For Dstillery, Geo-Data Is About More Than Mobile
Consumer profiles across Out Of Home and radio are increasingly being filled out by location analytics, says Lauren Moores.
Over the past two years, geo-data has gone from being the proxy for PC-based cookies on mobile to being one of the primary hubs of cross-device, multiplatform advertising. The problem remains though: how does mobile-based, geographical analytics inform traditional media channels.
Lauren Moores, VP of analytics at marketing tech company Dstillery, has been gauging the expanded role geo-data plays in online and offline advertising for years. She now sees a clear use case for media formats such as radio and Out Of Home, two of the most traditional forms of local, place-based marketing.
GeoMarketing: What is Dstillery’s approach to using geo-data. Is it, do you see it merely as another means to target someone when they’re near a particular business location? Is it for the big topic or do you think it’s just more about insight to understanding?
Lauren Moores: It’s definitely not the former, it’s the latter. The way I look at mobile data — and mobility data, in a larger sense — is the ability to not only help you find new audiences, and figure out who they are, what they’re doing, but also to measure how well your campaigns have performed. It plays in so many realms.
It’s not mobile/smartphone/tablet; it’s more about people leaving physical footprints in a trail of metadata signals that allow us to do so much more than just serve them an ad. Location data allows us to build better models from different audiences. Physical data patterns allow you to marry digital behaviors. You can take those footprints, which are received digitally, combine them with cross-device behavioral data to better understand the consumer path to purchase. This also allows you to create new audience models based off of location and go beyond the norm of using desktop data to reach mobile.
Geo-data allows us to add traditional channels to our existing digital channel sweet spots (desktop, mobile, native). Lately, we’ve been using digital to better understand fixed-location, traditional media such as Out Of Home and radio. We’ve also been using that data to inform TV audiences. It’s taking the patterns of where people have been over time and being able to create behavioral insights. We then use those insights to answer some of the harder questions that we’ve had. It fills in all the blanks.
What’s your sense of the state of geo-data quality? In some ways, it appears geo-data may be getting worse because of more and more people are using mobile networks throughout the day. And in large cities with tall buildings, there’s a lot of interference. So how do you know the geo-data you’re getting is valuable?
When we first started, the most actionable signal that you got from the mobile ad impression was time-stamp and IP address. Then publishers started adding location information. For some publishers when providing data through an SDK, data definition is not always done correctly. It’s done in a way that’s not dynamic enough, or with the wrong triangulation. Or it’s inferred from IP.
In any case, location data went from being 25 percent to being 70-75 percent of overall mobile ad impressions. However, with the growth of location came a lot of dirty signals. At Dstillery we’ve developed ways to analyze the data prior to usage in order to use only those signals we should actually be using for targeting and for analytics.
How do you separate the good location data from the bad?
One way is something we call, “The Manhole Cover,” where we see thousands of lat/long data points where they don’t make sense, like in the middle of the desert. We also track the timing of impressions, and find some mobile devices seem to move across country at the speed of sound. We throw them out, we don’t use them.
We’ve also done work to see if some apps are more likely to provide bad location data than others. You can easily find those that are your most egregious and block those apps from being included in any location modeling. I do think, like anything, once people think a signal is really cool, or a means for more money, they start providing it, but that doesn’t mean it’s good data.