🔥 Heat maps: how they work & how to read them

Source: Positions detected within the camera’s view (or inside a selected zone).

Aggregation: We count visits per pixel grid (x, y) across your chosen date range.

Smoothing: Density is widened via morphological dilation and then Gaussian blur—you see clusters, not exact footprints.

Scaling: Each heatmap normalises its own scale (0–255) for the selected time range. Intensity is relative within that view, not comparable across different days by colour alone.

Blue → low activity

Green/Yellow → medium activity

Red → high activity

Date range: The UI defaults to 30 days but any date range is supported.

Zones: Selecting a zone restricts the heatmap to that polygon; all widgets update to zone-specific data.

Locating hotspots/coldspots and aisle dwell patterns

Seeing the impact of fixtures or entrance placement

Guiding merchandising & flow experiments

Person-level tracking or exact positions

Cross-period colour comparisons without the same time filter

Inferring staff vs shopper (unless you use enterprise options elsewhere)

Look for persistent red paths → high flow corridors.

Red near a kiosk without matching kiosk events may signal attraction but low engagement.

If a refit reduces red in an obstruction area, you likely improved throughput.

No facial recognition. Full-body blur.

For heatmaps we store computed overlays + coordinates + metrics, not raw video frames. (Frame retention policies live in Privacy, data handling & retention.)

Accuracy drops with crowding, occlusion, harsh lighting, mirrors/reflective surfaces.

Mount cameras high, angled down, and verify results by a quick walk test.

If the camera goes offline, heatmaps won’t backfill for that gap (enterprise backfill optional).