# How is Occupancy measured?

- We have two ways of measuring occupancy in large spaces:

**Technique 1.** **Estimate occupancy using headcount**. This technique would involve instrumenting the entryways with sensors in headcount mode and subtracting # of exits from # of entries to *estimate* occupancy. So if a headcount sensor has sensed 100 entries and 90 exits in a space, the estimated occupancy would be 10 people at that point in time.

**Technique 2. Directly measured occupancy**. This technique involves aggregating the outputs of sensors that cover the entirety of the space with sensors in activity mode. So if a conference room has 2 activity sensors installed, with 1 of them showing 2 people and the other showing 1 person, the occupancy of that room is measured to be 3 people.

**Technique 1 vs Technique 2:**

- Technique 1 only calls for instrumenting entries and exits vs Technique 2, which needs to cover the entire space so Technique 1 can have a cost advantage.
- Technique 1 is more effective when the sample size is large and is prone to error or drift when the sample size is small:
- The Butlr headcount sensor's accuracy spec is 95%. So if there are 10 entries and 10 exits, it can be expected to count 19 of 20 and so the estimated occupancy of the room would be 1 instead of 0.
- The error/drift could also accumulate throughout a period of time (e.g.: 1 day) if there's an asymmetrical amount of errors between in-counts and out-counts. Some environmental and behavioral conditions might increase the accumulated error, for instance: drastic temperature difference between both side of the door; user hanging around at the door.
- Whereas if Technique 1 is implemented in a larger space where there are, say, 200 entries and exits, the 5% error is manageable relative to the sample size and there is smaller possibility of asymmetry of error and smaller error to sample ratio.