IoT-Based Visitor Tracking Application

Overview

What this project was about

Inspired by a visit to METRO supermarket in Bangalore, this concept project explored a simple question: supermarkets know almost nothing about how customers actually move through and experience their stores. CCTV exists for security, not insight. What if retailers had a tool analogous to Google Analytics — but for physical spaces? I designed the data model, dashboard architecture, and key screens for an IoT-powered visitor tracking application.

The Challenge

Understanding the problem space and identifying where the current experience was failing users.

The Problem Worth Solving

Digital businesses have extraordinary visibility into user behaviour: every click, scroll, hover, and conversion is tracked and analysed. Physical retailers are operating essentially blind. Traditional feedback mechanisms — rating boxes, occasional surveys — capture almost nothing about actual in-store behaviour. Valuable information about what customers look at, where they linger, what they pick up and put back down, and where they get frustrated is lost every day.

The inspiration — a visit to METRO supermarket in Bangalore raised the question: what happens to all the behavioural data generated here every day?

Supermarket operational challenges — The operational challenges facing physical retail — relying on CCTV for security only, while valuable customer insight goes uncaptured

Proposed Technology

The concept uses real-time video stitching from ceiling-mounted cameras across store sections, combined with AI processing to extract behavioural signals without capturing personally identifiable information. The insight: cameras are already mounted on ceilings. The question is whether they can be re-purposed from security to insight.

Multi-camera video stitching and processing diagram — Video stitching and AI processing — combining feeds from ceiling-mounted cameras to build a continuous store view

Customer movement path tracking visualisation — Customer movement tracking — visualising individual paths through the store, dwell time per section, and eye direction data

Combined data analytics metrics diagram — Data combination — merging path tracking with purchase history, billing, feedback, and inventory for a complete picture

Dashboard Design

Four dashboard views were designed to serve different decision-making needs — from individual visit analysis to real-time floor management to trend reporting.

Customer analytics dashboard wireframe — Customer analytics — individual visit summaries showing path through store, dwell time, and basket correlation

Key metrics real-time display — Key metrics display — real-time summary for at-a-glance store health, designed for managers who don't need to deep-dive daily

Interactive store layout with live customer positions — Real-time feed — aerial store view with live customer positions, giving floor managers immediate situational awareness

Store analytics dashboard with performance charts — Store analytics — interactive charts for campaign performance, seasonal trends, and layout optimisation over time

“How about a Google Analytics-like tool to track visitors on a physical store? Based on the customer movement, various metrics are tracked for the betterment of customer experience.”
Project brief·Concept exploration, 2018

Results

Outcomes & Impact

Dashboard views designed

Customer analytics, real-time aerial feed, store analytics, and key metrics — a complete system for retail behavioural insight.

Behavioural signals tracked

Dwell time, movement speed, eye direction, cart abandonment, queue efficiency, session replay, and more — all from existing ceiling camera infrastructure.

New hardware required

The concept leverages cameras already installed in stores for security, reorienting existing infrastructure toward business insight.

Reflection

What I learned

This concept project taught me that sometimes the best design work sits at the intersection of problem identification and technology feasibility. The real insight wasn't 'let's track customers' but 'cameras are already here—what insights are we leaving on the table?' Designing data models before interfaces forced clear thinking about what actually matters. I'd bring this constraint-first approach to more projects: physics and infrastructure often point toward better solutions than starting with a blank canvas.

Interested in working together?

I'm available for freelance design work, consulting, and speaking opportunities.

Get in Touch