“Honey, I’ve been lying
Honey, I’ve been jiving
And I am not signifying”
—The Rolling Stones, “I’m Not Signifying”
In digital advertising, we take for granted the usefulness of data. It’s hard not to — data has been intrinsic to media since the days of Arthur Nielsen, and it now swirls around us in seeming superabundance. Every mouse click, purchase, video view, song listen, Starbucks visit can be tracked, targeted in real time, and saved for future use.
If it is possible to fetishize an abstraction, then Big Data is our fetish object du jour, the foundation of the premise that the more data, the better. Most media discussions don’t even linger on the topic, but quickly move on to questions of more tactical salience: methods of data ingestion, data origins (1st-party? DMP? Which DMP?), etc.
But it pays to remember that Big Data is usually dumb data (as I’m not the first to point out), and that more is not necessarily better. The key to doing great marketing with a particular type of data is to a) quickly understand what it is useful for and what not; b) figure out what else it can most usefully be combined with.
A case in point: location analytics. In the early days of mobile programmatic advertising, many of us in the space (myself included) became obsessed with a brand new type of data being offered up in bid streams: device location. It turned out that between, between 5% and 20% of mobile bid requests contained actual latitude / longitude data.
This wasn’t your garden-variety IP address mappingm as happens with Zip codes on desktop browsers, but actual device lat / lon, sometimes written out to 6 decimal points! Here is some simple math to illustrate how ludicrously precise that seemed: if you divide the Earth’s circumference (40,000 km) by 360 degrees, then one degree equals 111 km. Taken to the sixth decimal point, that is 0.11 meters, or about 4 inches. 4 inches!
So location data was not only available in scale, it was precise — and therefore perfect for digital marketing. And many smart people wondered whether we could connect the dots between the places a person visited and the things they were likely to buy.
For example, Person A goes to Equinox gym in the morning, Le Pain Quotidian for lunch, and Whole Foods in the evening. Is that person more likely to buy a new luxury car than Person B, who goes to Equinox gym → Burger King → and Pathmark?
From a data science perspective, figuring this out entails analyzing a ton of passive, continuously-updated data, identifying meaningful patterns within that data, and testing the resulting hypotheses with actual campaigns. In other words, a lot of work. Unless you’re making Palantir-sized revenue, was boiling this particular data ocean worth the slightly higher CPMs?
Not really, as it turns out.
Taken in isolation, an audience’s patterns of location visitation reveal few secrets that the brand advertiser couldn’t already get from existing demo research. After all, mapping a particular place on the map to a particular demographic is nothing new; out-of-home and terrestrial radio advertising have been doing it for decades, and census tract data is public and mostly free.
The beauty of location targeting lies deeper down the funnel. Mobile location can be considered a new type of cookie which -- like a cookie – has an effective expiration date and is most useful when connected to other types of data. For example:
This last example is the area of greatest interest to me in 2015, because it encompasses a huge swath of media -- programmatic buying, digital out-of-home, addressable TV, offline retail data, and more nascent technologies like iBeacons and NFC (which, despite low adoption in the US so far, will see a big uptick over the next two years, thanks partly to Apple).Connecting these dots between offline and online, physical and digital, isn’t simply about Big Data. It’s about Smart Data – and having the discipline to know the difference.