 May 2027 | Rules 3, 5 to 16, 22 and 23 in force | Controls live and tested |
| 13 Nov 2028 | Earliest three-year mark under the Third Schedule | Erasure and 48-hour warnings running |
The months from February to May are for building on the map. Notices come straight from the itemised inventory. Consent management has to be tied to purposes and systems. A rights request process must be able to find every copy of a customer's data. Our sequencing plan for 13 May 2027 sets out the full order of work.
Large platforms should watch one more date. On 22 January 2026, Business Standard reported that MeitY had proposed a shorter window for Significant Data Fiduciaries. For certain provisions, it would be 12 months from notification instead of 18. The newspaper's sources expected large technology firms, Amazon among them, to be classified as SDFs.
We are in touch with the industry to further compress time required for compliance
— Ashwini Vaishnaw, Union IT Minister, as quoted by Business Standard
As of 12 November 2026, we have found no Gazette amendment to the commencement dates. Plan as if the proposal could still be notified. Finishing discovery early costs little if the date holds. It saves the programme if the date moves.
The question to ask before approving any DPDP spend
Before a Board approves a consent platform or a breach runbook, it should ask one question. Can we list every place our customers' data sits today, including at vendors and sellers? If the answer is no, every other control rests on a guess.
Discovery is the least visible part of a DPDP programme. Every other part stands on it.
This article is general information. It isn't legal advice for any specific situation.
Sources
- India Kanoon, "Section 6 in The Digital Personal Data Protection Act, 2023", 11 August 2023. Link
- India Kanoon, "Section 8 in The Digital Personal Data Protection Act, 2023", 11 August 2023. Link
- India Kanoon, "Section 11 in The Digital Personal Data Protection Act, 2023", 11 August 2023. Link
- ApniLaw, "Section 9 - Digital Personal Data Protection Act (DPDP) - Processing Of Personal Data Of Children", 11 August 2023 (Act text). Link
- DPDPA.com, "Rule 1 of Digital Personal Data Protection Rules, 2025", 13 November 2025. Link
- DPDPA.com, "Rule 3 DPDP Rules 2025", 13 November 2025. Link
- DPDPA.com, "Rule 6 of DPDP Rules 2025", 13 November 2025. Link
- DPDPA.com, "Rule 7 of DPDP Rules 2025", 13 November 2025. Link
- DPDPA.com, "Rule 8 of DPDP Rules 2025", 13 November 2025. Link
- DPDPA.com, "Rule 14 of DPDP Rules 2025", 13 November 2025. Link
- DPDPA.com, "Schedule 3 of Digital Personal Data Protection Act, 2023 DPDP Rules 2025", 13 November 2025. Link
- Data Security Council of India, "FAQs on Reasonable Security Safeguards", 2025. Link
- Business Standard, "Meity may cut compliance timeline for key DPDP rules to 12 months", 22 January 2026. Link
- MediaNama, "Meesho IPO: Offer Size, Financial Report Card, Risk Factors & Other Takeaways", 2 December 2025. Link
- Bar & Bench, "Consumer Protection (E-Commerce) (Amendment) Rules 2026: From consumer disclosure to platform governance", 28 September 2026. Link