An Australian apparel group had POS data, e-commerce data, and loyalty data. They had no reliable answer to the most basic question: how many people walked into the store, and how many of them bought something?
The problem
The group had been using thermal footfall sensors that drifted 15–20% within months of install. Conversion — the number that drives merchandising decisions — was effectively a guess. Regional stores were being evaluated on bad numbers, and the central merchandising team was allocating stock on bad forecasts.
Our deployment
We installed Canopy CU-1 units at the entrance of each of 38 stores, running a directional people-counting agent. Counts were tied back to the group's POS system (Lightspeed) in the console. Within a week, merchandising had a real conversion number: transactions divided by footfall, by hour, by store.
The result
The merchandising team reallocated stock between stores based on real conversion, not on modelled estimates. Three stores that had looked "underperforming" turned out to have normal conversion but low footfall — a property issue, not a buying issue. Two other stores had high footfall but low conversion, and got display redesigns.
Implementation details
Each store ran a Canopy CU-1 unit at the entrance, fed by the existing dome camera above the door. The directional counting model was calibrated by walking the entrance during a two-hour calibration window; counts were within ±3% of a manual audit by staff. The event stream was joined to the Lightspeed POS feed in the console on a daily batch — footfall by hour, transactions by hour, conversion by hour. No loyalty data was joined. No individual shopper was identified.
For the merchandising team, we built a simple Power BI report that showed conversion by store, by category, by hour. They could slice by region, format, and trading day. The report replaced three separate spreadsheets that had been maintained by analysts for years.
Queue handling
When a checkout queue exceeded three people, the system opened an extra till through the workforce roster tool.
Ongoing support
The rollout is now in its second year. We have added dwell measurement in three pilot stores, to understand which categories are actually being browsed. The merchandising team uses the data quarterly. The Canopy units sit in the store equipment room and require no operator attention.