Geopath Mobile Device Location & Movement Patterns

Owned by Kevin Narine
May 31, 2023
2 min read

Standard Data accesses device activity from hundreds of millions of devices across multiple location SDK aggregators. Data from these mobile devices can inform us about the movement of the population at the national and local level, such as traffic trends, average miles traveled, exposure to out-of-home inventory, and place of residence for additional population weighting purposes.

These data feeds provide robust information from more than 100 million unique devices in the United States every month. Intermx culls those data down to a statistically significant, representative sample of the best devices available to generate the data utilized by Geopath. This sample ranges between seven and ten million qualified devices each week that are used to understand the travel patterns of the population as a whole.

 

Location Data Processing

Below is an outline of the steps that are taken on a daily basis to generate the refined feed of device location data.

1. Importing

The processing pipeline starts with importing raw Mobile location feeds from our providers. These data feeds provide a basic time-series of location data from many unique mobile devices seen throughout the U.S. (other areas or providers may be turned on as desired). These feeds fundamentally provide billions of individual locations with attributes describing,

  1. Device ID: A Unique Identifier for each mobile device. This ID may be provided by the operating system, device hardware, or the application itself. Most commonly a user controlled advertising id has been used which is available from both the Android (AAID - Android Advertising ID) and IOS (IDFA - Identifier for Advertisers) operating systems.

  2. Timestamp: UTC timestamp when a location for the device was logged.

  3. Latitude and Longitude: The location of the device at the timestamp.

  4. Horizontal Precision: Reported precision of the location in meters.

2. Ingesting

In this step we ingest the imported datasets into a data warehouse for processing. During this ingestion, each location is considered against other locations for the device logged around the same time to remove the noise and thoroughly vet the accuracy of the provided locations. The locations with imprecise horizontal accuracy, impossible latitude and longitudes, and inconsistent locations relative to others around the same time are marked as error sightings and their location data is removed.

3. Combining and Aggregating

Next, we blend the data from all providers into a single collection. This allows us to reliably blend locations received across all providers to create the richest activity profile for a device possible. During this process the locations are again vetted for consistency to continue to increase our confidence in observing a single device in at specific location that can be substantiated by different providers independently. This leaves us with a refined set of device-locations with which to develop a to feed the population analytics.

Â