People counting accuracy: what the numbers mean by method
How accurate people counting is depends on method, placement and calibration, so no single percentage answers the question. This guide covers what accuracy means per method, what a quoted figure does and does not prove, and how any claim is verified.
Accuracy is the first question buyers ask about people counting, and the one suppliers answer most loosely. A single percentage figure means very little on its own, because accuracy is a property of the method, the placement and the question, not of the brand.
Accuracy by method
Every counting method has a sweet spot, and “accuracy” means something different at a doorway than across a floor:
| Method | Strongest at | Where it is not the answer |
|---|---|---|
| Wi-Fi | Whole-space coverage: dwell, paths and zone correlation across an entire store, mall or district, on hardware you already run | Exact line-crossing in and out at a single door |
| 3D sensor | Exact directional in and out at a doorway, real-time occupancy | Behaviour across a whole space |
| AI camera | Point counting with standard camera hardware, analytics-only, no images stored | Whole-space journeys without many sensors |
| LiDAR | Door and outdoor counting, privacy-clean line detection | Whole-space dwell and paths indoors |
What a percentage figure does and doesn’t mean
A figure quoted without the venue, the placement and the calibration behind it is a lab number, measured somewhere other than your store. The practical test is whether the count is accurate enough, and consistent enough, for the decision you are about to make. A count that is exact at one door will not tell you which zone lost visitors after the refit, and whole-space coverage will not settle a legal capacity limit at a turnstile.
How Bumbee Labs verifies accuracy
Every deployment is verified with manual control measurements matched to our own data, and the model is tuned to the space until the numbers mirror what actually happened. We apply the same verification to every method we operate. Our Wi-Fi method is the only footfall method in Europe approved by a data protection authority, the Swedish IMY.
Questions to ask any supplier
- Which authority has reviewed the method? “Compliant” is self-declared; a review by a named authority is a different category of evidence.
- What leaves the sensor? Counts and aggregated statistics, or identifiers and footage?
- How is the deployment calibrated? Are manual control measurements matched to the system’s own data?
- What does the quoted figure assume? Ask for the venue, placement, traffic mix and season behind it.
Choosing the right accuracy for the decision
A capacity limit at a turnstile needs point precision at that door. A refit decision needs whole-space dwell and paths, while a staffing decision needs the hourly curve. The methods are often combined, each doing the part it does best, with sensor ground truth calibrating the whole-space model. How to count visitors sets out the comparison.
Getting started
Book a demo and we’ll show accuracy against manual control counts on a venue like yours, method by method.
At Ray our motto is customer success, and we are excited about our partnership with Bumbee Labs. Complimenting the seamless Wireless provided by Ray, our customers will now be able to leverage the deep analytics provided by Bumbee Labs and take informed business decisions to aid their growth.
Frequently asked questions
How accurate is people counting?
It depends on the method, the venue and the placement, not on the brand. At a doorway, 3D sensors and AI cameras give exact directional in and out; across a whole space, Wi-Fi covers dwell, paths and zones that no doorway sensor can see. Every Bumbee Labs deployment is calibrated against manual control measurements so the numbers mirror what actually happened.
Which counting method is most accurate?
That depends on what you need to count. Each method is strongest at something: sensors and cameras at point counts, Wi-Fi at whole-space behaviour, LiDAR at privacy-clean line detection, cellular at district scale. Most venues combine methods, each doing the part it does best.
How is accuracy verified?
Each deployment is checked with manual control measurements matched to the system's own data, and the model is tuned to the space. A supplier that cannot describe its calibration procedure is quoting a lab figure that was never measured in your venue.