Fundle.ai: Loyalty Data Insights AI Under India's DPDP 2023

Overview of DPDP 2023 and Its Impact on Loyalty Data

The Digital Personal Data Protection Act 2023 rests on eight core principles: lawful processing, purpose limitation, data minimisation, accuracy, storage limitation, security safeguards, accountability and the rights of data principals. For a loyalty programme manager, each of these lands somewhere specific in the daily workflow.

Purpose limitation means you cannot collect a customer's purchase history for points calculation and then silently repurpose it for targeted advertising or share it with a co-brand partner without fresh, specific consent. Most legacy loyalty setups in India — whether running on Capillary, EasyRewardz or a homegrown POS-integrated stack from Wondersoft or GoFrugal — were not architected with purpose binding in their data models. Retrofitting that is non-trivial. Data minimisation means you cannot store every field your POS captures indefinitely. If Reliance Trends or Pantaloons collects a customer's date of birth for a birthday bonus, that data must be used only for that purpose and not retained beyond the campaign cycle unless there is a documented fresh consent.

Storage limitation and the right to erasure together create a new operational reality: customers can request deletion of their data, and the loyalty platform must be able to execute that end-to-end — across the transactional database, the campaign engine, the AI model training sets and any third-party integrations with partners like Petpooja for F&B or POSist for restaurant POS data. That is a systems integration challenge most operators have not yet mapped.

The consent mechanism itself must be granular, freely given, specific and revocable. A blanket enrolment checkbox will not hold up. Operators need a Consent Management Platform (CMP) that logs the exact consent version, timestamp, channel, purpose and scope — and that can replay that audit trail on demand for a Data Protection Board inquiry. Brands like Cafe Coffee Day or FabIndia running coalition or co-branded programmes face additional complexity: each brand in the coalition is a separate data fiduciary, and data flows between them require individual consent records, not programme-level assumptions.

The net effect is that DPDP 2023 forces loyalty operators to confront what their data actually is, who it belongs to and what they legitimately have permission to do with it. That discipline, applied correctly, dramatically improves the quality of the data that flows into AI models — because you end up with consented, accurate, purpose-tagged data instead of the noisy, stale, duplicate-riddled records that plague most loyalty databases today.

How AI Extracts Loyalty Data Insights While Maintaining Compliance

The instinctive fear among retail analytics teams is that DPDP compliance will shrink the usable dataset to the point where AI models lose predictive power. In practice, the opposite is true when the architecture is right. Consented, purpose-tagged data is structurally cleaner than raw PII pools, and clean data consistently outperforms large-but-dirty datasets in supervised learning benchmarks.

Consider RFM modelling — the backbone of most loyalty analytics work in India. Recency, Frequency and Monetary scores can be computed entirely on pseudonymised transaction records. The customer ID is hashed; the model sees spend patterns, category mix and visit cadence, not names or mobile numbers. A mid-sized mall running 80 tenants across Apparel, F&B, Entertainment and Services can build segment trees that identify, for instance, a high-frequency F&B visitor with declining apparel spend — a churn signal for the apparel wing — without ever exposing individual identity to the analytics layer. The output (segment assignment, churn score) is then married back to the consent record to determine which communication channel and which offer the customer has authorised the mall to send.

Propensity modelling for cross-category offers works similarly. Apollo Pharmacy's loyalty data, if shared in a coalition arrangement with a wellness brand, requires explicit consent for that cross-brand data flow. But within a single brand's consented dataset, AI can score a customer's likelihood to shift from OTC purchases to a wellness subscription with high accuracy — and trigger a personalised push notification on the app, which is a consented channel, rather than an SMS that the customer may not have opted into for marketing.

AI loyalty analytics in India must also account for the linguistic and behavioural diversity of the customer base. A model trained on tier-1 metro data from Select CITYWALK will not generalise cleanly to a tier-2 mall in Indore or Coimbatore. Fundle's AI Platform handles this through regional model fine-tuning — separate propensity and churn models calibrated on local spending patterns, regional festival calendars and preferred redemption categories. This is not a feature that point-solution vendors in the competitive set — Antavo, Xeno or Customer Capital — have productised for the Indian market at scale.

The critical architectural principle is that the AI layer must be downstream of all compliance gates. This means the model training pipeline must accept only anonymised or pseudonymised inputs, the inference layer must check consent status before triggering any customer-facing action, and every model output that influences a campaign must be logged with a reason code traceable back to a specific consent record. This audit trail is not just a legal requirement; it is the foundation of a repeatable, improvable AI programme.

Building a DPDP-Compliant Loyalty Analytics Framework

A compliance-ready loyalty analytics framework is not a single technology purchase. It is a stack of decisions across data governance, technology architecture, vendor contracts and internal process — and it needs to be operational before the Data Protection Board begins enforcement, not after the first notice arrives.

Start with a data inventory. Map every field your loyalty programme collects: name, mobile, email, date of birth, purchase history, location check-in, browsing behaviour on the loyalty app, tier status, referral activity. For each field, document the stated purpose at collection, the retention period, who inside the organisation accesses it and which third-party systems it flows to — your CRM, your campaign tool (MoEngage, WebEngage), your AI analytics layer, your POS integration (POSist, GoFrugal), and any co-brand partners. This inventory is the baseline from which you design consent architecture and data minimisation policies.

