Q&A: The Washington Post’s Programmatic Team Isn’t Afraid of Experimentation

Vp of programmatic advertising Jason Tollestrup dives into the company's partnerships, publishers and proprietary tech

The company prides itself on its 'silo-less' approach. Getty Images

When Jeff Bezos bought The Washington Post in 2013, he made it clear that he needed to bring this “national institution” of news into the digital age. Take the company’s Research, Experimentation and Development (RED) group, for example, which solves some common ad-tech gripes, like viewability and latency, for its advertisers.

“As a company, we’re continually experimenting—and sometimes failing— in order to find that next big thing,” said Jason Tollestrup, the Post’s vp of programmatic advertising. Since 2014, Tollestrup has helmed the company’s programmatic advertising team, morphing it from what he called a “lightly managed” series of ad networks, into a branch of the company that offers up innovative and unique solutions for every advertiser.  He spoke with Adweek about the Post’s experimental ethos, what they look for in their partners, and how they set themselves apart in the marketplace.

The following has been edited for length and clarity.

Adweek: There are so many publishers that all have something unique or buzzy to offer. How do you make yourself stand out from all of that noise?
Jason Tollestrup: One of the tactics that we’ve taken here at The Post is really focusing on seeding user experience—and that also goes for advertisers on the other side. Thanks to the state of programmatic header bidding, the more partners you add, the slower your site becomes, the slower your ads load, and your viewability drops. You might get revenue jumps in the short term, but your revenue shrinks in the long term. So instead of focusing on just loading up our site with as many header bidders or programmatic partners as we can in order to raise revenue, we’ve made increasing the quality of our ads and our ad experience the number one focus.

It sounds like you’re pretty picky about the vendors you choose to work with. What’s the logic behind those decisions?
We evaluate them by a few metrics; the first is whether they can do custom integrations with Zeus, our in-house ad serving script. It’s about one quarter of the size of your typical script, and that really speeds up our ad serving logic, which allows us to put really tight timeouts on our header bidders.

We also ask our partners how often they’re iterating their own product. We don’t want them using a version of an ad product that they’ve created three years ago and haven’t touched since then; we want them continually iterating on making your product faster.  There’s a specific amount of time where we say header bidder partner one, run an ad request—I want that request back in under 200 milliseconds. We only work with partners who are able to come in under our timeouts, and we do custom integrations with every single one of them. If they can’t make that work, we yank ’em out and find a partner that can.

How were you able to get your process so streamlined and cut out some of the messiness that’s inherent to ad tech?
There’s so many parts of the SSP or DSP calling process that are just redundant and can be cut. For example, sometimes an SSP will wait for an ad slot to get designed on the page, which can take an extra 200 milliseconds. So we decided, ‘Hey, let’s just cut that out, build a direct connection to the SSP and define it ahead of time.’ We literally looked at everything that takes time in the auction, and remove what’s unnecessary.

All of our ad loading logic is handled by Zeus as well. The default we work with is called “smart loading, ” which is a little bit more sophisticated than lazy loading. So for example if you’re scrolling too fast that we know the ad isn’t going to load, then we won’t fire the ad call at all, because if that ad never has the chance to be viewed then it’s just a worthless impression, and it drags viewability down. It takes scroll speed into account to figure out when to fire the ad, so that by the time you get there, it’s on-screen and loaded. And if viewability drops, all of that data is piped through our business intelligence tools, so we get those numbers in real time, we figure out why, and we make changes.

It sounds like your role on the programmatic team isn’t just about advertising, but actually dispersed across other teams. It sounds very holistic.
While I can’t speak for other publishers, I don’t think their programmatic people are typically integrated site-wide, but here we are. What I love about the Post is that everybody talks to each other. Engineering is embedded in the newsroom. Advertising is embedded into engineering. We talk to those teams all the time, we’re always running experiments, we’re always trying new things, and everybody has a seat at the table.

I was here before Jeff Bezos bought the company, and back then, everything was very siloed. He was good about changing that culture and getting us to break down those barriers. Honestly, I’m involved with experiments on user experience and stuff like that all the time. And the day that I’m not doing that anymore is the day that I’m gonna get worried, because that means we’re not innovating.


@swodinsky shoshana.wodinsky@adweek.com Shoshana Wodinsky is Adweek's platforms reporter, where she covers the financial and societal impacts of major social networks. She was previously a tech reporter for The Verge and NBC News.
{"taxonomy":"","sortby":"","label":"","shouldShow":""}