As Mobile Ad Volume Rises, Location Accuracy Decreases
Either developers aren’t doing enough to “get the data right” or they’re playing “fast and loose” with geo-data to capture higher ad prices, surmises Thinknear.
The ability to accurately serve a location-based ad to a consumer is getting worse, according to hyperlocal ad network Thinknear.
In the company’s third Location Score Index report covering Q4 2014, Thinknear analysts find the geomarketing industry increasingly challenged by the demand for more programmatic ad inventory on mobile platforms, particularly in-app advertising.
The amount of programmatic ad space has expanded significantly, with most major exchanges experiencing 2- to 3 times growth in volume over the past year, Thinknear estimates. But the upward trajectory for automated ad buys has also put a certain degree of pressure on developers.
Thinknear sees two possible scenarios at work: it’s possible that in their bid to offer more targeting options for advertisers, developers have been careless when it comes to providing an accurate path for location-based ads. The other possibility is that the people behind these apps are not being terribly honest with their advertising clients and are making claims of higher geo-targeting accuracy in order to charge higher prices.
For the most part, Thinknear would like to make it clear that in most cases, it’s not a matter of honesty as much as it is ignorance—developers see an opportunity to make more money, fast, and use the location they have on hand to do so, rather than trying to seek more accurate data.
The bottom line is that there is more high quality programmatic ad inventory available to marketers these days. The issue is that in finding that premium, more accurate location-based ad placements are even harder to find.
A Downward Trend
On average, the ad industry’s Location Score fell to 51 in Q4 2014 from 55 percent in Q3, according to Thinknear. The Index is based on a weighted 100-point scale. It’s worth noting that sequential comparisons like this one are sometimes not the best way to compare performance, since seasonality and other time-based issues can be a determining factor. But since Thinknear only introduced its Index last May, we’ll have to wait until more direct, year-over-year juxtaposition can be studied.
It should also be noted that Thinknear’s Index is based on “sampled and analyzed data from over a billion ad impressions and ran location accuracy tests on more than 500,000 consumer ad experiences.” The company’s expansive reach gives its analysis a great deal of credibility, but it is not necessarily the last, definitive word on the matter of location accuracy.
With that out of the way, there are many other reasons to heed Thinknear’s critique. For one thing, it clearly defines “Location Accuracy” as the “proximity of a user’s stated location per the ad request, compared to the user’s actual location in the real world.”
For example, as anyone who has even looked at a map on their smartphone and found the “blue dot” representing their location to be a block or two off can tell you, the “stated vs. actual location” of a target can vary by a matter of feet or even by hundreds of miles. “Identifying the user’s true location is the goal of any ad platform, as it enables enhanced targeting opportunities and better performance,” Thinknear’s report notes.
The company also notes that considering programmatic ads can be served in the fraction of a second, it doesn’t make much sense to claim to consistently track any given user down to a square meter tile on an internet-connected map.
Precision vs. Accuracy
Aside from trying to educate the industry on understanding how various sources of data affect location targeting, Thinknear’s other public service is to note the difference in meaning between “precision” and “accuracy.”
As Thinknear interprets the two terms, “precision relates to the granularity of the data: a DMA, a zip code, a city block, or a specific store location.”
On the flipside, “accuracy” has to do with determining if the data is actually correct or not. “So while a ‘precise’ ad request may specify a user’s location to within a matter of feet, that data is worthless if it is not accurate,” says Thinknear.
Consider The Source
When trying to tell the accuracy of a location-based ad, Thinknear advises considering the source first. Advertisers need to ask where the data actually came from — was it from an A-GPS (assisted global positioning) chip in a connected device, which is regarded as delivering the clearest signal about where an object happens to be in a given moment, or from wifi, which is widely deemed as the second most accurate connection.
However, as most people use networks at home or work, as opposed to walking or driving near a retailer, it’s not terribly actionable.
Cell towers are able to send off powerful data signals, by dint of their ubiquity in most densely populated locations. But with those dense populations come a large amount of tall buildings and structures, which creates equally high levels of static and can block a signal. And although computer IP addresses are very capable when it comes to loading and browsing websites, it’s a pretty weak data point in terms of pinpointing an individual phone in a large area.
But as Thinknear has often said, the worst method of extracting location data are the apps that ask for user registration. In this case, a developer asks for a zip code when a person downloads their app. But that zip code is centered around their home or work and therefore has no way of finding that person as they go about their day.
“Location data, regardless of the source, is typically translated into a latitude and longitude—a number followed by decimals, with each decimal point offering an extra layer of precision,” Thinknear notes. “One decimal point can narrow down a location to a city level, while five decimal points can theoretically narrow a location down to a square meter.
“While it is technically possible to identify a mobile user’s location within a single meter, this level of precision is generally not scalable and is often a result of inaccurate pre-processed data,” the report concludes.