A fashion retailer in Northern Europe had its people count verified at three stores between January and February 2026.

At every location, Indivd’s count was compared with an independent count of the same hours. Each store achieved at least 97.78% absolute accuracy.

This is the lowest result among the three stores, not their average. The verification covered 3,101 visitors and 54 hours across ten counting lines on five cameras.

Midwinter in Northern Europe, when daylight is short and visitors arrive in heavy coats. Three fashion stores, three different buildings, each with a people count running at its doors.

Within two winter weeks, all three stores had that count verified.

The retailer selected the locations, hours, and scope of each check.

The stores

One visitor count beneath every store decision

Fashion stores use visitor counts to plan staffing, assess window displays, calculate conversion, and compare performance across locations.

How many people should be rostered for the afternoon? Did a new window display bring more people inside? Which stores are gaining visitors, and which are falling behind?

Each of these decisions depends on the count at the door. Across the three stores, that count was produced by ten entrance and exit lines on five cameras.

Fashion entrances can be difficult places to count. Shoppers pause near the window without entering, groups cross together, and heavy winter clothing changes how people appear to a camera.

The challenge becomes more important when stores are compared. A performance table assumes that every location’s visitor count is equally reliable. It has no separate column showing the effect of counting errors.

Accuracy is not something a model has. It is something a deployment produces.

Three buildings mean three deployments. A consistent standard only exists if it is achieved at every location.

A like-for-like comparison built on unchecked visitor counts can rank the counters instead of the stores.

The check

Three stores assessed using one standard

The retailer decided where, when, and how much was tested at each store. The verification periods fell during two weeks of winter trading.

Across the three locations, 3,101 visitors were included in the comparison over 54 hours. Indivd’s count was compared with an independent count of the same hours.

The verification covered all ten counting lines on all five cameras. Each store’s complete selected counting set was assessed together.

Every store achieved at least 97.78% absolute accuracy. This figure is the lowest result among the three locations.

Reporting the minimum matters because it states the accuracy achieved at every tested store. An average could allow a stronger result at one location to conceal a weaker result at another.

What the number carries

Three stores that can be compared with greater confidence

Verified between January and February 2026: every one of three fashion stores in Northern Europe achieved at least 97.78% absolute people counting accuracy.

The verification covered:

• 3 fashion stores
• 3,101 visitors
• 54 hours
• 10 counting lines
• 5 cameras

The minimum result was 97.78%, which means every tested store achieved that level of accuracy or higher.

Ten counting lines on five cameras, installed across three separate buildings, were assessed using one accuracy standard. This shows that the result was achieved at each deployment rather than assumed from a single location.

For the retailer, the value is more dependable store comparison. Staffing decisions, conversion rates, and assessments of window displays can be based on visitor counts that were checked in the stores where they were produced.

Without verification, counting errors can distort the comparison itself. An undercount can make a successful window display appear less effective. An overcount can make conversion appear weaker and direct staff or investment toward the wrong location.

When each store’s visitor count is assessed using the same measure, differences in the results are more likely to reflect the stores rather than their counters.

In a separate engagement, KPMG audited Indivd’s counting accuracy at a major retailer’s store against manual ground truth. Every camera tested came within 1.17%, the strongest result of any system included in that audit.

The retailer can now ask a question that one store alone cannot answer: which location performed best during the winter, based on three visitor counts assessed in the same way?

FAQ