Skip to main content
Facilities teams need to know how spaces are actually used: peak hours, dead zones, desk utilization. CSI analytics provides anonymous utilization signals from the radio environment without installing cameras or tracking employee devices.

​
What to measure

Binary occupancy over time is the robust foundation. Aggregate it into four core metrics:
  • Utilization rate — fraction of time a zone was occupied during business hours.
  • Peak occupancy periods — when does this room actually fill up?
  • Dwell time — how long does occupancy persist once triggered?
  • Zone comparison — which floors, wings, or desks see traffic?
These metrics drive real decisions: rightsizing meeting rooms, adjusting HVAC schedules, reallocating desk inventory. None require knowing who was present.

​
Temporal aggregation

Raw CSI events are noisy at second-level resolution. Analytics pipelines aggregate over minutes to hours.
1

Apply debounce and hysteresis

Filter rapid toggles at the event layer, same as smart-home automation.
2

Bucket into time windows

Aggregate into 15-minute, hourly, or daily windows.
3

Compute utilization

Calculate occupied samples divided by total samples per window.
4

Visualize

Render as heatmaps, trend lines, or zone rankings.
Motion intensity adds a secondary axis (high-traffic versus low-traffic occupancy) when activity detection is available.

​
Anonymous CSI vs MAC analytics

MAC-address tracking logs device identifiers. This creates three problems:
  • One person with three devices inflates the count.
  • Five people sharing one screen undercounts.
  • Device identifiers may be personal data under GDPR.
CSI analytics reads how bodies disturb the signal. No identifiers are captured or stored. The output is zone-level utilization, not individual tracking. For regulatory context, this is closer to a people-counter beam than a WiFi probe request log.

​
Honest limits on headcount

Binary occupancy per zone is reliable. Exact person count is not: multipath superposition from multiple bodies creates ambiguous CSI patterns without spatial diversity. WiFree achieved 92.8% crowd-count accuracy with transfer-kernel learning on building-scale CSI, but that required calibrated multi-AP infrastructure, not a single ESP32 node. The detection ladder places person count at rung 7. Design dashboards around utilization and presence trends, not exact headcounts. Estimates are possible with multiple nodes and calibration, but treat them as approximate.

Occupancy detection

The underlying signal for analytics

Device-free occupancy sensing

Keeping analytics stable with baselines

Getting started

Deploy your first analytics pipeline