Digital marketers have finally started to address the fraud that runs rampant in the online advertising industry. But as users move to mobile devices en masse, the currency of digital marketing — and subsequently, the face of fraud — has evolved from fake traffic, generated by bots, to fabricated data, passed on by publishers.
According to a new report, roughly two-thirds of the location data which mobile publishers pass on to advertising exchanges, and eventually to marketers, is inaccurate. The report, released by Thinknear, the mobile advertising wing of Telenav, found that of the 67% of mobile advertising impressions, which include a latitude-longitude data, only a third can accurately predict the location of a user down to 100 meters. And Eli Portnoy, general manager at Thinknear, says that a “sizable number of these impressions” are inaccurate up to 100,000 meters — roughly the distance between Baltimore and Washington D.C.
The company released the study in conjunction with Location Score, a technology which Thinknear built to provide an rating to predict the accuracy of location data for each impression. The product, and many of the insights from the report, stem from a model the company built over the past year by buying millions of mobile ad impressions and asking those users to share their location separately. The company then compared the location information the publisher included in the ad request with the user’s real location, and identified the discrepancy.
Marketers in the U.S. will spend nearly $4.5 billion on location-targeted advertising in 2014, accounting for roughly 40% of all spending on mobile marketing this year. The research represents the first comprehensive analysis of a problem that has remained something of an open secret at mobile companies as the ad technology industry has matured.
The percentage of mobile advertising impressions which include location data has exploded in recent years, ballooning from roughly 10% in the early days to 67% today. The problem, says Portnoy, general manager at Thinknear, is that the growth far exceeds the rate at which publishers have gone through the process to formally obtain a user’s permission to obtain their location and pass that data to marketers.
“Publishers used to have very little incentive to collect location information because advertisers largely did not care about the data. But today, advertisers are willing to pay a substantial premium for inventory with location data, and so advertisers now have an incentive to pass on that information,” Portnoy told me in an interview. “The issue is that gaining access to a user’s location data is difficult, and so getting to that number [67%] is extremely improbable.
While it’s unclear to what extent publishers are actually fabricating location data, some techniques are more explicit than others. In many cases, it’s a matter of passing lower quality data off as something far more valuable. A publisher which does not have access to a user’s exact location can often still find the geographical center — or centroid — of the zip code or country which a user entered in a registration form. Many then pass those coordinates on to marketers as if they reflected the user’s current location.
In one of the more striking illustrations of the technique, Portnoy says a large portion of mobile advertising impressions reported originate from Topeka, Kansas. He says that publishers have converted users which they know are in the U.S., and are telling marketers that they are in Topeka simply because the city is the geographical center of the country.
The line between fraud and negligence is slightly less clear in mobile, partly because the the positioning algorithms that identify where we are in the real world vary in accuracy, and are inherently inconsistent. The core positioning algorithms use a hybrid approach which combines GPS data, wi-fi positioning, and cellular tower triangulation to to identify our locations — each of which has varying degrees of accuracy.
When a user opens an app indoors, for instance, the algorithm might rely on wi-fi positioning, a less accurate technique which triangulates a position based on the location of known nearby wireless routers, more heavily because he or she is out of reach of a GPS satellite. And while the technique is substantially less accurate than GPS, the data all looks the same to the marketer.
The emergence of standards is one of the critical challenges, which faces emerging marketplaces. In this sense, the mobile advertising industry is a victim of its own voracious ascent. Today, there are no standards requiring publishers to pass on so-called meta-data about the way in which the data was captured to marketers. Without that data, marketers are effectively buying a product in which there’s nearly always some variation in quality.
However, as marketers shift meaningful spending to mobile, fraud is increasingly translating into a meaningful financial liability. According to a December 2012 report from Nexage, one of the largest mobile exchanges, marketers pay between three and five times more for impressions with latitude-longitude data than generic run-of-network impressions.
Consider a hypothetical campaign in which a marketer spends $500,000 to buy location-based impressions to reach users around a competitor’s store. If 66% of the impressions are effectively inaccurate, then the marketer is paying $330,000 dollars for effectively run-of-market impressions, which should cost $110,000 (a third of the cost.) That means in a half million hyperlocal mobile campaign, a marketer will overpay by roughy $220,000, or 44% more than they theoretically should.
Portnoy declined to disclose the names of publishers which pass on the fraudulent data, but says that the company plans to reach out individually and if they do not cooperate may release more information accordingly. While he is hesitant to label the activities as fraud, he believes that it does pose a substantial threat to the growth of the industry.
“The problems feels a lot like the issues that the [online] advertising world had around viewability. If an ad is served, and no one sees it is that fraud or is it just shady practices,” says Portnoy, “Ultimately, it does not matter how you classify it — the advertiser doesn’t get what they paid for. The advertiser gets hurt, and so does the industry.”
Steven Jacobs is Street Fight’s deputy editor.