Most people counting evaluations focus on two figures: accuracy and price.

Both matter, but neither tells you whether the system will answer the questions your retail teams need to answer. Two vendors may offer similar entrance-counting accuracy while providing very different capabilities for conversion, customer journeys, benchmarking, data ownership and future development.

The 19 questions in this guide are designed for procurement, retail operations, IT and analytics teams. You do not need technical expertise to ask them. You need clear answers, defined evidence and written commitments.

Key takeaways

• Ask vendors to verify performance in your locations and during your busiest periods.

• Treat every percentage as incomplete until you know the metric, conditions and sample size.

• Check what happens when data is missing or the sample is too small.

• Confirm whether the system can answer questions about visits, journeys and conversion, not only entrance volume.

• Examine benchmarking rules, data protection and peer-group requirements.

• Compare five-year ownership costs, not only device prices.

• Put data export, exit terms and technology upgrades in writing before signing.

If you only ask three questions

For an initial vendor shortlist, start here:

1. What happens when a camera or data connection fails? This reveals whether the system protects data quality or quietly fills gaps.

2. How are comparable locations classified? This separates a real benchmark from a large collection of unrelated sites.

3. What happens to our data if we leave? This exposes lock-in before it becomes expensive.

The full evaluation should still cover all 19 questions.

1. Can the vendor prove accuracy at our locations?

A brochure percentage may come from a controlled entrance with favorable lighting and moderate traffic. Your locations may have reflective glass, wide entrances, vestibules, backlighting or groups crossing together.

What to look for: Verification at your own locations, checked against a defined reference count before billing begins. Results should be documented and available for review.

Red flag: The only evidence comes from a laboratory or another customer’s site.

2. Was the verification large enough and run under realistic conditions?

Peak periods are often the hardest counting conditions and the most commercially important. A quiet demonstration does not show how the system performs when groups enter together.

Sample size matters too. A high percentage based on a few dozen crossings is less useful than a properly scoped result based on several hundred. Ask the vendor to explain why the sample is large enough for the claim being made.

What to look for: Tests that include peak traffic, state the number of visitors and report sample size beside every accuracy result. As a practical starting point, Indivd recommends several hundred crossings per connected unit, often 300 to 900 or more depending on the site and test design.

Red flag: The sample size is missing or the vendor selects only a quiet testing window.

3. What happens when a camera or connection fails?

Cameras fail. Networks drop. Equipment is moved during store maintenance. The important question is whether the system makes the problem visible.

What to look for: Clearly marked data gaps, labelled estimates and a defined record of the affected locations and periods.

Red flag: The dashboard always shows a number, but the vendor cannot explain how missing data is identified.

4. What happens when the sample is too small?

A conversion or bounce rate based on very few visits can move sharply because of one or two people. Showing a decimal point does not make the result reliable.

What to look for: A stated minimum sample size below which a rate is hidden, suppressed or clearly marked as unreliable.

Red flag: Every zone always reports a rate, regardless of traffic volume.

5. What is the denominator?

Traditional retail conversion divides transactions from the till by entries from a separate people counter. That ratio can be useful, but it cannot explain who visited, what they did or why they left.

What to look for: A denominator that can be defined for the business question, such as genuine visits, target customers or visitors who reached a relevant zone.

Red flag: The vendor cannot explain whether a falling conversion rate reflects weaker selling or a larger number of low-intent visitors.

6. Can the system show who visited, not only how many?

Entrance volume explains how busy a store was. It does not show whether the store attracted its intended audience or whether that audience converted.

What to look for: Privacy-preserving audience classification that can support a retailer-defined target customer profile, alongside a separate view of the audience that converted.

Red flag: Every result is calculated across all visitors, with no way to compare relevant audiences.

Any audience classification should be supported by transparent governance, appropriate privacy safeguards and documentation for the relevant jurisdiction.

7. Who controls the definitions?

Terms such as “engaged visitor” or “qualified visit” depend on the retailer’s format and objectives. A useful threshold for a fashion store may be unsuitable for a convenience store.

What to look for: Qualification rules and target customer definitions that your organization can set, review and govern consistently across locations.

Red flag: Definitions are fixed by the vendor or changed independently by each store.

8. Does the system forecast or only report?

Reporting explains what happened. Forecasting can support staffing, stock and campaign planning, but only when the source and limitations are clear.

