Excerpt from Yield, Ari Paparo's upcoming book

Unless you’ve been living under a rock (or blocking me on X) you probably know I’ve got a book coming out. Well, its finally shipping this week! I’m also doing a reading in NYC on Thursday the 7th (RSVP here).

In advance of getting your hands on a copy, here’s an excerpt from Chapter 13, which explores Google’s reaction to the growth of header bidding.

Jonathan Bellack forwarded the seminal AdExchanger article introducing the world to header bidding to his broader team, and in this one email chain we can see the mixture of dismissal, negativity, arrogance, and internal conflict that would characterize Google’s coming approach to this rapidly spreading innovation.

Aparna Pappu, VP of engineering, offered a quick take: “The fight for the tag on the page continues, and suboptimal setups continue.” This comment reflects Google’s obsession with keeping their tag on the publisher page as a way of avoiding disintermediation. While header bidding participants certainly liked having their tags on the page, it was not the primary impetus for the movement; it was much more about getting access to inventory.

Drew Bradstock, a product manager on Bellack’s team, dismissed the revenue effects and offered that “pubs are gambling and can’t actually see the opportunity costs of doing this.” Drew seemed to have taken the point of view that latency or other problems with header bidding have hidden costs that publishers are ignoring. While this might be accurate, it also reinforced the parochial attitude toward publishers that existed inside Google at this time, that publishers did not know what was good for them.

Nunzio Thron, an engineer on the team, got to the heart of the matter that others were trying to avoid: “Publishers are doing it because their ad server doesn’t do what they want (e.g. getting inventory to compete on both Rubicon and AdX). It’s very rational for publishers to want to integrate all possible demand sources, and if our answer is ‘you don’t need those demand sources’ rather than ‘here’s how to integrate those demand sources smoothly with high yield and low latency’ I believe we’ll lose publishers.”

Tellingly, the more customer-facing team members diagnosed the problem accurately as well. Michelle Sarlo Dauwalter, a specialist in the support organization, offered bluntly, “If we are committed to competition, and we believe competition drives revenue, we should allow for all sources of demand to compete fairly, in real-time (or as close to real-time) as possible. While I can help our Sales team build a story around latency and lost impressions . . . the fact remains that if we allowed for real-time competition across all demand sources, we wouldn’t even need to have that conversation!”

And finally, bringing to the fore some of the optimistic and visionary spirit that originally pervaded Google but that seems to have dried out, Dauwalter concluded, “If all demand sources competed fairly, we could start stripping away the waterfalls, would immediately gain more access to inventory, make more money for publishers, increase ad viewability and engagement, and ultimately make the internet a better place.”

“If” indeed.

The dislike for header bidding at Google was not entirely commercial and self-interested. For the engineering-led culture of the company to see publishers running their advertising business on what was perceived as hacked-up JavaScript just felt deeply wrong.

The biggest actual problem with header bidding was increased latency. Every time a publisher page would load, a bunch of different vendors would load their own code, each of whom would then make a call to their servers, wait for a response, and then finally submit their bid into the ad server. In a worst-case scenario, this process could add up to a full second or two to the page load. From the earliest days of the development of Google Search, founders Larry Page and Sergey Brin were obsessed with optimizing the speed of their pages. Until very recently, Google searches still included the number of milliseconds it took to deliver results.

The latency argument had validity but was also self-interested. One publisher executive complained that the Google ad server was actually the slowest part of his entire stack. “Given the latency of [the ad server], I don’t think they have a leg to stand on talking about header bidding latency,” he added. Another example of latency hypocrisy related to the ad network bidding into AdX. While other bidders, including DSPs, were given strict time-outs on every auction, when the ad network bid into the exchange, it would get extremely loose time-out restrictions, a fact that was not disclosed.

Header bidding was also wreaking havoc on Google’s ad server infrastructure. Instead of a single ad for each source of demand, you might have thousands or tens of thousands of ads from each source to allow for the use of key values that reflected each ten-cent increment of targeting. Even for the vaunted Google engineering systems, the software was simply not built to sustain this use case. Early adopters of header bidding would often run into limits on the number of “line items” allowed in the ad server and had to struggle with their account management teams to get the limits raised.

Jonathan Bellack raised concerns about billing and reconciliation. There was no way, he argued, for a publisher to audit whether the bids coming into the header (and thus winning the auction) matched the payment the publisher got at the end of the month. There was no real evidence this was happening, but it was a real concern at the time.

In addition to Rubicon and others being disadvantaged by header bidding, FUD—fear, uncertainty, and doubt—was also coming from Google. While everyone agreed that latency was the main problem with the technology, Google execs would frequently bring up ad quality and fraud as a drawback, even though there was no evidence to back that up. Payam Shodjai, a product leader for the Google DSP, suggested “a carefully crafted marketing campaign . . . primarily targeted to buyers and couples spam + [header bidding] into a single story.”

In November 2015, at a conference focused on advertising operations, Jonathan Bellack got on stage with Tom Shields from AppNexus to informally debate header bidding. Tom was a legend in advertising, arguably one of the inventors of the first ad server (even before DoubleClick), and had recently sold his company, YieldX, to AppNexus. It was a cordial and gentlemanly conversation among erstwhile rivals, with Tom emphasizing the revenue to be made and Jonathan bringing up the antiheader talking points of latency, bad ads, and financial reconciliation. The video remains on YouTube but cuts off before audience questions. According to witnesses, the most interesting part was after the debate, when Joey Trotz from Turner confronted Bellack. Joey was a longtime and loyal customer of the DoubleClick ad server and regularly attended client events to provide feedback on the products and generally build bridges with the team. His question was simple, asking more or less, “This is all nice, but your customers are making more money from this and need your help. What are you going to do to help us?”

Jonathan is at heart a publisher guy, and by all accounts he really cared about the well-being of his customers. While he worked under Neal in the same organization as Brad and the ad network team, as well as the newly ascendant DSP team, he had to look out for his customers to feel like he was doing his job. There was no doubt that he was still constrained and conflicted, as the rejection of prebid from the IAB illustrated, but he had to do something. He reframed the problem as not how to kill header bidding but how to make something better. His customers were doing something that made them more money, they weren’t going to stop using, and they wanted his product to support. Whether it was distasteful, or against the interests of another Google team, was secondary.

Google could not just ignore this problem either. With header bidding, suddenly a big part of the value of an ad server was being commoditized. If publishers were making many of their most important ad decisions in JavaScript on the webpage itself, it was not that big a leap to imagine putting the rest of their ad decisions through that mechanism and dropping the ad server entirely. The ad server was the linchpin of Google’s publisher strategy, so any threat to its dominance was taken very seriously. If you squinted just hard enough, you could imagine a competitor like Facebook or Amazon teaming up with a header bidding solution to peel publisher customers off Google.

Yield, How Google Bought, Built and Bullied its Way to Advertising Dominance is available tomorrow, August 5th, on Amazon or wherever you get your books. Learn more about Yield at aripaparo.com.