Worker Activity Index
“ Worker Activity” is a standardised metric that allows the comparison of worker activity levels between office buildings, irrespective of building size and vacancy level. The data is updated weekly in propella.office, so subscribers can profile office workers and follow trends in changing patterns of office worker behaviour.
Worker Activity is a function within the application that provides users with a measure of worker device activity within the geofenced building area. Device activity can be thought of as "pings" from the de-identified devices (mobile phones) that our algorithms have associated with likely workers within the subject building. The Worker Activity Index (WAI) is a derived metric that aggregates the worker device events for a time period, normalises this data against other people counting data to remove some underlying variability in the mobile data, and then accounts for the area (net lettable area, NLA) of the building to standardise across buildings of different sizes.
The methodology to derive the Worker Activity Index is as follows:
(1) Building/Mobile Data Intersection: We create a digital boundary unique to each office building which is referred to as the building’s ‘geofence’ – refer Figure 1. We then use the geofence to collect a statistically significant sample of mobile data found within the building envelope over the study period.
Figure 1: Example of a geofenced building showing mobile events detected within the building.
(2) Worker Identification: By observing each unique mobile device’s interaction with the geofenced building over the study period, we probabilistically categorise devices that exhibit ‘worker’ or ‘visitor’ behaviour using our proprietary “worker” algorithm.
(3) Data Aggregation and Smoothing: The mobile data is then subset to only consider activity from worker devices before it is aggregated into daily bins. A three week simple moving average is then applied to the data to provide sufficient sample size to reduce volatility in estimates.
(4) Global Sample Normalisation: Within the mobile data there exists random fluctuations in mobile activity linked to macro behavioural trends associated with app usage, mobile software updates & population sample size. If left uncorrected, studies using mobile data can become biased toward periods of favourable sample size. To correct for this, propella.ai uses composites of unbiased real world signals (e.g. people counting data provided by local council) to normalise our mobile data and remove such bias.
(5) Worker Activity Indexation: The normalised worker activity data is divided by the square meters of NLA for the building to produce the Worker Activity Index. This is a numeric value typically between 200 and 0. It is NOT an attempt to measure the percentage of building occupancy. This new metric allows comparison of worker activity levels between office buildings and the CBD average, irrespective of building size and vacancy level.