Advertising on Facebook and Google is a foregone conclusion for most brands. Hotel booking site Hotels.com does, of course. But it was also looking for an alternative.
“There is bias in the walled gardens,” said Helen Cameron-Heslop, senior manager of ecommerce analytics at Expedia’s Hotels.com brand.
On Wednesday, Hotels.com announced a year-long partnership with cross-device company Tapad after a multiple month bake-off against two other vendors.
The nature of that walled-garden bias is inherent and not dependent on size. Google might have billions of users, but the only cross-device connections it can legitimately claim take place within the Android ecosystem. An iPhone user might use Google, but not be logged into Chrome.
By the same token, most of Facebook’s usage is on mobile, leading to the inverse: far fewer users logged in on desktop.
Regardless, all of that raw data is tucked away behind the garden walls anyway.
“As an analyst, I find it difficult to work with data that I cannot validate myself,” Cameron-Heslop said.
And that can create a somewhat fraught dynamic.
“Facebook and Google are closed ecosystems with the goal of selling more ads to marketers and, in the process, increase the price of their product,” said Tapad CEO and founder Are Traasdahl. “A marketer’s goal is to drive the highest ROI, and we also see that the advanced marketers need full transparency, unification of identity across all media sources and, lastly, they need complete access to cross-device data so they can tie back to their own data.”
Which is why Cameron-Heslop was attracted to the concept of a more open relationship with a probabilistic cross-device vendor. But there were concerns.
“Our business is our data and we don’t want to hand it out willy-nilly to people,” she said. “We also knew that by taking this outside of just our own logged-in user data, we would probably have to make a compromise on accuracy if we wanted scale, or compromise scale if we wanted accuracy, especially considering how much we think about privacy.”
Cameron-Heslop declined to name the other two vendors tested, but there are only so many cross-device vendors in the space, which makes it fairly likely that Drawbridge and perhaps Crosswise were in the consideration set. [Updated September 15, 2016: Crosswise reached out to say it was not one of the companies that participated in the bake-off. September 16, 2016: Qualia got in touch to say the same.]
After three months, each vendor returned a device graph created from a subset of Hotels.com’s users. Hotels.com then validated the graph against its overall data set looking for two types of matches – connections seen by both Hotels.com and the vendor and incremental matches the vendor could provide beyond Hotels.com’s own data.
Hotels.com carefully evaluated those incremental matches because while each additional match might seem like added gravy, “we also needed to strike a balance between that and making sure the matches weren’t … overzealous, that someone wasn’t getting a little creative with the matching process,” Cameron-Heslop said.
In other words, the matches needed statistical integrity.
Going forward, Hotels.com will work with Tapad to help quantify the success of its marketing efforts and put numbers behind gut feelings.
Cross-device attribution is particularly important in the travel vertical, where the consideration period is long and consumers often spend a lot of time researching trips on mobile before booking on desktop later.
“Our anonymized user data is fractured – we can see the conversions happening and what device they happen on, but there are a lot of journeys that don’t seem to convert and make it appear like we’re wasting our marketing budget when we know that’s not the case,” Cameron-Heslop said.
Mobile is clearly a critical part of the mix, even if the final conversion usually takes place elsewhere. But Hotels.com wanted to rely on more than intuition to inform its ad spend.
“By its very nature, mobile is more challenging in terms of the identification process because people move around between IP addresses, and that makes it more difficult to link anonymous users and behaviors together,” she said. “That’s what we’re using Tapad for – to understand those non-converting journeys and see the value that they actually bring to us.”
