The Extrapolated Visitor Report attempts to estimate real-world foot traffic based on the use of a panel high-emittance devices acting as a stand-in for the general population of smartphone-carrying adults in the United States.
Over the last few years, with the emergence of specialized SDKs built solely for the purpose of gathering mobile location data, Near's data store has grown more dense; the points-per-unique-device have grown from the early days when Near started collecting data in 2014. As a result, some types of analysis are now possible which weren’t able to be performed with confidence previously.
One such analysis is the Extrapolated Visitor Report which attempts to estimate real-world foot traffic based on the use of a panel high-emittance devices acting as a stand-in for the general population of smartphone-carrying adults in the United States. This tech note outlines the methodology used for this report
The Observability Metric
Near has created a proprietary device observability metric, conveniently called the Near Observability Metric, which provides a rubric for scoring devices by how many data points the device emits over the course of a set period of time. In layman’s terms, Near's observability metric asks if the device visited a location for some duration, what is the probability that the visit was observable, that is, that at least one of the observations from the device occurred during the visit?
In order to answer this question, a deep understanding of the underlying device behavior must be considered, allowing for various SDK types which collect data using varying rulesets, as well as accounting for the nonuniform probability of visits throughout the day. Thus, the observability metric is based on the temporal distribution of observations from a device over some time span, and does not depend on any particular model for the frequency or time dependence of the observations.
The Panel
Each week, after all observations from all sources have been received (which can be delayed by 5-7 days, every device in the Near ecosystem (200 Million in the United States) is scored using the Observability Metric. A panel is created each week of the devices which score the highest observability scores. An easy way to think of this panel is the most always-on of Near's devices.
The Extrapolation Report
When an Extrapolated Visitor Report is requested from the Vista Platform, a list of all the visitors to the study geography is first created. This list of visitors is then compared to each week’s panel visitors, identifying the list of panelists who visited a specific location in the timeframe. Each of these visitors acts as a stand-in for real-world foot traffic. Given the general representativeness of the geographic distribution of devices within the Near dataset¹, this straight-forward extrapolation from panel to real-world behavior is warranted.
The Future
The initial version of the Extrapolated Visitor Report uses an advanced scoring technique, a fixed-size panel, and thus a straight-forward extrapolation model. The Near Data Science team is working on creating a superior method of empaneling devices, that doesn’t rely on a fixed-panel size, and uses a more sophisticated share of panel methodology. This can be used to not only extrapolate to real-world visitation, but also be used to predict potential future sales volumes. Stay tuned for more progress in the coming months.
¹A white paper discussing the representativeness of Near's data is available here.