Benchmarking Foot Traffic Normalization
We have published a dashboard in GO Analysis, benchmarking the performance of our foot traffic normalization algorithm, vs a ground-truth dataset of TSA security checkpoint foot traffic at 438 airports.
The following chart from the dashboard shows the percent changes in the normalized foot traffic timeseries from GO, compared to the raw data and TSA ground-truth. This demonstrates how the normalization algorithm provides a more accurate signal, compared to using the raw data alone.
Normalized foot traffic correlates better to TSA ground-truth, compared to raw data
The dashboard is available in GO Analysis under the demos/documentation/geolocation folder, as the Geolocation_Data_Performance_Metrics notebook.
Background on the foot traffic algorithm
The foot traffic algorithm counts unique devices at an AOI, using mobile geolocation data sources. GO provides normalization to correct for panel shifts in the raw data, which result from devices and apps being added / removed from the data source (a common occurrence).
Benefits of normalization for GO users:
- Remove noise in the data while preserving signal, allowing users to more accurately assess changes in foot traffic at a given location
- Better match the absolute metric of foot traffic, allowing users to better answer questions about the number of people at a given location rather than the number of device pings