AI-Driven Segmentation in Indian Retail Loyalty: The DPDP Compliance Playbook

DPDP 2023 Overview and Its Impact on Retail Loyalty Data

The Digital Personal Data Protection Act, 2023 is India's first comprehensive data privacy legislation and it borrows structural concepts from GDPR while adapting them to the Indian retail and digital commerce reality. The Act defines a 'Data Principal' (the customer), a 'Data Fiduciary' (the brand or mall collecting data) and specifies that processing personal data is lawful only when the Data Principal has given informed, specific, unambiguous consent — or when processing falls within a narrow set of 'legitimate uses' that do not include marketing communications or AI-driven profiling.

For a retail loyalty operator, the practical impact is significant across four dimensions. First, consent must be granular: a customer enrolling in a Pantaloons loyalty card cannot be assumed to have consented to their purchase history being shared with a co-tenanted brand inside the same mall ecosystem. Each processing purpose — transactional reward calculation, behavioural profiling, cross-brand offers, third-party data sharing — requires its own consent signal. Second, consent must be revocable: the Data Principal has the right to withdraw consent at any time, and the Data Fiduciary must stop processing within a reasonable period. For a real-time AI segmentation engine, this means member records must carry live consent flags that the model honours in every scoring cycle. Third, the Act mandates data minimisation: you may collect only what is necessary for the stated purpose. Loyalty programmes that have historically vacuumed up every available attribute — from vehicle registration to social media handles — will need to audit and prune their data models. Fourth, and critically for AI segmentation, the Act's provisions on automated decision-making mean that personalised offers and tier upgrades driven purely by algorithmic scoring, without human review options, may need a consent and transparency layer of their own.

Indian retail operators who have been benchmarking against Capillary, EasyRewardz or MoEngage need to ask a pointed question of every vendor: does your data collection architecture produce a per-member, per-purpose consent record that can be produced to a Data Protection Board on request? The answer, for most legacy CRM stacks, is currently no. Brands like Café Coffee Day or Lifestyle, running high-volume loyalty programmes with millions of enrolled members, face the most acute exposure because the sheer size of their member base amplifies non-compliance risk. DPDP penalties can reach ₹250 crore per breach instance — not a theoretical risk that can be deferred.

The strategic implication is this: the brands and mall operators that invest now in a consent-native data infrastructure will not merely become compliant — they will own the cleanest, most trustworthy first-party data asset in Indian retail. And clean, consented data is not a constraint on AI segmentation; it is the precondition for AI segmentation that actually predicts behaviour rather than amplifying historical noise.

Importance of Consent Management in AI Segmentation

Most Indian retail marketers think of consent management as a one-time event: the checkbox at sign-up, the SMS opt-in at POS. This mental model is catastrophically wrong for AI-driven segmentation, and understanding why requires looking at how modern cohort engines actually work.

An AI segmentation model in a mature loyalty programme does not simply group members by spend tier. It continuously re-scores every member across multiple dimensions — purchase frequency, category affinity, channel responsiveness, churn probability, occasion proximity, household composition signals — and reassigns members to dynamic cohorts that change weekly or even daily. A customer who visits a Lenskart kiosk inside a mall twice in 30 days may be cohorted as a 'high-intent eyewear upgrader' and immediately targeted with a personalised bundle offer. A Tanishq member whose purchase gap has crossed 14 months may be moved into a 'lapsed high-value' cohort triggering a win-back sequence. These are not static segments; they are living, model-driven classifications that draw on every data signal the platform can access.

Here is the consent problem: each of those data signals — dwell time from WiFi probes, cross-category purchase signals from a mall-wide loyalty scheme, device identifiers from the loyalty app, historical communication response rates — may have a different consent basis. If a member consented to 'earn and redeem points' but not to 'behavioural profiling for personalised offers,' then using their dwell time data to power an AI affinity model is a DPDP violation, regardless of how technically sophisticated the model is. The AI engine does not know what it is not supposed to know unless the consent layer tells it.

