As Location-Based Marketing Matures, Has It Gotten Any More Accurate?

The answer, as you might expect, says Thinknear’s Brett Kohn is: “it depends.”

The role of location signals to power real-time ad targeting, social media analytics, and Connected Intelligence voice-activated assistants has become an integral part of all marketing, as digital mapping supports roughly $1 trillion in business around the world.

Still, trying to determine how accurate those geo-data signals are remains a challenge for the brands, agencies, and platforms that are increasingly reliant on the promise of reaching the right consumer at the right time as marketers seek to develop ever more detailed profiles of how, when, and where people conduct their shopping.

For the past four years, location-based mobile ad platform Thinknear has issued its Location Score Index (LSI) to get a handle on how the variety of signal sources and datasets are performing. (The report can be downloaded here; registration required.)

In its latest report, Thinknear puts the overall LSI at… “43.”

While that number is in line with the Telenav-owned company’s Q2 LSI in 2016, it is notably lower than the 55 LSI in Q3 2015.

So does that mean location accuracy is getting worse? Or is it just getting better in some areas versus others? (This past week, The LSA’s Greg Sterling even put a price on “bad location data” and came up with an estimate of $1,000 — at least in one instance.)

Attempting answering those questions, Thinknear Co-President and GM Brett Kohn tells GeoMarketing, “It depends.”

And now, we’ll explain the levels of nuance at work here.

What Is Thinknear’s Location Score Index?

Thinknear’s Index is based on a weighted 100-point scale, with “100” being perfect geo-data accuracy.

Similar to an index such as the S&P 500, Thinknear describes the number as “an annual assessment for location accuracy, but over time it becomes a benchmark for how location data quality in the programmatic ecosystem is trending.” It’s worth noting that the LSI is based on a 100-point, non-linear scale, meaning that it’s easier for the industry score to grow from 25 to 35 than it is to grow from 75 to 85.

In addition, it’s important to note that Thinknear is measuring only programmatic bid stream data in the U.S. and Canada.

Location Signals: Which Is Better/Worse?

Furthermore, Thinknear’s findings reflect the complicated connection between multiple factors that affect location data accuracy, such as signal source (GPS signals, wi-fi, cell tower triangulation), environment (area density, skyline view, indoor or outdoor location), and personal use (location data access enabled, type of mobile app used, operating system usage).

The most important distinction impacting geo-data quality is whether it’s coming from an ad call or through an SDK.

With an “ad call,” the data is collected when someone opens a news app or a page in a mobile browser and an ad appears, thereby capturing that device’s location. When geo-data is gathered through an SDK, then the signal is coming from an app for weather, navigation, or social media and the user has opted in to always have their phone’s location data services turned “on.”

In the SDK example, the device is constantly updating and sending a geo-data signal – and as such, that data tends to be more accurate as a result.

Bigger Haystacks That Keep Multiplying

But the bottom line – and we are getting to it – is whether location data is getting better or worse.

In getting to its “43” score, Thinknear looked at the volume of ad inventory across five areas: Hyperlocal, Local, Regional, Multi-Regional, and National.

It found that most accurate segment was Regional, which showed that 31 percent of location data targeted in a “regional area” was accurate between 1,000 and 10,000 meters of the user’s true real-time location – or roughly, within 6 miles.

After that, ads seeking targets within a Hyperlocal geography was 30 percent accurate to within 100 meters of the user’s true real-time location, meaning that the designated target was found somewhere within a space the size of a football field.

Meanwhile Local volume was 5 percent accurate within 0.6 miles, while Multi-Regional was 21 percent (less than 60 miles), and National was 13 percent (a span greater than 60 miles).

While not quite apples-to-apples, geo-data analytics specialist PlaceIQ, in a May 2016 report, looked at 150 specific locations in five major cities and found that location data obtained via mobile smartphones is accurate up to 30 meters (or 93 feet. Or imagine seven Volkswagen Beetles parked one in front of the other.)

For Thinknear’s part, the goal it had was larger than that example, but also narrower than the location technology world beyond programmatic advertising.

“I would say the real message we’re trying to drive home with this report is twofold,” Kohn says. “One is there is actually a ton of high quality, highly accurate data available to marketers in the mobile ecosystem. But you really need the right tool-sets to find it. That’s been somewhat consistent over the years. The good data is there, but you really need good tools to find it. I would almost equate it to kind of a needle in the haystack metaphor. There are a lot more needles out there now to find than there were three or four years ago. But the size of the haystack keeps getting bigger and bigger.”

Location Accuracy Growing Pains

So as more platforms, more marketers, more devices, more locations, go into more programmatic advertising exchanges, the expansion of the space can make it seem as if the overall accuracy is declining. But it’s simply reflects the growing demand for more data by more platforms across more areas.

