It’s not rare to use automatic people counters with low accuracy¬†when performing capacity control, and thats due to the incorrect estimation of the margin of error.


It’s believed that using a 95% accuracy counter in an event where 50.000 attendees are expected generates a maximum error of +/-2.500 people counted (5% * 50.000 people), but usually it is not the case. Although this error could be accepted, the problem is that the actual percentage will probably be more than double. Thus, we have to point out that the error should be calculated on the number of entrances and exits, as an error can occur each time a person goes through an entrance. The real number of attendees in a big crowded event is counted with automatic people counters placed in the entrances, which is more economic than covering all the event’s area with sensors in the ceiling. Otherwise, if it was calculated using the already installed security cameras, results would be far from the accuracy desired due to the low cameras’ resolution.
Now, following the previous example, we see that the actual expected error in a mass-event such as a concert, where at least everyone will enter and exit once, is 5.000 people (5% * 100.000 entrances and exits), instead of 2.500. Looking at it like this, there’s a chance this figure is no longer acceptable. And there’s more, we can safely say part of the visitors will go out of the enclosed area and later go back inside, so let’s assume half of them do that. Then, the expected error is 7.500 people (5% * 150.000 entrances and exists) and this represents 15% of attendees expected.


Finally, the expected error is higher when more entrances and exists take place and this often occurs in small spaces, as there’s normally a lot of traffic in both directions. An example of this situation can be found in subway platforms. Let’s say a platform has a capacity of 200 people, assume 100 people arrive at the platform every 5 minutes, and a every 5 minutes a train arrives from which 100 people get off. Then, the expected error from a 95% accuracy counter, can reach +/-120 people in only one hour. In a situation like that, the human eye would be a better guarantee than a people counter. That is why, in small spaces like subway platforms or small conference rooms, the best solution is to cover all the area with automatic people counters in the ceiling. In that case, the expected error will be calculated with respect to the average capacity expected.


To conclude, we want to highlight the importance of knowing how the automatic people counters’ expected error is calculated and taking it into account when considering the maximum capacity allowed without assuming any risks.