Engagement is dwell with a threshold. A person who walks through a zone in front of a display registers a dwell of a second or two, indicating they are just passing through. We can infer that a person who stops for three seconds or longer has, at minimum, looked. Engagement counts the second group: the number of dwell events in a zone that exceed the threshold, and the share of people passing or entering the zone that we infer engaged.
The exact threshold is a convention that can be tuned to match the environment. Three seconds is a common choice because it is longer than a glance while walking and short enough to catch a look at a screen or a window. Longer thresholds suit longer interactions: thirty seconds for a fixture where people handle products, several minutes for a seating area. What matters is that the threshold is fixed, stated, and the same across the zones and stores being compared.
Engagement rate, engagements ÷ people passing the zone, is the display-level equivalent of capture rate. For a window it measures whether the display stops people; for an in-store screen or bay it measures whether the content holds them. Combined with sales data for the products on the bay, it shows whether attention turned into purchases.
In retail media, engagement is the metric that distinguishes a screen that was merely in view from one that was watched, and it is reported alongside impressions, which count everyone who passed.
Glimpse reports engagement from zones on existing CCTV and from Glimpse Sensors mounted on a display or screen, with the threshold set per zone and 3 seconds as the default. Each engagement is a dwell event, which is privacy-safe; nothing about who engaged is recorded.