Data quality: what affects accuracy
Session based model
Counts are session based, not unique people.
Kiosks dedupe by session ID on upload.
Cameras do not track identities and do not backfill offline gaps.
Known edge cases
Mirrors and reflective surfaces can inflate counts.
Heavy crowding or occlusion reduces detection accuracy.
Extreme lighting or motion blur can lower detection confidence.
Poor camera placement or narrow field of view leads to missed movement.
Clock drift and timing
Device clock drift is reduced using server receipt time, IP based location, and heartbeat monitoring.
Analytics use UTC so comparisons are stable across daylight changes.
Offline behaviour
Kiosks: events queue in localStorage and upload later; numbers update on next refresh.
Cameras: no buffering when offline; gaps appear as lower totals.
Quality signals in the UI
The dashboard does not surface sampling warnings or thresholds. If numbers look unusual, check device online status, placement, lighting, and your filters.
Practical ways to improve quality
Mount cameras high and angle slightly downward to reduce occlusion.
Avoid mirrors and direct backlighting in the scene.
Validate with a quick walk test after drawing zones or lines.
Keep zones simple: prefer 4 or fewer per camera for clarity.
Use the same time window when comparing periods.
Retention and deletion
Analytics events and configuration
Sessions, experiences, and campaigns use soft delete archiving by default.
No fixed retention on raw analytics events unless agreed; enterprise retention can be configured.
Camera frames
If frame storage is enabled for security or validation, frames are deleted after 6 months.
Heatmaps
We store computed overlays, coordinates, and density values, not full video.
Exports
CSV files you download are under your control. Store and delete them according to your policy.
Right to erasure
Submit requests through support. We will act on applicable records and logs as your Processor.
Privacy posture
No facial recognition and no identity tracking.
Cameras are processed server side and anonymised immediately after detection runs.
Kiosk analytics redact sensitive inputs. Local caches can be encrypted at rest for enterprise.
Access is role based: Admin global, Manager and Viewer scoped to assigned countries.
Transport security: HTTPS with TLS 1.2 or 1.3.
Roles and access
Admin: view all analytics, audit logs, and configuration.
Manager: read and write within assigned countries.
Viewer: read only within assigned countries.
Checklist for compliance and quality
Post an in-store notice about analytics in use.
Run a DPIA if required by your regulator or policy.
Review placement and lighting for each camera.
Keep kiosk media optimised and devices online daily.
Set internal retention rules for exports and reports.