Start with the decision the measurement will support
Do you want to know when staff are needed, which product area attracts activity, or whether people finish an interactive experience? These questions need different measurements. A large entry count does not explain a product's appeal, and a long stay can mean either engagement or waiting.
Write the intended decision beside each metric. Then ask which sensor, application event or observation can provide evidence for it. Discuss the plan through visitor tracking, with an explicit request to confirm the current solution's capabilities for your layout. A generic service name is not a measurement specification.
Keep the units distinct
| Measure | Operational definition to specify | What it does not establish by itself |
|---|---|---|
| Entry events | Crossing a named boundary in a specified direction | Unique people when someone can enter again |
| Unique participants | People deduplicated by a defined, suitable method over a stated period | A result obtainable from every counter |
| Occupancy | Estimated number present in a defined area at a time | Total attendance over the day |
| Time in zone | Duration between defined arrival and departure events, or a documented estimate | Attention, enjoyment or purchase intent |
| Interaction | A named application action such as session start or completion | A unique person or qualified sales lead |
| Conversation or contact | A separately recorded event with an agreed definition | Automatic attribution to every nearby screen or game |
Axis distinguishes line crossings from occupancy within an area and recommends validating installations against observations. Its counting white paper explains those measurement choices. It does not establish the features or performance of another product.
Make a small measurement worksheet
For each zone, record its boundary on the plan, the metric, the data source, the reporting interval and the person responsible. Add inclusion rules for staff, passers-by and repeat entries. If a system cannot distinguish a category, state that limitation rather than silently removing it from the report.
Use consistent time periods when comparing zones. Note closures, demonstrations, content changes and equipment interruptions. A quiet half-hour during a closed demonstration area should not be presented as evidence that visitors disliked its product. Keep missing data separate from a measured zero.
Worked example: entries and a product game
Imagine a hypothetical two-hour period with 240 recorded entry events at the stand boundary, 90 game starts and 72 game completions. The game reports one start and at most one completion for each valid session, and the counters cover the same period without an interruption.
The game completion rate is 72 ÷ 90 = 80% of recorded starts. This describes session completion under the stated event rules. The 90 starts divided by 240 entries gives 37.5 starts per 100 entry events. It is not a person-level participation rate: returning visitors may create several entries or sessions, and the measurements are not linked.
Report the raw counts and the definitions alongside these ratios. To investigate a lower completion rate later, inspect the application flow and host notes. The count alone cannot tell you whether the cause was confusing instructions, an early departure or a software problem.
Validate the installed arrangement
Agree a practical observation test with the supplier at representative traffic levels. Compare the same boundary, direction and time interval. Include people entering together, pausing near the boundary and returning. Record discrepancies and the conditions of the check rather than copying a general accuracy percentage from a brochure.
For an open stand, confirm what the system can observe at every accessible edge. For application analytics, test each named event, including interrupted sessions and restarts. Align clocks and define the event time zone before combining exports.
Agree data handling before collecting it
Ask what raw data is generated, whether identifiable images or identifiers are involved, where processing happens, who can access the information and when it is deleted. Have the responsible privacy lead assess the proposed setup and any required communication to visitors. These decisions belong to the actual deployment; do not assume that a diagram labelled “anonymous analytics” resolves them.
What to request from a supplier
A metric dictionary and marked measurement zones.
Confirmed capabilities, exclusions and installation assumptions.
An example report with units, intervals and missing-data flags.
A validation procedure and a record of configuration changes.
Data-handling information for your responsible reviewer.
Share the intended decisions with the tracking team. A small set of well-defined measures is more useful than a dashboard whose totals cannot be explained.