Fundle.ai: AI First Party Data Platform for Retail Loyalty in India

Marketing Challenges for Indian Retail Loyalty

Walk into the marketing war-room of any mid-to-large Indian retailer and you will find a specific kind of chaos. There are three to five separate systems capturing customer data: the POS (often POSist, Petpooja, or Wondersoft in food and fashion), a mobile app, a WhatsApp Business API connector, a rudimentary CRM — sometimes EasyRewardz or Capillary — and a campaign tool like MoEngage or WebEngage. Each system has its own customer ID. Nobody has done identity resolution at scale. The result is that the same shopper who bought ethnic wear at Manyavar, grabbed a coffee at Cafe Coffee Day in the same mall, and then browsed eyewear at Lenskart appears as four different people across four databases.

The second challenge is consent and compliance. The DPDPA 2023 requires explicit, purpose-specific consent before any personal data is processed for marketing. This is not the GDPR copy-paste that legal teams initially feared, but it is real, and it has teeth. Mall operators running programs across 200+ brand tenants face a particular compliance complexity: who is the data fiduciary — the mall or the brand? How is consent inherited or re-sought when a shopper moves from the mall's loyalty wallet to a brand's offer engine? CMOs who have not answered these questions are one audit away from a crisis.

The third challenge is attribution. Indian retail operates across at least three commerce surfaces simultaneously — physical store, brand D2C, and marketplace (Myntra, Nykaa, Amazon). Loyalty touchpoints span all three, but most platforms track only in-store transactions. When a Pantaloons CMO asks 'what is the incremental revenue contribution of our loyalty program this quarter?', the honest answer from most current platforms is: we do not know, because we are missing 35-40% of the purchase signal.

Fourth, and perhaps most consequential, is the AI readiness gap. Most loyalty vendors in India — including several well-funded ones — sell rule-based segmentation dressed up as 'AI.' Assign a customer to a Gold tier if they spend ₹25,000 in a year. Send a birthday SMS. Re-engage if inactive for 60 days. This is 2009-era CRM logic. The gap between this and genuine machine-learning-driven next-best-action is enormous, and it is the gap that the AI first party data platform for retail loyalty is designed to close.

AI Use Cases for Customer Segmentation and Insights

The phrase 'AI-powered loyalty' is overused to the point of near-meaninglessness. So let us be precise about the five AI use cases that actually generate measurable ROI in Indian retail loyalty, and what each requires in terms of data infrastructure.

First is behavioral micro-segmentation. Traditional RFM (Recency, Frequency, Monetary) produces five to ten static buckets. A proper ML clustering model — run on 18 to 24 months of transaction history enriched with category, time-of-day, basket composition, and channel signals — produces 40 to 80 dynamic micro-segments. A Phoenix Marketcity operator, for instance, could distinguish between the 'weekend family anchor tenant visitor' who spends ₹8,000 per trip across food court, multiplex, and fashion, and the 'weekday solo professional' who spends ₹2,200 primarily at electronics and quick-service restaurants. The offer strategy for these two segments is completely different, and no human analyst is going to hand-code that distinction across 2 million members.

Second is next-best-offer prediction. Given a member's current session context — the brand they just visited, the time elapsed, their tier status, their remaining points balance — a trained recommendation model can predict, in real time, which offer from which partner brand will have the highest redemption probability. This is how Fundle's AI engine personalizes offers across 270+ partner brands: not by sending every offer to every member, but by ranking offers per member per moment.

Third is churn prediction and pre-emptive re-engagement. Indian loyalty programs bleed members silently. A member who visited Select CITYWALK four times in Q1 and has not appeared in Q2 is not 'churned' in the traditional sense — they may still be shopping, just not visiting. A churn model trained on dwell-frequency decay, spend-velocity change, and seasonal baselines can flag at-risk members 45-60 days before they fully lapse, when re-engagement is still economically viable.

Fourth is wallet-share estimation. This is underdeveloped in Indian loyalty but enormously valuable. By modeling the relationship between a member's declared income band, category purchase pattern, and peer-cohort behavior, the platform can estimate what share of a category's total spend the retailer is capturing. A member spending ₹3,000/month on apparel at Lifestyle but showing browsing signals consistent with a ₹12,000/month apparel budget is a wallet-share expansion opportunity, not just a retention case.

Fifth is channel and campaign attribution. Multi-touch attribution models that account for the non-linear Indian omnichannel path — discovery on Instagram, trial in-store, repurchase on the D2C app — are only possible when the first-party data is unified and the identity graph is clean. Without this, every campaign claims 100% credit and the CFO stops believing the marketing team.

Combining Privacy Compliance with AI Personalization

The single most common objection a retail CMO raises when the conversation turns to AI-driven personalization is this: 'Our legal team has put a hold on any new data initiative until we are DPDPA-compliant. We cannot move forward.' This is a false dilemma, and it stems from treating privacy compliance as a blocker rather than a design principle.

The Digital Personal Data Protection Act 2023 is, at its core, a consent architecture law. It does not prohibit personalization. It requires that personalization be grounded in freely given, specific, informed, and unambiguous consent — and that the data fiduciary be able to demonstrate that consent at any point. For a mall operator running a multi-brand loyalty program at a property like Select CITYWALK or Phoenix Marketcity, this means the consent framework must clearly explain to the shopper: (a) what data is collected, (b) for what specific purpose, (c) by whom, and (d) how it is shared with brand tenants.

