Numbers you can defend in front of a CFO.
Every metric has a fixed definition, every headline number is proven against a human count before we sell, and measured and modeled never mix.
What each metric means, precisely.
Reach
Unique visitors who entered the activation during operating hours, counted at the entrance. Staff excluded.
Capture rate
Reach divided by the number of people who passed the activation.
Engagement rate
Share of visitors who spent at least 30 continuous seconds in a brand zone (entrance excluded).
Cost per engaged visitor
Production budget divided by engaged visitors. Shown when you share the budget.
- Zone visits / unique zone visitors
- Entries into a zone, counted per entry, always shown next to the number of different people.
- Zone penetration
- Share of visitors who entered a given zone.
- Dwell per zone
- Zone entry to zone exit, per visit; mean and median.
- Time in activation
- First entry to last exit, per visitor; mean and median.
- Bounce rate
- Share of visitors with under 3 seconds total.
- Arrivals per hour
- New visitors entering per hour, operating hours only.
- Occupancy
- People inside at a point in time.
- Heatmap
- Where people stood, operating hours only, staff excluded.
The rules the numbers follow.
Measured vs. modeled, never mixed.
Every number is tagged measured (counted from what the cameras saw) or modeled (calculated from assumptions). They never share a table or a chart. The report can only claim what was measured.
Proven by people, not by the model.
People count the same footage the system counted, independently and without seeing the system's number. Their count is the reference. See "How we prove it" below.
Proven for each type of activation.
Accuracy depends on the space, the crowd, and the camera placement, so we prove it separately for each type of activation (pop-ups, booths, out-of-home) before we measure it for clients, and spot-check every event after that.
How we prove it.
Before we sell, we prove it. Once we sell, we check every event.
We're in stage 1 now, on footage from our first activations.
Stage 1
Prove it first
- Two people independently count every entrance crossing from recorded footage. A third settles disagreements.
- Annotators box every person in sampled frames, including staff, people half-hidden, and people at the back of the room.
- Part of the footage is locked away before any tuning. The system is graded only on footage it's never seen.
- The bar: within 5% of the human count on a full day, every time. Until we clear it, we don't sell.
- Every format is proven on its own: pop-ups, booths, out-of-home, vehicles.
Stage 2
Check every event
- Frames pulled at random times throughout the event, counted by hand.
- Three random 15-minute windows at the entrance, counted in and out.
- The system has to land within 5%. The result goes on the report.
- Miss, and we recount the whole day before anything ships.
How we keep the check honest
- The times we check are picked at random by software and logged before anyone sees a result.
- At least a third of the checks come from the busiest hour, when counting is hardest.
- The person counting never sees the system's number.
- If a check fails, we recount the whole day before the report goes out.
- The person counting never tunes the system.
- Every result gets logged, including the misses.
What a measurement actually looks like.
Measurement questions
How accurate is it?
Our launch bar is visitor counts within 5% of a human count on a full day, on footage the system has never seen, and we won't sell a report until we meet it consistently. After launch, every event is spot-checked and the result is on the report.
How do you check every event without recounting everything?
We prove the system on full human counts first. After that, each event is spot-checked: frames captured at random times through the event are counted by hand, along with short random windows at the entrance, and compared to the system. If they don't match within 5%, we recount the whole day before the report goes out.
Is 95% the same everywhere?
No, and anyone who says so is quoting a brochure. Accuracy depends on the space, the crowd, and camera placement, so we prove it separately for each type of activation before we measure it for clients. Ask any vendor quoting "up to 99%": measured how, on what footage, against what reference?
Can we compare numbers across events and agencies?
Yes. That's the point of fixed definitions. An engaged visitor means 30+ continuous seconds in a brand zone at every event, so two activations measured by Campaign Vision are comparable no matter who produced them.
What happens with very small crowds?
Aggregates stay aggregates: heatmap cells with fewer than three people are suppressed, so a quiet moment never exposes an individual. Metrics for tiny samples are shown with appropriate caution rather than dressed up as trends.
Where do the report's claims come from?
Only from the event's measured data. The report never states a number that isn't on the dashboard, and modeled figures, if any, appear separately and are labeled.
Do you count staff?
No. Staff are identified before the event and excluded from every visitor number. The staff count is shown separately.
Why do you count at the entrance?
People cross a line one at a time, and a crossing at a line is the most reliable thing to count. Wide cameras then measure what happened inside.
Make your next budget conversation easier.
Tell us what you need to prove: an activation, a market, or an out-of-home placement. We'll show you how we measure it and how we prove the count.