Precision can be a tool for truth, but it can also be used to convincingly stretch the truth. People intuitively assume that very precise numbers are for some reason more accurate than less precise numbers. It’s what we were all taught in high school math and science classes. That’s why we call the digits following the decimal place “significant figures.”
Sportswriter Gregg Easterbrook frequently addresses this obsession with significant figures in his “Tuesday Morning Quarterback” column for ESPN.com. “When one draft prospect is said to run a 4.58 and another a 4.59,” he wrote in May, “draftniks have something to talk about—though only in track and swimming might hundredths of seconds merit attention. Americans love absurd precision.”
It’s precisely — pun intended — for this reason that the location-based marketing world is full of marketing-speak, emphasizing an ever-increasing degree of precision rather than the more important degree of confidence. The location-based marketing space is crowded; competitors are continually looking for differentiation. The most commonly evolving arrow in their quivers is the precision with which a mobile user’s location can be identified and targeted. The industry has evolved from targeting zip codes, to neighborhoods, to 100-meter tiles, and now we see claims of targeting within four feet of a mobile user. Ever-increasing degrees of precision are appealing to marketers, and of course it’s exciting that technology permits this level of exactitude, but mobile campaign managers should further explore the realities (and complexities) of location data.
Let’s get the facts straight: is it possible to consistently and accurately target an individual down to a four-foot level of precision? If we’re talking beacon technology with a dedicated app in a limited-scale environment, then sure, it’s possible. But the real discussion is about the high-volume programmatic location data the ad-tech industry uses to target ads based on a device’s system-based location services — a combination of cell tower triangulation, geo-tagged WiFi networks, and/or GPS. This, not beacons, is what we mean when we say “location” for the majority of mobile ad inventory. The degree of precision calculated for location services-derived data is measured through the number of decimal places included in a lat/long of an ad request. Three decimal places give us a user’s location to within around 100 meters. Five decimal places imply precision to within around a meter (about three feet). And beyond that, we’re splitting hairs.
Ironically, inaccurate sources of data such as user registration are often presented with the greatest level of precision (i.e., the most lat/long decimal places), because they have been pre-processed as opposed to being pulled in real time. Lower-quality data sources involve indirect methods for tagging location such as reverse-geo lookups from IP address, or user registration data, which tends to be very outdated and often resolves to centroids associated with fairly broad geographic areas. In order to get both the degree of precision given by these lesser sources of data and the accuracy provided by more superior forms—so that the promised four feet is a reliable four feet—publishers must expend much in cycle time, battery payload, and bandwidth.
The good news for brands is that targeting to within four feet isn’t necessary. There are certainly use cases in which a brand may want to trade off the scale of a campaign for increased precision, and in those cases they should work with a capable and transparent technology provider. But the 100-meter “tile size” has become the location industry’s default norm for a reason. Not only is 100 meters sufficient for most location-based marketing needs, that distance presents an accuracy level that is likely to be of high fidelity.
We spend a lot of time focusing on accuracy, because that’s what matters. Location data is the core of mobile advertising, and brands will benefit significantly more from an ad-tech industry focused on delivering accuracy over absurd precision.