This is why consent management must be architecturally upstream of the AI segmentation engine, not downstream of it. Platforms like Xeno, WebEngage or Almonds.ai provide strong campaign execution and segmentation tooling, but they typically receive data from a CRM or CDP that was populated before DPDP requirements were codified. The consent metadata often does not travel with the member record. The result is a segmentation engine that is technically capable but legally exposed.

A ConsentFirst architecture — the model Fundle's platform is built on — inverts this relationship. Every member attribute entering the system carries a consent tag: what purpose it was collected for, when consent was given, and whether it remains active. The AI segmentation engine only reads attributes where consent is active for the relevant processing purpose. When a member withdraws consent for behavioural profiling, their record is immediately flagged, and the next model scoring cycle excludes those attributes. The cohort membership of that member updates automatically. This is not compliance theatre; it is consent-aware AI that produces outputs the brand can stand behind in a regulatory audit.

How Fundle's ConsentFirst CMP Ensures Compliance

Fundle's ConsentFirst CMP is the consent and data governance layer built natively into the Fundle AI Platform. It was designed from the ground up to meet the structural requirements of DPDP 2023, not to retrofit GDPR patterns onto Indian retail workflows. The distinction matters because Indian retail loyalty operates very differently from European e-commerce: enrolments happen predominantly at physical POS terminals, QR codes on shopping bags, or mall kiosks — not via a digital cookie banner. The CMP must work in low-bandwidth environments, be operable by a retail floor associate in under 45 seconds, and produce a legally defensible consent record without requiring the customer to navigate a complex multi-screen flow.

Fundle's ConsentFirst CMP ensures DPDP-compliant data collection for AI segmentation across 123+ malls and 270+ brands. That scale is not incidental — it reflects the architecture's ability to handle multi-brand, multi-operator consent complexity. When a customer is a member of a mall-level loyalty programme that aggregates data from Reliance Trends, FabIndia, Apollo Pharmacy and a dozen F&B brands under one roof, each brand may be a separate Data Fiduciary with its own consent obligations. The ConsentFirst CMP manages this by maintaining a consent graph — a per-member, per-brand, per-purpose consent map — that the Fundle AI Platform queries before any cross-brand segmentation signal is used.

The CMP operates across three consent collection touchpoints. At enrolment, a simplified consent notice in the member's preferred language (Hindi, Tamil, Kannada and 6 other regional languages are supported) presents processing purposes as distinct, toggleable permissions — not a single blanket consent. The associate or the digital enrolment flow captures these as structured consent records, not free-text acknowledgements. Second, at each re-engagement point — app login, offer redemption, annual programme renewal — the CMP surfaces a consent refresh prompt, ensuring that long-standing members who enrolled before DPDP came into force have an opportunity to review and update their permissions. Third, the Fundle loyalty app includes a self-service consent dashboard where members can view exactly what data is held, for what purpose, and withdraw specific consents without needing to contact customer care.

On the AI side, Fundle AI Agents query the consent graph before executing any segmentation or personalisation workflow. If a Fundle AI Workflow is designed to build a 'high-value lapsed member' cohort using cross-category purchase data, the workflow will automatically exclude members who have not consented to cross-brand data sharing, build the cohort from permissioned attributes only, and log the consent basis used in the cohort's metadata. This creates a fully auditable, consent-aware segmentation operation that marketing teams can run at the pace of AI without waiting for legal sign-off on every campaign.

Balancing Personalization with Data Privacy in Indian Retail

There is a persistent myth in Indian retail marketing circles that privacy regulation and personalisation are in fundamental tension — that DPDP compliance necessarily means blunter, less effective segmentation. The data does not support this. In fact, consent-based first-party data consistently outperforms inferred or scraped data in AI segmentation models across every metric that matters: prediction accuracy, campaign response rate, and revenue per communication.

