Privacy-Safe Footfall Counter for Small Shops Using ToF Sensors

The problem small shops have with footfall

A small shop, pharmacy or cafe knows its sales but not how many people walked in. Without footfall you cannot tell whether a slow day was fewer visitors or fewer visitors buying. You cannot plan staff for busy hours, test whether a new window display brings people in, or compare branches. Large retailers answer this with overhead people counters, but those cost far more than a corner shop will pay. The cheap alternative, a camera with a counting app, brings a different problem: owners and customers are uneasy about cameras analysing faces, and privacy law in many countries treats images of people as personal data.

What already exists and the gap

• Xovis makes stereo 3D sensors for retail and airports with very high accuracy, priced at hundreds of euros or more per sensor before software.
• Density and Butlr use depth and thermal sensors that avoid identifiable images, aimed at offices and larger buildings with subscription pricing.
• Infrared beam counters sold cheaply online count beam breaks, cannot tell in from out and miscount groups.

The gap is a privacy-safe counter that knows direction, is honest about its accuracy, costs close to the price of a beam counter, and comes with a dashboard a shop owner actually reads. A single-door time-of-flight sensor is enough for most small shops.

What you build

• Counter device: an ESP32 with a time-of-flight distance sensor mounted above the door, looking down. The sensor can be split into two regions of interest, so a person crossing produces a pattern that tells you the direction of travel.
• Firmware: detection logic, a local counter, a ring buffer for offline periods, and upload over MQTT or HTTPS.
• Backend: an API and a PostgreSQL table of counts per door per 15 minutes.
• Dashboard: footfall by hour and day, busiest hours heatmap, and conversion when sales numbers are added.

Roles are simple: an installer who calibrates the device, and an owner who reads the dashboard.

The technically interesting part

Turning distance readings into reliable in and out counts is the real work. People walk at different speeds, pause in the doorway, walk side by side or carry large bags. You will build a small state machine per zone that tracks a person entering zone A, then zone B, and only counts when the sequence completes, with timeouts for someone who steps back out. Tuning thresholds against real data, and measuring the error honestly, is what makes this worth a final year project. Show your supervisor the raw traces of tricky cases (two people together, a child, a pram) and how the algorithm handles them.

How to evaluate it

• Count accuracy against a human counting with a clicker for at least three hours at a real door, reported separately for in and out, and for single people and groups.
• Offline reliability: cut Wi-Fi for two hours during trading and check that uploaded totals match the device totals exactly.
• Cost: final bill of materials per door.
• A short interview with two shop owners after a week of use: did they change anything based on the data?

Suggested plan

1. Weeks 1 to 3: sensor bring-up, region-of-interest reading, logging raw traces.
2. Weeks 4 to 6: direction detection state machine, tuned on recorded traces in a campus doorway.
3. Weeks 7 to 9: offline buffering, backend API and database.
4. Weeks 10 to 12: dashboard and calibration mode.
5. Weeks 13 to 15: real shop trial, accuracy study and report.

Path to a product

The first customer is a pharmacy chain or a small clothing retailer with two to five branches that wants to compare them. Charge a one-time hardware price and a small monthly fee per door for the dashboard. Shopping mall managers are a later market because they want counts across many entrances. The same device fits libraries, clinics and university labs that must report visitor numbers.

Scope warning

Do not add cameras, face detection or demographic guesses: that defeats the point. Do not build a full point-of-sale integration; a CSV import of daily sales is enough. Do not try to track individual customer journeys inside the shop.

Learn what you need on Sefism

• Data analytics roadmap : how to turn counts into busy hours, trends and conversion rates people act on.
• Backend roadmap : APIs and databases for the part that stores and serves counts.
• PostgreSQL Documentation : time-bucketed queries and indexes for per-door counts.
• From Data to Viz : picking the right chart for hourly and weekly patterns.
• EU data protection rules for business : why a camera-free design matters and how to write the privacy notice.

Get help with this idea

The counting algorithm is where most teams under-plan, so bring recorded traces if you have them. If you want a second opinion on scope, hardware choices or how to set up the evaluation before you commit, bring your plan and your bill of materials and we will pressure-test it together. New here? Create a free account first.

• Discuss this idea with a mentor
• Send a query to Tauseef
• Book a 1:1 session