Deterministic vs. Panel Data
Deterministic data is data built on actual consumer behavior. In contrast, panel data is modeled and based on a sample set of users. If deterministic data is available – use it. Deterministic data, which reflects what consumers really do in their daily lives, will always be more accurate than panel data, which makes assumptions about the behavior of consumers with similar characteristics. However, not all deterministic data is created equal. For this reason, you’ll also want to understand how your data provider is validating their results.
Let’s imagine that we’re launching a new campaign promoting an exercise bike. We want to reach folks who regularly go to spinning classes, since we know this set of consumers bikes for exercise. There are multiple techniques used by location data providers to record a consumer attendance at a location, and some techniques are more precise than others. If the geo-fence at that location is tightly drawn – ideally, a hand-drawn polygon that reflects the actual size of the studio – you can be reasonably sure that mobile devices captured within the geo-fence reflect studio attendees.
If the geo-fence is dropped as a pin with surrounding radius, however, the geo-fence may overlap adjacent businesses, resulting in attendances captured at the children’s toy store or insurance company next door. Some data providers use grid systems that register the activity of consumers observed at all businesses within a city block-sized area. While useful to understand the composition of a shopping center or neighborhood, it’s not a good fit for our sample campaign.
For this reason, it’s essential to understand the precision of the geo-fences used to capture consumer attendances. Polygonal geo-fences are the most precise, and come closest to reflecting the actual size of the venue in question. Other methods will yield audiences with a greater margin of error, reflecting the geo-fence’s overlap with adjacent venues or businesses.
It’s also important to understand what, if any, metadata is used to validate a consumer attendance. Returning to our example campaign, we’ve verified that we have a well-drawn geo-fence around the spinning studio. But what happens if the same studio also hosts yoga, ballet and self-defense classes? Unless the location data provider also has a record of associated events and related metadata, the geo-fence will capture all attendances at the venue, and not just those who attend spinning classes. Take the time to ask what processing is being done on the backend to exclude people who live in the apartments above, or who work at the front desk every day. These attributes will affect the quality of your audience – and the performance of your campaign.
QUESTIONS TO ASK:
Is your data deterministic in nature, or panel-based?
What type of geo-fences are used? Grids, pin and radius, polygons?
What metadata is provided for each geo-fence?
Does your metadata include events scheduled at the venue?
Next: Scale & Reach
Curious about Gravy Audiences? Download the full list of our ready-to-use location-based audiences today, or contact us at email@example.com for more information.