The reason is straightforward. A customer who has explicitly consented to behavioural profiling for personalised offers is, by definition, signalling intent and receptivity. They have cognitively opted in to a value exchange: their data for relevant, useful communications. This self-selection effect means that consented cohorts have structurally higher engagement baselines than cohorts built on data the member is unaware is being used. When Manyavar runs a pre-wedding season cohort using only members who have consented to occasion-based profiling, the cohort is smaller but the conversion rate is materially higher than a broad 'male, 25-40, high spender' segment built from unaudited transaction data.

The practical playbook for balancing personalisation and privacy operates at three levels. At the programme level, the value proposition of consent must be made explicit to members: 'Share your preferences and we'll make sure you never miss an offer that matters to you' converts better as a consent prompt than a legal notice. Brands like Tanishq have long understood that their customers value the feeling of being recognised and remembered — consent framing that speaks to this emotional benefit sees opt-in rates 15-20 percentage points higher than compliance-language framing in A/B tests.

At the data model level, privacy-preserving AI techniques — federated learning, differential privacy, cohort-level modelling rather than individual-level scoring — allow brands to extract segmentation signal from data without requiring individualised profiling consent for every inference. A mall operator running Fundle Mall Loyalty can model cross-category affinity at cohort level, identifying that 'members who shop at athleisure brands also index highly for premium food court spend' without needing to process individual members' complete purchase histories under a behavioural profiling consent basis. At the execution level, consent-aware personalisation means that every member receives communications calibrated to their consent scope — full hyper-personalisation for those who have opted into profiling, category-level recommendations for those who have consented to transactional data use only, and generic benefit reminders for those with minimal consent. This tiered approach maximises reach while never breaching consent boundaries.

How Fundle solves this

Fundle's approach to AI-driven segmentation in Indian retail loyalty is built on a single architectural principle: consent is the first data point, not the last compliance check. Vineet Narang's founding vision for Fundle was that Indian retail's loyalty programmes had been built on borrowed time — accumulating rich data assets whose legal basis was fragile and whose consumer trust foundation was untested. DPDP 2023 has simply made that fragility visible. Fundle was built to be the platform that makes consented, AI-powered loyalty the default, not the exception.

The Fundle AI Platform integrates the ConsentFirst CMP directly into the loyalty enrolment and engagement stack. There is no separate compliance module to sync, no third-party CMP that needs to be API-connected to the segmentation engine. The consent graph is native to the Fundle Loyalty data model, which means that Fundle Brand Loyalty programmes for individual retail chains and Fundle Mall Loyalty programmes for multi-brand mall ecosystems both operate on the same consent-aware infrastructure. A Fundle Mall Loyalty deployment at a 200-brand mall property handles the multi-fiduciary complexity of cross-brand consent automatically, maintaining distinct consent records for each brand while allowing the mall operator to run aggregated, consented cohort analysis across the portfolio.

Fundle AI Agents are the execution layer. These are purpose-built AI agents that operate within predefined consent boundaries, executing segmentation, personalisation and campaign triggering workflows without requiring manual compliance review for each action. A Fundle AI Agent running a 'lapsed high-value member win-back' workflow will, before building the cohort, query the consent graph, exclude members who have withdrawn profiling consent, apply data minimisation rules to the attributes used, and log the consent basis in the workflow execution record. This is Fundle Agentic AI in practice: autonomous enough to operate at marketing speed, governed enough to operate within legal boundaries.

Fundle AI Workflow allows marketing teams to design segmentation and personalisation workflows visually, with consent guardrails embedded as workflow steps rather than external approvals. A retail marketing head at a chain like Lifestyle or Pantaloons can build a pre-season re-engagement workflow in hours, with the Fundle AI Workflow engine automatically flagging any step that draws on attributes not covered by the target cohort's consent profile. The result is a marketing operations capability that is simultaneously faster and more compliant than the manual, spreadsheet-and-legal-review processes most Indian retail CMOs are currently running. Fundle.ai is not just making compliance manageable; it is making consented AI segmentation the competitive advantage that it should always have been.