Location Inaccuracy Is A Bigger Problem Than Fraud

badlocationThere are myriad reasons why location data can be defective – and fraud is just one small piece of that growing problem.

Growing because the lack of industry standards around the collection and use of location data leads to a fair amount of misunderstanding, unintentional discrepancies and widespread inaccuracy.

Location fraud is the easy part,” said Ran Ben-Yair, CEO of location data startup Ubimo.

By easy, Ben-Yair is referring to the quashing of unusable location data.

While including location data boosts the price of inventory on the open exchange, most companies that peddle location-based advertising services use algorithms and rule-based logic to easily filter out the obviously bad data.

But the nuances around precision and accuracy pose the greater challenge, making fraud one spoke on an ever-turning wheel of inaccuracy.

X Marks The A Spot

Location seems like a simple concept – what could be more binary than a place? You’re either there or you’re not, but mobile introduces a complexity. “Nobody stands still,” said Mike Schneider, VP of marketing at location services company Skyhook Wireless.

This is where the interplay between precision and accuracy enter the scene, Schneider said. Precision is the range of possible error, while accuracy is whether or not you hit your mark. It’s possible to be highly precise and highly inaccurate at the same time, for example, as in when an advertiser is able to precisely target a specific location – it just happens to be the wrong location.

Low precision and low accuracy means the data is all over the map and therefore of no use. Low precision and high accuracy is like peppering the area around a bullseye with bullets without hitting the target, whereas high precision and low accuracy would mean that you’re in the wrong place, but you’re hitting that incorrect target over and over again.

High precision and high accuracy means you were spot on – but that doesn’t happen in most cases. A location quality index released by Telenav-owned hyperlocal data company Thinknear in May found that the quality of location data leaves something to be desired. Only 35% of impressions were accurate within 100 meters, while 19% of impressions were six miles off target.

There are umpteen ways that bad location data can be introduced into the system.

Truncated coordinates, for example, can pose a problem. According to location data provider Factual, coordinates with less than four decimal points can only be accurate to within 100 meters, which would, for example, make real-time geofencing impossible. Each deleted decimal point makes the resultant location more and more general.

“You really want to store all your decimal points on latitude and longitude,” Schneider advised.

In some cases, coordinates get inadvertently transposed, which could mean the difference between the middle of the ocean and a retail location in Buenos Aires.

In other cases, an access point – say a much-used Wi-Fi hotspot or cell tower that is actually located in a highly populated area – is mistakenly coded as being located in the middle of nowhere. One example uncovered by Factual corresponded to an isolated section of farmland in Kansas that had somehow started to rival major metropolitan cities for volume of mobile traffic.

Speaking of coding, some developers will purposely tinker with how their app accesses location services with UX in mind. Running location services in the background at all times drains a user’s battery. So, rather than tracking users as they approach a Best Buy location, for example, the app will identify them as having already arrived at the Best Buy when they’re just in the vicinity.

And sometimes, your phone just gets the jitters – the digital analogy to a dropped call. An app might not load properly or it might suffer an interrupted data request, so the user appears to be standing at the top of the Eiffel Tower one moment and in Montmartre 30 seconds later, and then back again.

“First, think about how many times a day your phone doesn’t perform optimally and then think about what that means from a data collection perspective,” said Shaina Boone, managing director of marketing decision sciences at Omnicom media agency OMD. “Each time that happens, it causes a disruption in the data source.”

Golden Opportunity

The fact is that bad location data isn’t always necessarily generated by bad actors – but it does put down the welcome mat for opportunists. According to BIA/Kelsey, location-based ad spend is set to hit $6.7 billion this year, so there’s money to be made.

“Mobile use is increasing and location inaccuracy is such a difficult problem to navigate that the perpetrators of fraud can piggyback on the problem because it’s hard to detect. Location data fraud is like a gold mine,” said Boone, who noted that location is still too much of the “Wild West” right now for her to suggest that clients include it in their media plans in any kind of meaningful way.

Stamp out the inaccuracy, however, and location data fraud becomes far less lucrative for fraudsters and shady publishers.

But are publishers, who see short-term revenue gains from trading inaccurate or fraudulent location data, incentivized to fix the problem? Ken Harlan, CEO and founder of mobile ad network MobileFuse, which uses location to send contextual messages, thinks the industry is.

“As programmatic buying platforms become more aware of the accuracy of data and are connected to multiple vendors, there is a shift of budgets toward sources with more accurate data,” Harlan said.

Ubimo’s Ben-Yair is seeing a similar trend.

“Location fraud is something that will decrease over time as exchanges make an effort to clean up the data and stop working with bad actors,” Ben Yair said. “There is a competition between the exchanges, they all want the best-quality traffic. They’re cooperating with us and they’re cooperating with the buy side. For example, they listen to our feedback about who is misbehaving.”

While no location standards exist at the moment, several industry committees are plugging away at the problem, including the IAB’s Mobile Location Data Working Group, and the Mobile Marketing Association’s Mobile Location Data Accuracy Group, on which Ben-Yair and Ubimo maintain a seat.

That said, location accuracy isn’t a quick fix, and in the meantime it’s better safe than sorry, Dstillery VP of media ops Alec Greenberg told AdExchanger in a previous interview.

"We throw out 50% to 70% of all the GPS coordinates we see every day because they’re questionable,” Greenberg said. “That’s a huge percentage.”


Popular On AdExchanger Right Now:

2 Comments

  1. Able_Magwitch

    I worked in sales for the "largest independent mobile advertising network" and I can tell you that the volume of inaccurate location calls was staggering. When agency folk ask for a geo-targeted campaign this company would creatively obscure the number of ads served to "unknown." That being said, most media buyers either did not have the bandwidth to delve into and/or didn't care. Don't get me started on "location expert" networks either as they're mostly marketing spin. The IAB did a nice white paper on how location is determined in mobile -- buyers should read that first and THEN take the meeting with the reps!

    Reply
  2. We did a trial run with a mobile carrier which provided us with their location data based upon cell tower information.
    It turned out that 98% of all location data from adexchanges was utterly useless.
    In order to boost their revenue, some adexchanges have been adding location data based upon IP databases.
    For mobile devices IP databases have absolutely no value because there is no relation between IP and the actual location of the device.

    We found 35,000 people to be in the carrier's datacenter around 22:30 according to the adexhange location data.
    Interestingly, this data center is off limits after 20:00 for security reasons.
    Not that 35,000 people would even fit in this place anyway.
    So beware what you're paying for.

    Reply

Add a comment

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>