A couple of years ago I spoke with Thinknear CEO Eli Portnoy about location accuracy in mobile advertising. My recollection is that he said between 5% and 10% of impressions carried a lat-long and many of those were inaccurate — often derived from a centroid and converted into a lat-long.
Thinknear was acquired by Telenav in 2012. The company currently offers what it calls “location filtered” mobile impressions, most of which come from exchanges. It’s positioned as a local-mobile ad network competing with YP, xAd, JiWire, Verve and others.
Portnoy and I recently revisited the lat-long conversation over the phone. He told me that today roughly 68% of mobile display impressions carry a lat-long. In part this is because of increasing demand for location among marketers. (Marketers are starting to “get” location.)
Portnoy went on to explain, however, that — one again — most of these impressions are inaccurate. This is the “dirty secret” (my word not his) of mobile advertising when it comes to location.
Thinknear not long ago conducted a test to see what percentage of mobile display impressions carrying a lat-long were accurate. The company examined “tens of millions of impressions.” The specific inquiry was, of the 68% of lat-long enabled display impressions, what percentage were accurate to within 100 meters?
What it found was the following (of the 68%):
- 32% of impressions were accurate to within 100 meters
- 42% were off by 3,200 meters or more (2 miles)
- 26% were off by more than 10,000 meters, which is more than 6 miles
The implications of this are fairly obvious. Mobile marketers may think they’re reaching their targets but in a majority of cases this may not be happening.
One of the principal ways that location is being used by mobile marketers and ad networks is as an audience targeting mechanism. Audiences are inferred from location over time (or in the moment sometimes). Yet if those inferences are based on bad lat-longs the audience being reached may be quite different from the audience targeted, as illustrated in the following (humorous) Thinknear slide:
I asked Factual CEO Gil Elbaz about this problem and he assured me that Factual uses algorithms and other methodologies to correct for location inaccuracy. Other location-based platforms and ad networks would likely say similar things, although I haven’t put this conundrum to each one.
Thinknear makes a number of suggestions to address the problem. Among them:
- Asking users to validate location
- Scoring apps re location accuracy (which the company does today)
- Publisher audits to determine passed-location accuracy
Getting location right for mobile advertising is an obvious imperative. In case the “why” is not obvious, however, Thinknear (like others) reports big improvements in engagement, CTR and post-click activity based on proximity.
For networks and data providers building audience profiles, using location as a foundation, accuracy is also critical as the “grandma” slide above illustrates.
An MMA sponsored webinar last week featuring xAd, YP and Verve illustrated that use of location in mobile ads is a powerful but complex and nuanced matter. Strategies and tactics will vary by consumer use case and vertical. Indeed, location targeting “best practices” are emerging but haven’t yet fully emerged.
Regardless, for the consumer, location signals relevance. Location or context (location + time, weather, etc.) can be used to customize targeting, ad creative and messaging in ways that make ads much more relevant and likely to perform. Getting location right is the cornerstone of all of this.