CBRE's 2026 Global Workplace and Occupancy Insights contains two numbers that look like they cannot both be true. Average global office utilisation is 53%, the highest since March 2020. The global occupancy ratio is 111%, meaning more employees are assigned to a location than there are physical workplaces. So the typical office is simultaneously half empty and oversubscribed. The resolution is timing: average peak occupancy sits at 80%, and that peak lands on Tuesday to Thursday. Mondays and Fridays are quiet. Anyone managing a building on the 53% average is planning for a day that does not exist.
That is the case for workplace occupancy analytics in one paragraph. Not a dashboard for its own sake, but a way to see where the 80% day actually hurts: which desks, which small rooms, and what time the canteen queue peaks. Below is how that plays out across the three spaces that generate the most complaints and the most avoidable cost.
Most organisations already have data. JLL's 2025 Global Occupancy Planning Benchmark found that 74% of organisations collect utilisation data, yet only 7% rate that data as excellent. The reason is usually the source. Booking systems, badge swipes and Wi-Fi association tell you what people intended or where their laptop connected. Presence sensing tells you what happened. The gap between the two is where the missing capacity sits, and it is a large gap: ghost bookings alone are estimated to waste 30 to 40% of meeting room capacity in most offices.
There is a second problem that only shows up once you install hardware. A desk occupancy sensor triggered by any object will happily report a laptop bag or a coat as an occupied desk. Presence has to mean a person is present, not that something is on the chair. Vemco Group has counted people since 2005 from its headquarters in Fredericia, Denmark, and the platform is sensor-agnostic: VemSpace and VemCount work with sensors already installed, including Xovis (Vemco is a GOLD partner), Milesight, Hikvision and AXIS. That matters for corporate real estate teams because the sensors bought for a pilot two years ago rarely need to be replaced to get person-level presence rather than object detection.
CBRE reports that 48% of organisations now target a desk sharing ratio between 1.01 and 1.49 people per seat, up from 21% in 2024, and that no respondents target one desk per person. Sharing is the plan everywhere. The friction is behavioural. Some organisations use a rule that if you leave your desk for a meeting longer than 45 minutes, you clear it so a colleague can use it. It is a sensible rule. It also fails without measurement, because it depends entirely on goodwill, and the person who follows it is the one who loses their desk while the colleague who left a jacket on the chair keeps theirs.
Occupancy monitoring at desk level changes the incentive. When live availability is visible on a floor plan, staff stop hunting and start going to the free cluster. Vemco customers embed live occupancy figures as a widget on an intranet page, so employees see current status themselves before they walk over. Staff exclusion keeps cleaners and facilities rounds out of the figures, otherwise a 6 a.m. cleaning pass looks like early adopters on every floor.
On accuracy: the contractual minimum is 96%, and typical results are 98 to 99% when lighting, layout and movement patterns allow. Two things follow from that range. It is comfortably enough to decide whether to consolidate a floor or adjust a sharing ratio. It must never be used to check whether a named employee was at their desk. Vemco counts people; it does not identify them, and there is no facial recognition anywhere in the stack. HR teams should insist on that boundary before the first sensor goes up.
Freespace's 2026 data shows 75% of meetings globally involve only two or three people, while most offices are still dominated by large formal meeting rooms. The result is predictable: the eight-seat rooms sit half booked and one-third used, while the two-person mute boxes are fought over from 10 a.m. onward. Space utilization data will show you the mismatch within a fortnight. Fixing the no-show pattern in the small rooms is harder, because a mute box that shows "booked" on the calendar is unavailable whether or not anyone is inside.
The established fix is sensor-triggered auto-release combined with a check-in requirement. Vemco's implementation: if nobody is detected in the space within the first 15 minutes of a booking, the booking is cancelled and the slot is released back into the customer's own booking system, so colleagues see it as free in the calendar they already use. Here is the observation most implementers learn the hard way. Detecting an empty room is the easy half. Returning the released slot to Outlook or Google Calendar, rather than to a separate facilities app nobody opens, is the half that actually changes behaviour. If the release only shows on a wall panel outside the room, you have solved the reporting problem and not the capacity problem.
Before 2020, according to Controliza, 85 to 90% of employees ate in the canteen daily and demand was predictable enough to plan on headcount. Hybrid attendance broke that. JLL puts 67% of office workers in the office one to four days a week, with only 15% in five days. A kitchen that still produces for 100% of headcount against roughly 60% actual attendance is wasting about 40% of production.
The arithmetic is not abstract. In a 2,000-person building at EUR 3 raw material cost per serving, 600 excess servings a day is about EUR 1,800 a day, EUR 36,000 a month and close to EUR 430,000 a year. That is before labour and disposal. Building occupancy counts from the entrance, available by 9:30, give the kitchen a same-day production number instead of a forecast from a booking tool that half the workforce ignores.
The employee side is simpler still. Nobody wants a canteen report. They want to know whether to go now or in twelve minutes. Real time occupancy monitoring shown on the intranet widget flattens the lunch peak with no staggered-lunch policy, no emails and no enforcement, because people make the choice themselves when they can see the queue.
An occupancy monitoring system that only produces reports is a cost. The return comes from what the data connects to, and from waiting long enough before acting on it.
Vemco's 55,000+ installations across 98+ countries count more than 60 million people a day, most of them in retail. Offices are quieter and better lit, and the questions are narrower: which desks, which boxes, what time for lunch. The method is the same. Count people, not objects. Measure what happened, not what was booked. Then give the answer back to the people who can act on it, in the tool they already have open.
What is an occupancy sensor? An occupancy sensor is a device that detects whether people are present in a defined space, such as a desk, a meeting room, a canteen or a whole floor. In workplace analytics it is used to record actual presence over time rather than bookings or badge data. The important distinction is between sensors that detect any object and sensors that detect people, since only the latter gives usable desk-level figures.
How do occupancy sensors work? Most workplace sensors use either infrared, radar or 3D stereo-vision to detect human presence and, in the case of counting sensors at entrances, the direction of movement. The sensor sends anonymised counts or presence states to a platform such as VemCount or VemSpace, which aggregates them by space and time. No images or personal identifiers leave the device in a counting-only setup, which is what allows the data to be shared openly with employees.
If your office is running at 111% on paper and 53% in practice, the missing capacity is already in the building. Talk to Vemco about a desk, mute box and canteen pilot using the sensors you already have, with auto-release wired into your own booking system and live occupancy on your intranet within the first six to eight weeks.