Location observations collected from mobile operating systems contain a horizontal accuracy attribute that defines the level of certainty of the recorded latitude and longitude — the smaller the horizontal accuracy, the more precise the observation.
GPS and related technologies allow for the measurement of a device's specific location. The accuracy of the measurement is not constant. At the time of the measurement, the device will also calculate the horizontal accuracy of the measurement and record that as part of the observation.
Horizontal accuracy is typically measured in meters, and it represents the radius of the margin of error of the measurement. Users may be familiar with the concept based on how it is presented in popular map applications. The smaller dark blue circle represents the latitude and longitude as measured by the device. The lighter blue circle represents the horizontal accuracy. In simple terms, the device thinks it is where the dark circle is, but it may not be exactly there, but rather within the lighter blue circle. The larger the light blue circle is, and the greater the value of the horizontal accuracy, the less confidence the device has in the stated latitude and longitude.
Uses in Narrative
Horizontal accuracy is an important concept related to location data in Narrative's Data Streaming Platform. It allows users of the platform to apply filters to line items based on accuracy. Additionally, buyer's of location data can receive the horizontal accuracy as an attribute on any location observations they purchase.
The most common use case for leveraging horizontal accuracy within Narrative is to use it as a filter on buy orders and sell orders. Filters allow customers to make sure they are only buying or selling observations that meet certain conditions.
Buyers will most often use this filter to make sure they're only buying location records that they have confidence in. If a record's horizontal accuracy is greater than 500m, for example, its usefulness in predicting behavior may be limited.
While it can be useful to filter horizontal accuracy to ensure precise measurements, the lower the value of the horizontal accuracy filter the less data that will be available.
Sellers can use filters to create different go to market and pricing strategies. As an example, a seller may sell records with a low margin of error -- 5m horizontal accuracy -- at a greater price than those with a high margin of error.
Differences by Operating System
Major mobile operating systems have some nuances in how the data is collected and reported, which are essential to understand for anyone leveraging horizontal accuracy as part of their data model. Essential for both measurements is that the horizontal accuracy does not represent full confidence that the device is within the radius, meaning it may be outside of the given value.
Android defines horizontal accuracy as the radius that provides a 68% (1σ) confidence that the device is within that radius.
iOS does not define the probability that observations occur within the horizontal accuracy radius. A consensus exists that it is at least as precise as Android.