Researched by BEAST Library Curator | Verified by Harper | Quality: 8/10

Occupancy's Three-Way Disagreement — A 2026 Reconciliation Methodology When Sensors Say 80%, Badges Say 45%, and Booking Says 60%

BLUF. The 2026 occupancy-intelligence stack has converged on a multi-source default: sensors + badge + WiFi + booking, all in one dashboard. The trouble is they routinely produce three different numbers for the same Tuesday at 2pm. VergeSense's May 2026 "Workplace Assistant" release explicitly tries to reconcile them; Occuspace, Butlr, XY Sense, and Density each take a different methodological view of who is right. For an FM, the question is no longer "which sensor should I buy" — it is "when my three data sources disagree, which one do I take to the lease decision?" This piece lays out the four-rule reconciliation methodology BEAST's CRE Soft-Service Squad uses to flag occupancy KPI Theater before it lands in a CFO deck, with explicit APAC privacy guardrails (PDPA, Taiwan PIPA, GDPR Article 4).

1. Why the disagreement exists — each source measures a different thing

The first move is to stop treating the three numbers as competing measurements of "the same thing." They are not. Each measures a distinct physical event:

SourceWhat it physically measuresTypical 2026 accuracySystematic bias
Sensor (thermal / radar / CV) Bodies physically present in a zone 92–97% (Butlr, XY Sense, Density 2026 benchmarks) Counts visitors, vendors, contractors as "occupants"; counts a multi-hour focus session as "1 desk used" regardless of how long
Badge / access control Entry events through credentialed turnstiles ~98% on entry; 60–70% as a proxy for "in the building right now" (tailgating, no badge-out) Misses tailgaters; misses everyone who badged in and went home at noon; double-counts re-entries
Booking system (Robin, Condeco, OfficeRnD) Intent — desks/rooms a person reserved ~99% as intent capture; 50–75% as occupancy proxy (no-show rate is the gap) No-show rate runs 25–50% on hot desks; counts ghost bookings until the auto-release fires
WiFi (Cisco Spaces, Aruba Location, Juniper Mist) Connected device count, AP-level 85–90% for portfolio trend; ~70% at room-level granularity Double-counts laptop + phone + tablet per person; misses anyone who turned off WiFi

The largest single source of FM confusion in 2026 is treating these as four competing measurements of "how full is the floor." They are four measurements of four different physical events. The reconciliation methodology starts by naming which event you actually care about for the decision in front of you.

2. Pick the right number for the decision you are about to make

Here is what I would do if this were my building. Match the source to the decision, not the other way around:

DecisionPrimary sourceWhy
Right-size the seat count before lease event Sensor (P85 peak) cross-checked with badge entry count You are sizing for physical bodies on the worst day, not for booking intent
HVAC pre-cool / setpoint schedule Sensor zone-level, 15-min granularity HVAC control loop needs body-presence ground truth, not bookings
Cafeteria / amenity capacity planning Badge entries + booking system for events You need the building's daily population, not the floor-level distribution
Hot-desk pool sizing Booking system actuals + sensor no-show detection You need the gap between intent and attendance to right-size the booking pool
RTO policy compliance reporting Badge data, sensor second Only badge data carries the audit trail HR needs; sensor data lacks identity binding (correctly)
Tenant experience / NPS root cause Sensor peak data + booking no-show rate Surveys complain about Tuesday seat-deficit; sensor + booking jointly explain it

This single move — naming the primary source per decision class — defuses 80% of the "but the dashboards disagree" CFO conversation before it starts.

3. The four reconciliation rules

When the same decision could legitimately use two sources, apply these rules in order:

  1. Sensor wins on physical-presence questions. If the question is "how many bodies are on the floor at peak," the sensor count is ground truth. Badge data systematically under-reports (no badge-out, tailgaters). Booking data systematically over-reports (ghost bookings).
  2. Badge wins on identity-bound questions. If the question is "did our RTO policy hit the 3-day target" or "is this team meeting their attendance commitment," only badge data carries the identity binding. Sensor data is anonymous by design (and should stay that way).
  3. Booking wins on intent / future questions. If the question is "what will Wednesday look like" or "should we open more amenity slots Thursday," booking intent is the leading indicator. Adjust by your measured no-show rate (typically 25–50% on hot desks; lower on desks assigned to specific people).
  4. WiFi wins on portfolio breadth. If the question spans a portfolio you don't have sensored end-to-end, WiFi-derived trend data fills the gap. Use it for portfolio-wide pattern detection, not for room-level decisions where it carries ~70% accuracy.

VergeSense's May 2026 Workplace Assistant explicitly encodes the first three rules; its unified analytics view tags each metric with the source-of-record so FM teams can show their work to finance. Occuspace and XY Sense ship comparable reconciliation surfaces. The methodology is converging across the vendor pool.

4. Spotting occupancy KPI Theater — three red-flag patterns

BEAST's CRE Soft-Service Squad runs a daily check we call occupancy KPI-Theater detection — three patterns that signal a number is being manufactured rather than measured:

If any of these three patterns appears in a deck headed to the CFO or board, the right FM move is to push the deck back for a reconciliation pass before it leaves the building.

5. The APAC privacy reconciliation gotcha

Multi-source reconciliation works on paper. The legal layer is where APAC deployments stumble:

The practical rule: in APAC deployments, run the badge–WiFi join behind a k-anonymity floor of k=5 minimum per zone per 15-min window. Below that threshold, suppress the joined count and serve the sensor-only number. The companion BEAST CRE-EN Privacy Broker writeup covers the differential-privacy noise budget needed for floors below k=5.

6. The 90-day reconciliation playbook

WindowActionVerification
Days 1–30 Inventory which of the four sources you have live. Tag each existing dashboard with its source-of-record. Disclose no-show rate next to every booking-derived "utilization" figure. Every FM dashboard has a source label. No utilization figure ships without its source named.
Days 31–60 Run a 2-week reconciliation diary: same Tuesday 2pm, log all available source numbers. Compute the spread. If >15 points across sources, you have a Pattern-A/B/C theater risk to fix. Reconciliation diary doc with measured spread per zone.
Days 61–90 For APAC sites, layer the k=5 anonymity floor on any joined badge–WiFi view. For non-APAC, document the GDPR processing register entry. Re-baseline the lease-decision number off P85 sensor peak, not booking or WiFi average. Lease-decision memo cites sensor P85 with badge cross-check, privacy review signed.

7. What this maps to in the BEAST stack

BEAST's CRE Soft-Service Squad runs occupancy KPI-Theater detection as a daily check; the CRE Space-Planning Squad consumes the reconciled P85 sensor number as its sizing input; the CRE-EN Privacy Broker enforces the k-anonymity floor before any joined view leaves the squad boundary. The methodology in this piece is the same one those squads execute on every multi-tenant brief. Ask our CRE AI Agent for a reconciliation diagnostic on your portfolio's current dashboards.

Sources

Have a question about this topic? Ask our CRE AI Agent →