Counted, not estimated.
A camera on each screen counts the people in front of it while your ad plays.
Why a count beats an estimate.
A model can say how many people usually pass a screen. A count says how many were there while your ad played. That is the difference between a guess and evidence, and it lets you compare a screen slot against other media on the same terms.
Where counting sits on the impression ladder.
Out-of-home measurement uses a ladder of terms. Higher rungs need more evidence. These are the industry terms (IAB and MRC).
| Term | What it means | Evidence needed |
|---|---|---|
| Gross or rendered impression | The ad played on a working screen with people likely around. | Proof of play plus a footfall estimate |
| Opportunity to see (OTS) | A person was in the exposure zone while the screen was active. | Usually modelled |
| Likelihood to see (LTS) | People were detected in the exposure zone while the screen was active. | Sensors or tracking on site |
| Audience impression | A person who showed actual viewing behaviour. | Attention signals such as gaze |
From presence to attention.
People present
People detected in front of the screen while an ad plays.
Dwell time
How long people stay near the screen.
Unique reach
Distinct people over a period, so one person is not counted again and again.
Attention signals
Whether someone is facing the screen. This is the step from presence towards audience impressions, and it is being rolled out.
One person, one impression per play.
A person standing in front of a screen for a thirty-second ad is seen on hundreds of video frames. Gridcast tracks them as one person and counts one impression for that play, not one per frame.
Counting happens on the device.
The model runs on the box beside the screen. Video is processed there and does not leave the device. Only anonymous counts are sent to the dashboard. See privacy by design for the details.
Measurement questions
Is a person counted twice if they stay?
No. A person is tracked across the play and counts once per play.
How accurate are the counts?
Counts depend on camera placement, lighting and how busy the venue is. We are validating them in our early network and will publish how we measure.
Do you identify who people are?
No. Gridcast counts people. It does not identify them.
What can I compare it to?
Because counts use standard OOH terms, they can be set beside other out-of-home media on the same basis.