And as Kohn notes, the company’s score is meant to assess the quality of the data before platform companies filter out the “bad data.”

“Not all data is created equal,” said Joao Machado, director of Mobile at OMD, in a Placed report on geo-data accuracy in Sept. 2017. “You have to be very diligent in determining the accuracy of location data coming off the exchanges. Quality beats scale all day long and 1st party data is the gold standard in quality of location data. A horoscope app dumping location data into a Supply-Side Platform before landing in the exchanges is the example of what leads all of us to scratch our heads around lack of value around so much of what goes on in the market.”

Placed, which was acquired by Snap last summer to provide attribution services for Snapchat ad clients, based its report on its Placed Attribution product, which analyzes an audience of over 150 million devices that generate over 140 billion latitudes and longitudes on a monthly basis.

Distance + Density

In Thinknear’s LSI report, the company does take a closer look at three “case studies” to show from a practical standpoint how effective geo-data can be.

For example, automotive brands in the U.S., can derive significant value by mapping competitors’ dealership locations and assessing market-by-market dynamics.

“In the auto case study, what we were trying to do is find what data in the automotive space can be used to better understand the local market dynamics on a city by city basis,” Kohn says. “In the past, you might have a tier one or tier two automotive customer that says, ‘Give me a geofencing strategy across the board and nationwide.'”

In terms of showing those “competitive dynamics” from one place to another, in Los Angeles, for example, the average distance from one dealer to the next is about .6 miles. The city has highly concentrated groups of dealerships. Therefore, the way one might apply geotargeting in Los Angeles, and the types of audiences a dealer is interested in, is going to be different than what you would do in, say, Chicago whose area dealerships are more spread out.

“We look at things like demographics by neighborhood, not zip codes, but down at the hyperlocal level, to understand which locations are driving, say, Lexus purchases versus Hyundai purchases,” Kohn says.

“The way we target the neighborhoods, the way we target the audiences really do differ market by market based on how far people travel to get to dealerships, the population density around the dealership groups, how spread out these dealership groups are, and what you find is, depending on how you set up the strategy, the efficiency of the campaigns can vary greatly. If you apply a one-size-fits-all approach, your efficiency drops down quite a bit and you end up with a lot of impressions hitting a lot of people,” Kohn continues. “If you apply a strategy for Los Angeles in the exact same manner in Chicago, you end up hitting all of the wrong people.”

The second study looks at how far urban, rural, and suburban shoppers are willing to go to shop at a big-box retailer.

Here too, the answer is complicated, and understanding the connection between population density and store location distance are part of the equation that geo-data can understand.

In its third case study, Thinknear examines Amazon’s acquisition of Whole Foods to see how the deal has affected minimum threshold levels for radius targeting to cover 70 percent of each store’s consumers since the sale of the grocery retailer.

“The case studies highlighted in our latest Index show some very smart ways that marketers can leverage location data to more effectively spread their message,” Kohn says. “These examples are just the tip of the iceberg as location data continues to influence various aspects of the marketing cycle.”

Locating Optimism

In sorting through the effectiveness of geo-data that so many providers present to brands, location technology platforms like Placed, PlaceIQ, Foursquare, Factual, NinthDecimal, Ubermedia, Blis Global, GroundTruth, and Thinknear all have various mechanisms and methods for filtering out “noise” of bid stream data, another reason the industry often notes that the issue of accuracy is not as bad is it can seem. Ultimately, it’s up to agencies to compare which ones get it the most right.

“We’re looking at the overall programmatic mobile industry, so it’s not just taking your data that we buy on our campaign,” Kohn says. “When we first did the report four years ago, our thought was everybody’s jumping on this. The Mobile Marketing Association is involved. Public servers are getting involved. This problem will sort itself out within 18 months. That was optimistic.”

So is Kohn optimistic about the direction of where the industry is headed?

“I’m optimistic about it in the sense that four years ago, when there were only a handful of players in the space and location was something that was new in coming onto the scene, you certainly had cases where people ran into scale issues of, ‘Gosh, do I have enough high-quality inventory out there to buy with good-quality data?’ That was an issue.

“That issue today is almost non-existent in terms of, “Is there enough inventory to buy?” And so I’m very optimistic on that front. But I think given the growth and amount of inventory, it’s that much more important for brands to be working with partners or have tools that enable them to find those increasing numbers of needles in the increasing numbers of haystacks. There are plenty of those in there. The hard part really is finding it.”

About The Author
David Kaplan David Kaplan @davidakaplan

A New York City-based journalist for over 20 years, David Kaplan is managing editor of A former editor and reporter at AdExchanger, paidContent, Adweek and MediaPost.