‘We get audience data at virtually no cost’: Confessions of a programmatic ad buyer

What do publishers get wrong about their user data?
It’s an asset that they generate that they should get compensated for. But they’re not. We’re able to cherry-pick their audiences by going around them.

How?
Publishers fork over a lot of demographic and behavioral information about their users to programmatic exchanges. We can buy a few impressions, and layer over some third-party data we purchase from Acxiom or Oracle to build our own audience segments that we would pay publishers a big premium for if they had better control over their data and sold this information to advertisers directly.

What’s an example?
Large consumer-facing publishers have business-to-business readers that they don’t take advantage of. There are executives that read publications like Sports Illustrated and GQ, but publishers don’t typically split the data up that way. Our B2B clients will pay $30 CPMs or more to reach their audience targets. But instead of paying the publisher $30, I build my own lookalike model with the data I glean from exchanges, and then I go buy the inventory for under $10 CPM.

Aren’t you worried that the third-party data is full of errors?
It certainly isn’t foolproof. That’s why we’d pay the publisher more for this data if they sold it directly to us. But the third-party data helps us get such a discount on the audience segments that it pays off in the long run. We test buying across all of our segments before we launch them across big budgeted campaigns. If an audience segment we built isn’t providing return on investment, we will tweak it or scrap it. But enough of them do provide a good ROI that we get a good bang for our buck from the data companies.

Maybe publishers aren’t selling advertisers this data directly because they don’t want to violate users’ privacy.
They might say that because it sounds nice, but that’s a fantasy. Publishers are already selling their readers’ data left and right. But they do it through third parties. If you look into the code on publishers’ webpages, you’ll find that most are littered with 20 or 30 different ad tech companies that are dropping pixels on the site to track users or retarget ads. I think they are exchanging dollars for dimes with this approach.

Why does data leakage remain such a problem for publishers?
The popularization of header bidding has kept it in the spotlight since buyers can now access user data by simply bidding. You’ve got buyers that bid without the intention of winning the auction because they just want to sift through the publisher’s data. Publishers see that their CPMs go up when they adopt header bidding, but the benefit goes to the buyers because we’re able to get publishers’ audience data at virtually no cost.

How can publishers fix this problem?
They should talk to data consultants about how to best present their data on exchanges without giving too much away. They should build their own data management platforms to make sure they’re the ones in charge of their own information. They should invest in hiring data scientists that can help them figure out which audience segments they can get the most money for.

Why don’t they do those things?
Part of the problem is that it is expensive to invest in tech and many publishers are strapped for resources, so they just depend on vendors to fix their problems because that’s easier. The talent that publishers need to get their data in order is in short supply, and programmatic is very complicated so these fixes aren’t as simple as retraining traditional salespeople to memorize ad tech acronyms.

But I think the bigger issue is that many publishers are stuck in the past and infatuated with maintaining their old sales models. If publishers put as much effort into getting their data in order that they do in taking 25-year-old media buyers to lunch, they would figure this stuff out.