People counting vendors often lead with an accuracy percentage. You may see claims of 98%, 99% or higher presented side by side, as if they describe the same test.

They rarely do.

One figure may come from a controlled laboratory. Another may describe a single reference site. A third may be the vendor’s own test, carried out once under favorable conditions.

The percentage is not enough to compare them. You also need to know what was measured, where the test took place, how the ground truth was established, who carried out the work and whether the evidence is still available.

This guide gives you a five-part test for getting those answers. It also applies the same test to Indivd’s own accuracy claims.

Key takeaways

• An accuracy percentage has little value without a defined scope.

• Entry counting, occupancy and zone movement are different measurements and should not share one accuracy claim.

• Results from one well-controlled site do not represent performance across every deployment.

• Manual review of recorded footage provides a stronger ground truth than an unexplained internal test.

• “Independently verified” can describe different levels of independence. Ask who designed the method as well as who ran it.

• Accuracy should be checked again when a camera is moved, refocused or re-angled.

• Vendors should retain test results and make them available for third-party review.

Why accuracy percentages are difficult to compare

People counting accuracy is not a fixed property of a product. It is the result produced by an installed system under a specific set of conditions.

Mounting height, camera angle, entrance width, lighting and traffic density can all affect counting performance. A system may perform exceptionally well at one entrance and less well at another, even when the same hardware and software are used.

This means a best result, an average and a minimum performance level are different claims. Each can be valid, but each answers a different buyer question:

• Best result: What has the system achieved under favorable conditions?

• Average: What does the system typically achieve across the sites included in the sample?

• Performance floor: What level can a buyer expect across nearly all deployments?

For procurement and operational planning, the performance floor is usually the most useful. It sets an expectation that can be tested at your own locations.

The five-part accuracy test

An accuracy claim is complete when it answers five questions:

1. What was measured?

2. Where was it measured?

3. What was used as ground truth?

4. Who designed and carried out the test?

5. Where is the evidence?

If one of these answers is missing, the percentage is incomplete.

1. What was measured?

Start by asking what the accuracy percentage describes.

A system may measure:

• entries and exits across a counting line;

• occupancy at a particular moment;

• movement between zones;

• unique visits;

• visitor journeys through a location.

These measurements have different failure modes. A result earned on entry and exit counting cannot automatically be applied to occupancy or zone-level journeys.

Ask the vendor to name the exact metric, the unit being compared and the way the error was calculated.

2. Where was it measured?

Counting conditions vary from one location to another.

A wide, well-lit entrance with an ideally positioned camera is different from a narrow doorway with strong backlight, dense group traffic or limited mounting options. A result from the first site does not predict performance at the second without further evidence.

Find out whether the published figure describes:

• a laboratory test;

• one reference location;

• a selected group of locations;

• an average across deployments;

• a defined performance floor across production sites.

Also ask how many entrances, cameras, hours and visitors were included. A short test during a quiet period is not equivalent to a test that includes peak trading conditions.

3. What was used as ground truth?

An accuracy percentage compares the system’s output with a record of what actually happened. That reference is the ground truth.

Manual annotation of recorded footage, with each crossing reviewed person by person, provides a clear basis for comparison. Whatever method is used, the vendor should state:

• how the reference count was created;

• who completed it;

• how much footage was reviewed;

• how many visitors were included;

• whether both quiet and peak periods were tested.

“Internal testing” does not answer these questions. It names the organization, not the method.

4. Who designed and ran the test?

The phrase “independently verified” needs clarification.

There are two useful but different forms of independence:

• Independent methodology: An external party designs the test and carries it out using its own method.

• Independent execution: An external party carries out a method supplied by the vendor.

Both can produce meaningful evidence. They should not be presented as the same claim.

Ask who designed the method, who selected the sample, who ran the test and who interpreted the result.

5. Where is the evidence?

A published figure should be supported by records that can be reviewed later.

Ask whether the vendor retains:

• the system count;

• the reference count;

• the difference between them;

• the location and camera details;

• the dates, duration and traffic volume;

• the final accuracy calculation.

You should also know how long these records are kept and whether a customer or third party can inspect them.

Evidence retention matters because it makes the claim repeatable and accountable. Without it, the result cannot be checked when conditions change or questions arise.

Why accuracy should be checked again

Accuracy describes an installed camera under current conditions. It does not describe the product forever.

A camera may be moved during store maintenance, refocused after an installation change or affected by a new fixture near the entrance. Lighting and traffic patterns can also change.

Ask the vendor what triggers a new accuracy check. At minimum, the process should cover changes to camera position, angle, focus or the physical counting area.

Re-verification should also be possible when store teams see an unexpected change in traffic or conversion data.

Applying the test to Indivd’s claims

Indivd makes two accuracy statements with different scopes.

The general deployment claim

Indivd’s general claim is counting accuracy of at least 97.5% in 99 out of 100 deployments.

This is presented as a performance floor, not as Indivd’s highest recorded result. One quality-assurance process is applied across deployments. Variation between sites may result from camera placement and physical conditions, rather than from different testing standards.

The customer-commissioned review

A separate customer-commissioned review, executed by KPMG, reported approximately 99% accuracy.

The scope is important. The review was independent in execution, not in methodology, because KPMG applied Indivd’s quality-assurance method rather than designing a separate test method.

Indivd states that the results supporting both claims are retained indefinitely and can be checked by a third party.

These two figures should not be merged into one headline number. The first sets a broad deployment expectation. The second is a specific result from a defined review.

A buyer’s accuracy checklist

Before a pilot or contract, ask the vendor to provide the following in writing:

• the metric covered by each accuracy claim;

• whether the figure is a best result, average or performance floor;

• the locations and conditions included;

• the number of cameras, hours and visitors tested;

• the ground truth method;

• who designed and executed the test;

• whether the independence relates to methodology, execution or both;

• the events that trigger re-verification;

• how long the underlying results are retained;

• whether the evidence can be checked by a third party.

Do not compare percentages until these details are available. Two vendors can publish the same figure while offering very different levels of evidence.

Treat accuracy as evidence, not a headline

People counting accuracy affects every metric built on the count, including conversion, staffing and performance comparisons. It deserves more scrutiny than a percentage in a comparison table.

The most reliable vendor is not necessarily the one with the largest headline number. It is the one that defines the metric, states the scope, documents the ground truth, explains who carried out the test and keeps the evidence available for review.

That is the standard buyers can apply to every vendor, including Indivd.