How Fynd helped Reliance Retail uncover a 62% funnel drop-off with Retail Jarvis

Reliance Retail runs hundreds of stores, each generating a steady stream of small signals - a customer who browses but doesn’t buy, a long queue, a trial room left unattended.

Traditionally, the only way to understand how a store was truly performing was by asking customers directly through surveys or feedback forms. But surveys often capture only a fraction of the story.

Reliance Retail partnered with Fynd to create Retail Jarvis - a system that observes what’s really happening in a store without asking the customer a single question.

The challenge: Customer experience was a guessing game

Feedback only came from customers who chose to give it

Surveys, ratings and complaint forms reached only those customers motivated enough to speak up. The quiet majority, who might lose interest and leave without a word, left no trace.

Nobody could see where the journey actually broke

A store might miss its footfall targets, but the cause was unclear. Was it staffing, product availability, queue times or something happening in the trial room? Without full visibility, every explanation was just a guess.

Store issues and customer sentiment lived in separate worlds

Operational data like staffing, cleanliness, and stock levels were in one system, while customer behavior, if tracked at all - was in another. Nobody could link, for example, a messy shelf on Tuesday to a dip in conversions on Wednesday.

The solution: An engine that watches, not asks

Four principles, before a single score

Fynd built Retail Jarvis around a simple idea: customer experience must be observed rather than asked about. It begins with assessing operational readiness, then captures what happens inside the store, watches how customers respond, and finally interprets what it all means.

One engine, three sources of signal

The system draws from three sources: offline store data (footfall, queues, product zones), online and app data (search, checkout, delivery) and external context (competitor pricing, local promotions, sentiment). All this data flows through four stages: readiness measurement, journey capture, behavioral analytics and AI Inference.

Every signal collapses into four scores

All of that data narrows down into four clear outputs.These signals converge into four clear scores - Store Readiness, Customer Experience, Journey Friction and Churn Risk each linked to the team best equipped to act, from store operations to CRM.

A console built around one real store

Fynd launched five operator dashboards in a pilot store in Mumbai, tracking the customer journey funnel, walk-in behavior, trial room activity, billing, and an action tracker that follows issues through to resolution.

A loop that keeps learning

The system is designed to learn continuously. After teams address an issue, the same signals are measured again and scores update accordingly. This feedback loop helps the model improve with every cycle instead of remaining static.

The impact: Problems that used to be invisible, now visible

20.6k walk-ins tracked in a single week

Retail Jarvis captured detailed walk-in activity during the pilot - 9,433 in the first week and 10,265 in the second providing the store team with a real-time, continuous understanding of footfall rather than relying on monthly estimates.

62% drop-off between footfall and billing, now visible

The customer journey funnel showed 10,265 walk-ins narrowing to 7,185 browsing customers, then 3,593 in trial or selection, and finally just 1,383 at billing. This gap was always there but never clearly seen before.

30% of customers walked out in under 2 minutes

About 6,186 customers walked out almost immediately after entering - a silent loss no survey would have detected.

A staffing gap during peak hours, quantified

During busy trial-room times, staff ratios stretched to 1:12 compared to 1:6 during quieter periods. Only 15% of walk-ins received staff interaction, while the remaining 85% navigated the store without assistance.

98% of returns traced to a single, fixable cause

Retail Jarvis identified that nearly all returns stemmed from customers not finding the right size. This insight transformed a vague “returns problem” into a specific issue the store could address.

Retail Jarvis is still in build, anchored on a single pilot store in Mumbai. These numbers come from that anchor demo, not a full production rollout - the platform's next step is scaling across more stores and formats.

Want to see what your stores are really telling you? See what Fynd can do for you.

Book a demo today