Skip to main content

👣 Footfall metrics & live counts

How Merlin Cloud cameras measure in-store footfall: what we count, how live people are shown, busiest hours, plus accuracy and limits. More.

Leo avatar
Written by Leo
Updated over 5 months ago

What we measure

  • Camera analytics (whole frame): Counts everything within the camera’s field of view.

  • Zones (optional): Draw polygons on the floor plan to see counts for specific areas.

  • Entry/exit lines (optional): Draw lines on doorways to track ingress/egress events for stores or zones.

You can use any combination of the above. For multiple doors, draw a line for each door; totals aggregate across configured lines.

Metric glossary

  • People now - Live count of people currently visible under the camera view.

  • Total People - Sum of people detected within the selected time/date filter (camera frame or selected zone).

  • Busiest time - The top hour within your current filter window (commonly the selected date).

Counting model (important)

  • We count sessions, not unique individuals (GDPR-first design).

    • If someone exits and re-enters, this produces two sessions.

  • Staff filtering: Not enabled by default. Enterprise clients can request worker identification options.

  • Side/staff doors: Add entry/exit lines on those doors if you want them included (or excluded by omission).

Entry & exit lines: how counts are produced

  • Create an ingress/egress line across a doorway in the camera view.

  • When movement crosses that line, we record an Entry or Exit event (directional).

  • Use separate lines per door to compare entrance performance.

Accuracy guidance

  • Typical accuracy is 85–90% under normal/ideal conditions; real-world results vary.

  • Accuracy may drop with crowding, occlusion, extreme lighting, mirrors/reflective surfaces, or poor camera placement.

Offline behaviour

  • If the camera goes offline, counts are not currently buffered or backfilled. (Enterprise backfill can be added if required.)

Privacy posture (summary)

  • No facial recognition.

  • Full body blur applied; detection is processed server-side and anonymised immediately after the detection script runs.

  • We report derived metrics, not raw identities.

Setup tips for reliable counting

  • Mount cameras high and angled slightly downward to minimise occlusion.

  • Ensure consistent, even lighting; avoid backlighting and reflective surfaces where possible.

  • Test zones and door lines: walk through them during setup to confirm expected counts.

  • Start with ≤ 4 zones per camera for clarity (more are supported but can be harder to manage).

“No data” state

  • New cameras show No data until events are recorded. Verify power/network, confirm the camera view covers your area, and test by walking into frame.

Did this answer your question?