What to look for: Forecasts based on your own historical data, clearly labelled as forecasts wherever they appear.

Red flag: Forecasts are presented as facts or are based on broad market estimates rather than your operating record.

9. Can it follow a visit from entrance to exit?

A visitor journey connects a person’s path through the location. It is different from a collection of unrelated zone counts.

What to look for: Journey performance verified in your building, against human observation and during busy periods.

Red flag: The claim depends only on an unspecified external benchmark. Ask what was measured, how ground truth was established and who carried out the test.

10. What are our locations compared with?

An estate average shows how your stores perform against one another. It cannot tell you whether the best store in your estate performs well against the wider market.

What to look for: Comparisons with relevant locations beyond your own estate, using a stated peer definition.

Red flag: The only benchmark is your estate compared with itself.

11. At what level does benchmarking happen?

Whole-store comparisons are often too broad to support action. The decision may concern a fitting room, department, concession or product zone.

What to look for: Like-for-like comparison at the level where decisions are made, such as fitting room against fitting room or fashion zone against fashion zone.

Red flag: The vendor has many installations but no classification system for determining which locations or zones are comparable.

12. How is benchmarking data protected?

Useful benchmarking should not expose another operator’s commercial data or allow a small peer group to reveal individual performance.

What to look for: Relative positions rather than identifiable values, no visibility of peer identities and minimum group sizes before a comparison is shown.

Red flag: Protection depends only on policy rather than rules enforced by the system.

13. Is the comparison a market index or a peer position?

A national footfall index provides market context. A position among comparable locations can support a specific decision. They are not interchangeable.

What to look for: Clear labelling of the comparison type. For an index, ask for its sample, definitions and category structure. For peer positioning, ask how comparable locations are selected.

Red flag: A broad market index is presented as evidence about your particular stores or trading format.

14. What is the five-year cost of ownership?

Price per device is only one part of the cost. Include installation, cabling, network changes, hardware replacement, software fees, IT support and the internal analysis needed to use the data.

What to look for: A five-year cost model that includes installation, ongoing support and the expected refresh cycle.

Red flag: The proposal quotes only a unit price and does not state when hardware must be replaced.

15. Can it use our existing infrastructure?

Some systems require new equipment and cabling at every entrance. Others can use compatible cameras and network infrastructure already in place.

What to look for: A clear description of what will be installed, network requirements, whether video is retained, where processing occurs and what IT must maintain.

Red flag: Ongoing IT responsibility is unclear or excluded from the cost model.

16. How do existing sites receive improvements?

If capability is tied to a hardware generation, every improvement may require a new device and site visit. Platform-based improvements may be deployed centrally.

What to look for: A documented upgrade model, including which improvements can be delivered centrally and which require hardware changes.

Red flag: The answer depends on replacing equipment at every location. Ask what changed during the last 12 months and what existing customers paid to receive it.

17. What happens if we leave?

Exit terms are easy to overlook during procurement and difficult to negotiate later.

What to look for: Your historical record returned in a documented, readable format, with the delivery method, timing and cost written into the agreement.

Red flag: Data export and deletion terms are absent or left to future discussion.

18. Where does the analytics capability live?

Some capabilities are fixed in the installed device. Others run in a central platform and can improve across the estate without replacing hardware.

This affects how quickly new models, definitions and analytical features can reach existing sites.

What to look for: A clear division between device-level and platform-level processing, including how each is upgraded.

Red flag: Important capabilities are tied permanently to the installed hardware generation.

19. Is the data structured for consistent analysis?

Automated analysis and assistants depend on shared definitions. If every location names zones, visitor types and events differently, estate-wide comparison becomes unreliable.

What to look for: Shared definitions enforced across the estate. Every answer should state the measure, time period, scope and level of confidence. Access rights should be enforced when the answer is produced.

Red flag: Definitions rely on free text, or an assistant answers questions that the underlying data cannot support.

The pattern behind the 19 questions

The strongest signal is not the number of features. It is how precisely a vendor describes the limits of its system.

A reliable vendor verifies performance in your locations, marks unsupported periods, explains what a metric can and cannot show, protects benchmarking data and commits to an exit process in writing.

Take these questions into the evaluation meeting. Record the evidence behind each answer and note where a commitment is verbal rather than written. The result will tell you more about the system and the supplier relationship than a feature checklist alone.