Next, deploy a Consent Management Platform that is natively integrated into every customer touchpoint: the loyalty app, the web enrolment flow, the in-store QR-based sign-up kiosk, and the IVR or WhatsApp onboarding flow. The CMP must present purpose-specific consent options in plain language, in the customer's preferred language (DPDP requires regional language support to be genuinely accessible), and must record the exact version of the consent notice presented. Every subsequent data use must be validated against the active consent record before execution.

Third, redesign your AI model training pipelines for anonymised inputs. Work with your data science team or platform vendor to ensure that model training jobs pull from a pseudonymised replica of the transactional database — never the PII-bearing production database. Implement differential privacy techniques where sample sizes are small enough to risk individual re-identification. For mall operators running Fundle Mall Loyalty, this is handled natively in the platform's data pipeline architecture, removing the burden from the mall's internal IT team.

Fourth, establish a Data Principal Rights workflow. When a loyalty member from Lifestyle or Manyavar submits a deletion request, your systems must be able to execute that deletion — or flag the record for deletion — across every system in the data map within the timeframe DPDP specifies. This requires tested API integrations between your loyalty platform, CRM, campaign tool and any external analytics vendors. Manual processes will not scale and will fail audit. Finally, appoint a Data Protection Officer (DPO) or designate an internal owner with clear accountability. The DPO does not need to be a lawyer, but they must understand the data flows, own the consent records and be the escalation point for breach response.

KPIs to Track for DPDP-Compliant AI Loyalty Analytics

Compliance does not mean you stop measuring commercial outcomes. The right KPI framework runs two parallel scorecards: a compliance health scorecard and a commercial performance scorecard. Both matter. A loyalty programme that is fully compliant but commercially inert is not a success; neither is a high-ROI programme that is one enforcement notice away from ₹250 crore in penalties.

On the compliance side, track consent coverage rate — the percentage of your active loyalty base with valid, purpose-specific consent records on file. Best-in-class programmes should target above 90% consent coverage within 12 months of DPDP operationalisation. Track consent refresh rate: how many members have been re-consented after a material change in data processing purposes. Track erasure request fulfilment time against the statutory deadline. Track the number of AI model training jobs audited for anonymised inputs in the quarter. And track the completeness of your data inventory — percentage of data fields with documented purpose, retention period and consent basis.

On the commercial side, the metrics that matter for an AI-powered loyalty programme are: incremental revenue per active member (target: 15-25% lift over a non-personalised control group in Indian retail benchmarks), redemption rate (a healthy DPDP-era programme should see 35-45% redemption rates when offers are genuinely personalised based on consented behavioural data), churn rate among top-tier members (AI-driven early intervention should keep this below 8% annually for mid-to-large retail chains), and cross-category purchase penetration (the share of members who buy across two or more categories — a direct output of AI-driven cross-sell propensity scoring).

A metric that bridges compliance and commercial performance is the consented data depth score: for each active member, how many purpose-specific consents are active? A member with five consents (analytics, personalisation, email, WhatsApp, third-party wellness partner) is a more valuable AI target than a member with one (basic points). Driving consented data depth through value exchange — better offers, exclusive access, personalised experiences — is the commercial strategy that makes DPDP compliance a growth driver rather than a cost centre.

How Fundle solves this

Fundle AI Platform was architected from the ground up for the consent-first, AI-driven loyalty era that DPDP 2023 is forcing Indian retail into. Where legacy platforms treat compliance as a module bolted onto an analytics engine built in a different regulatory era, Fundle treats the consent record as the primary data object — everything downstream, including AI model inputs, campaign triggers and partner data flows, is gated by the consent state of the individual member.

Fundle's ConsentFirst CMP is the operational centrepiece of this approach. It is embedded natively in Fundle Mall Loyalty and Fundle Brand Loyalty enrolment flows — app, web, WhatsApp, in-store QR — and it presents purpose-specific consent options in 12 Indian languages. Every consent event is logged with version, timestamp, channel and member ID. When a loyalty member revokes a consent, the propagation across Fundle's AI pipeline, campaign engine and partner API layer is automated within minutes, not days. This is what makes the claim credible: Fundle's ConsentFirst CMP ensures DPDP compliance while powering insights for 123+ malls.

Fundle AI Agents operate on anonymised behavioural signals — visit frequency, category spend mix, redemption patterns, app engagement — to generate churn risk scores, cross-sell propensity rankings and next-best-offer recommendations. These agents are purpose-bound: an agent trained on F&B transaction data cannot access or influence apparel campaign decisions unless the member has consented to cross-category personalisation. Fundle Agentic AI goes further, running autonomous campaign optimisation loops that test offer variants, measure incremental lift and reallocate campaign budget in real time — all within the guardrails of the member's active consent scope.

Fundle AI Workflow connects the compliance and commercial layers operationally. When a mall operator at Phoenix Marketcity or a retail CMO at a Lifestyle-scale chain needs to run a Diwali cross-sell campaign, the workflow engine checks consent coverage across the target segment, flags members who lack the required marketing consent, routes those members to a re-consent journey, and releases the campaign only to the verified-consented segment. The audit log generated at each step is exportable for regulatory review. Vineet Narang's founding vision for Fundle was that AI in Indian retail loyalty should be commercially powerful precisely because it is trustworthy — and that trustworthiness is an architecture decision, not a policy statement. For retail CMOs and loyalty programme managers navigating DPDP 2023, Fundle's integrated approach is the difference between loyalty data insights AI that compounds in value over time and a compliance liability waiting to be triggered.