How Tamoco’s Normalized Footfall Methodology Measure Visits More Accurately

The problem:

Many visitation or footfall measurement methodologies that leverage location data come with some inherent flaws. Often, there is a compromise to be made between scale and accuracy.

Without the necessary scale, it can become difficult to identify actual trends in the data, with no guarantee that small sample sizes will dramatically skew any insights generated.

Higher accuracy filters ensure that false visits are removed from the dataset. Still, in many cases, this also filters out ‘true’ visits that do not pass the strict filters that we use to constitute a ‘visit’. 

The challenge:

These data points can contribute to more conclusive pools of visits data, which can then be used to measure footfall within a POI more accurately.

With this in mind, Tamoco set out to create a better visitation/footfall measurement solution. We needed to ensure that we could generate relevant scale without compromising the accuracy and precision that underlies our data methodology.

In this case, the challenge was to measure visitors to stores for a large chain across the US.

The data:

Our methodology for this project was as follows:

Step 1: Attribution of devices to POI

• We measure visitors (unique devices) that have been proximal to a store.
• To establish the area to monitor, we are drawing a smart polygon based on the streets surrounding the given store, plus limiting the areas to 150m (where polygons are too large).
• We filter matching data points by opening hours of the store.
• This gives us a large pool of devices that have been seen while the store is open either on the road coming into the store, in the store parking lot, or inside its building.

Step 2: Creation of control groups

• We establish a relationship between the number of visitors (found in step 1) with the number and the behaviors of selected unique devices for which the probability of visit to POI is near 100%.
• We use a clustering method to find those highly probable visitors and create a control group out of it.
  • These control groups are unique to each POI
  • The control group is created for the most “active” day of each month (based on Tamoco DAUS)

This is an example of a control group for the same store.

Step 3: Normalisation of data, creation of the footfall study

• We are then normalizing the visitors from step 1 using
  • Changes in daily active users in a larger region (in this case within the US state)
  • The control groups from Step 2
• As a result, Tamoco can understand:
  • the activity trend at the Store level by each hour, day, month, year
  • what % of visitors visited the store x number of times
  • what % of visitors also went to another POI
  • what % are living/working in the area/city

This dataset will then make up our data delivery. This can be delivered raw, or we offer powerful visualization products to bring the data to life.

This project has been an excellent opportunity to create a more robust visitation measurement product. While other methodologies that have smaller data pools might affect reliability in visits, our methodology takes a normalization approach, ensuring that these risks are removed from the measurement process.

For our customers, it means that they can more reliably identify trends in visitations to a POI. This project works great for retail brands looking to get footfall data that represent what is happening on the ground more accurately.

Get started with Tamoco Footfall measurement