A genuine privacy-first loyalty platform India builds this into the data model from day one, not as a compliance checkbox but as a trust signal. Research consistently shows that consumers are willing to share more data — and more accurate data — when they understand and trust the value exchange. A shopper who understands that sharing their purchase history across brands in a mall unlocks genuinely relevant offers (rather than spam) will maintain their consent and even expand it over time. This is the data flywheel that separates a privacy-first architecture from a surveillance-style CRM.

The technical implementation requires three elements that most legacy loyalty vendors do not have. First, a consent management layer that sits upstream of every data flow — ingestion, processing, and activation — and gates each operation against the stored consent record. Second, a data minimization engine that ensures AI models are trained only on the data necessary for the stated purpose; a churn-prediction model does not need a member's declared income, for example. Third, a right-to-erasure workflow that can propagate a deletion request across the identity graph, all downstream AI models, and all partner brand systems within the 72-hour window the DPDPA implies. This is genuinely hard engineering, and it is a meaningful differentiator between platforms.

The payoff for getting this right is not just legal safety. It is a consent-rich, high-quality first-party data asset that compounds in value every quarter, while competitors who ignored compliance scramble to re-paper their data practices under regulatory pressure.

Measuring Success and ROI from Your Loyalty Data Platform

The measurement problem in Indian retail loyalty is structural. Most programs are evaluated on vanity metrics — enrolled members (a stock, not a flow), total points issued (a liability, not an asset), and app downloads (a channel metric, not a business metric). None of these tell the CFO whether the loyalty program is generating more revenue than it costs. The consequence is that loyalty budgets get cut in downturns because they cannot defend themselves with numbers.

The right KPI architecture has two layers. The first layer is program health: active member rate (members with at least one transaction in the trailing 90 days as a percentage of total enrolled), redemption rate (points redeemed as a percentage of points issued, with a healthy Indian retail benchmark at 55-70%), and member ARPU versus non-member ARPU (the ratio should be 1.8× to 2.4× for a well-run program). These metrics tell you whether the program is alive and whether members value it.

The second layer is business impact: incremental revenue per active member (measured via holdout), loyalty-attributable repeat purchase rate, cross-brand penetration rate (in a mall program, what percentage of members transact with three or more anchor tenants — a key indicator of the data graph's commercial value), and cost-per-incremental-transaction. For a mid-size Indian mall operator with 1.5 million active loyalty members, a one-percentage-point improvement in cross-brand penetration rate — from, say, 22% to 23% — can mean ₹8-12 crore in additional annualized GMV across the tenant mix, at near-zero incremental acquisition cost.

AI-specific metrics matter too. Track model precision on next-best-offer recommendations (what percentage of served offers result in a transaction within 48 hours), churn model recall (what percentage of members who eventually lapsed were flagged as at-risk 45 days prior), and consent quality score (the percentage of active members with full, multi-purpose consent — a leading indicator of the AI models' future data richness). These metrics tell the CIO whether the AI investment is compounding or stalling.

The discipline of incrementality measurement — using randomized holdout groups — is non-negotiable. Without it, every campaign appears to 'work' because the members who received the campaign were already your best customers. True incrementality is the only honest answer to the CFO's question: 'What would revenue look like without this program?'

How Fundle solves this

Vineet Narang founded Fundle on a specific thesis: that Indian retail's loyalty problem is not a points problem or a technology problem — it is a data architecture problem dressed up as a marketing problem. Every capability in the Fundle AI Platform is designed to answer one question: how do you turn a fragmented, multi-brand, multi-format Indian retail environment into a single, consented, AI-ready customer intelligence asset?

The Fundle Loyalty Platform unifies the enrolment, consent, transaction, and engagement data across the entire brand or mall ecosystem into a single identity graph. For mall operators, Fundle Mall Loyalty handles the specific complexity of multi-tenant consent — distinguishing what data the mall operator holds as fiduciary versus what flows to brand tenants, with full audit trails for DPDPA compliance. For enterprise retail brands, Fundle Brand Loyalty connects in-store POS data (integrating with POSist, GoFrugal, and Wondersoft out of the box), D2C app events, and WhatsApp engagement into a unified member profile that updates in near real time.

The AI layer — the Fundle AI Brain — runs five production models on this unified data: behavioral micro-segmentation (updated weekly), next-best-offer ranking (real-time, triggered at checkout or app session), churn probability scoring (daily batch), wallet-share estimation (monthly), and multi-touch attribution (continuous). The Fundle AI Agents execute the downstream activation: pushing the ranked offer to the right channel — push notification, WhatsApp, in-store display, or POS prompt — at the moment of highest conversion probability. The Fundle Agentic AI layer means these agents do not wait for a human campaign manager to approve every send; they operate within policy guardrails set by the CMO and execute autonomously at the scale Indian retail requires.

For CMOs who want to build custom workflows — a cross-brand birthday campaign that triggers only when a member has consented to cross-brand data sharing and has a churn score above 0.65 — the Fundle AI Workflow engine provides a no-code canvas to compose these logic chains without engineering tickets. The result is a platform where the marketing team moves at the speed of insight rather than the speed of IT sprints. Across its live deployments, Fundle's AI engine personalizes offers across 270+ partner brands — generating the kind of cross-brand behavioral signal that transforms a single-brand CRM into a genuine retail intelligence platform. That is the infrastructure the Indian retail CMO needs to win the next decade.