"Data-Driven Thinking" is written by members of the media community and contains fresh ideas on the digital revolution in media.
Today’s column is written by Tom Phillips, CEO at Dstillery.
We hear a lot about device match rates these days. Maybe too much. It started a few years ago, when the advertising industry decided it needed to find all those great retargeting candidates on their mobile devices.
You see, consumers still booked travel, researched car purchases, signed up for cell phone plans, credit cards and insurance, and bought all kinds of stuff on their laptops and in browsers, contributing to the rich array of cookie-enabled intelligence that has been a boon to marketers of all stripes.
But for the ad tech industry – and its clients among digital ad agencies – there was an alarming trend. Consumers still “converted” the way they always had, but they were communicating, searching, browsing and playing games in a whole new way: on their smartphones and tablets, divorced from all that conversion activity. This threatened the virtuous circle of advertising campaigns tying to measurable customer acquisition.
The answer appeared to be device matching. It’s a set of predictive technologies for associating the growing number of devices used interchangeably by consumers. The technique originated as a way to find that plum retargeting candidate as she moved from laptop to smartphone and tablet and back again.
The approach had some bonus advantages. That long-simmering issue of the single consumer using multiple browsers suddenly had a nice resolution. It even addressed cookie-clearing, as that browser with the cleared cookies could be found again because it continued to be associated with a particular smartphone or tablet.
A new golden age, right? Kinda. People, including press and agencies, started asking about the match rate for these predictive technologies. And those inventive companies who developed the matching technologies found themselves in a bind. (Disclosure: Dstillery bought one of them last year.) How to measure the match rate? There was no reliable truth set to determine the exact match rate.
Then the number 80% was floated. Taken literally, it meant that for every 10 browsers, eight could be associated with a specific mobile device. And for every 10 smartphones, eight could be associated with a specific browser. Really? How does anyone know? Who’s doing that study?
The secret to success in applying predictive technologies may be in knowing which questions to ask, which can be answered and where to look beyond the obvious to determine real intent. The accuracy of device-matching predictions is still interesting, but the business imperative is finding relevant audiences for marketers. We should put the evaluation of every industry metric through that filter: Are we getting to the ultimate objective of improving audience relevance, and at scale?
Here are some questions that are germane to the issue of device-matching accuracy. The confidence level of the answers you receive will make them meaningful:
- Do the device owners observed at airports exhibit a materially higher proclivity to visit websites on their laptops that relate to travel and travelers, or to working in the travel industry?
- Do the people who overindex for visits to a given cruise line website tend to reside in places where that cruise line has port departures?
- Do visitors to the physical locations of a big-box retailer have higher-than-average conversion rates at that retailer’s website?
- Do the people whose devices are observed at university campuses exhibit a higher-than-average interest in the websites of no-fee credit cards?
With the right approach, the expected correlations will hold true and the audiences will make sense – logically, and for marketers.
That doesn’t mean there isn’t any room for surprises. Take that first question above, for instance. It’s a valid assumption that people who go to airports a lot would tend to look like business travelers in their Web browsing behavior. But what was discovered is that those airport regulars look a lot more like airline workers. That makes sense, but it obviously changes the approach for that audience set.
The bottom line: The match rate is fundamentally a valid concept. But if the goal is to establish meaningful audiences for marketers, let’s ask the questions that get at that objective and for which we can develop hypotheses and then test results. Let’s focus on matching audiences to brands rather than devices to devices.