A few weeks ago, Google made an announcement that surprised the advertising technology community. Google, long known as the 800 pound gorilla of the ad tech world, said that after years of keeping their DFP ad server closed to outside real-time buyers, they will open it. The move has the potential to solve a fundamental problem that’s plagued automated digital ad buying (also known as “programmatic” buying) for the last 20 years. But the announcement was also met with some distrust, and it raises important strategic questions for many of the players in the digital advertising system today.
A Fundamental Problem In Programmatic Ad Buying
Programmatic ad buying is the process by which a website’s ads are purchased by advertisers automatically in real time. Theoretically, it’s pretty simple: a user clicks to a site, indicating that an ad slot is about to be served. The site offers the ad slot up to an auction (known as an “exchange”), where advertisers use technology to assess it and offer a bid. The highest bidder ends up showing their ad on the site, all in a matter of milliseconds. In the big picture, programmatic buying is a hugely important advance, allowing no-cost, instant and intelligent transactions between media companies and advertisers. (As a matter of comparison, under the old paradigm “Sales and Marketing” routinely makes up 35% of the total expenses of running a media company.)
So though programmatic buying depends on an auction, the fundamental problem has been that there’s never been a real auction Or, rather, there are so many different auctions that the purpose is defeated — the best price often doesn’t win. The reason for this is that technology websites use to serve ads – the ad server – wasn’t developed with programmatic ad buying in mind. Ad servers use simple pre-set rules to pick ads from a set of pre-selected set. These are linear decisions, like “serve Ad A vs. serve Ad B”, rather than dynamic ones, like “serve whichever of A or B offers the highest price.” When programmatic, auction-based buying came around, many different auctions sprang up, and websites were stuck choosing between them blindly. Their ad servers could only decide to send the ad to Auction A or Auction B based on pre-set rules. They did not know in advance what the highest bidder in each auction would pay. They just had to guess, and guessing wrong meant leaving money on the table.
A Single, Unified Auction For Ads (…?)
What publishers need is obvious: a single, unified auction for their ads. With all the bids being processed together, publishers are sure to get the best price. And it seems like Google is creating just that. From their announcement:
<blockquote> Exchange bidding in Dynamic Allocation will allow publishers to invite trusted third-party exchanges and SSPs to submit real-time prices using industry-standard RTB calls. These prices will be considered along with bids from the DoubleClick Ad Exchange and the publisher’s reservation campaigns to pick the highest-paying ad.</blockquote>
Industry observers have been calling for exactly this since 2010, but rather than jubilation, there’s been significant caution and mistrust around the announcement. The reason for that requires some explanation about how, for a long time, Google was part of the problem.
Dynamic Allocation And The Google Advantage
Google owns the most popular ad server, Doubleclick for Publishers (DFP). They also own the largest ad exchange (AdX). For a long time, they benefited from this by bundling the two together, calling it Dynamic Allocation This allowed real-time, auction-based bids from AdX into DFP, but not bids from any other exchanges. This helped publishers a little bit, because some programmatic competition is better than none (though certainly worse than full competition). It certainly made AdX more appealing to buyers, because they got to cherry-pick the best impressions without full competition. It was therefore also certainly bad for anyone running a competing ad exchange.
These competing exchanges fought back against Google’s advantage, offering a workaround known as “header bidding” that allowed them to compete more fairly. Combined, this hack had the effect of unifying the separate auctions anyway, outside of Google’s control. This led to better results for competing exchanges, and better yields for publishers. The downside is that they were hacky, requiring many pieces of third-party code in the header of the site which slowed down load times and degraded user experience.
Facebook, The Open Web, And Google’s End Game
Some see Google’s opening of DFP to everyone’s real-time bids as a capitulation to header bidding, and that’s definitely part of the story. Google must have known that Dynamic Allocation would be a short-term advantage, and that a robust, open marketplace is the future of programmatic advertising, because that’s most efficient for the buyers and the sellers. Of course, they probably also knew that running that meta-exchange would not be a high-flying business. The interesting question is whether Dynamic Allocation was simply a prelude to a grander strategy, or a short-term advantage they lost and will try to regain another way.
The answer will come as Google’s plans unfurl. If, as some have speculated, Google seeks to charge a “tax” of any sort on the outside bids going into their meta-auction in DFP, it’s likely they are pursuing a short-term plan to try to defeat their rival exchanges and consolidate more and more of the market into the Google ad technology stack. If this is the case, header bidding will still prosper, and the industry will not have made much progress.
However, it’s likely Google sees a bigger picture. If they make it free and frictionless for all the other ad exchanges and ad buyers to participate in a unified auction in DFP, these players will have no reason not to participate. In fact, they will be compelled to by publishers.
This latter strategy would raise prices for publishers by creating a truer, more efficient market. Perhaps even more importantly, it would commoditize access to supply as a strategic advantage. In the old system, each ad exchange (except Google’s) was seeing a unique set of ad impressions. This meant buyers had to be on those exchanges to access that unique supply. Further, exchanges and other supply-side-focused platforms would jockey for that access, and tout it as a unique benefit. Surely, with the move, Google will commoditize itself, but they are also commoditizing an entire segment of the ad tech industry built on supply-side access. This is to their benefit for two reasons. First, Google is also a media seller with a huge amount of their own ad inventory, which they can hold out as a continuing reason to use their AdX exchange. Second, if everyone has the same access to a bulk of web ad impressions, then the differentiator becomes not supply-side access, but buy-side intelligence. And Google, with their trove of data and technology advantage, is well positioned to win that fight. We are already seeing other programmatic platforms bulk up their buy side business in anticipation.
Finally, it’s possible that Google’s strategy has nothing to do with ad tech competitors, and everything to do with Facebook. The new walled gardens, led by Facebook’s Instant Articles, are in a zero-sum competition for users with the open web, where Google dominates. Ad tech’s failings encouraged the Instant Article’s rise, as bloated, low-quality ad creatives pushed users to block ads, and header bidding slowed down sites. Until present, no one had the leverage to control the problem, but Google, running an open, free meta-exchange in DFP, would.
The combined take of the rest of the ad tech industry is small compared to Google’s revenue. However, the signs of poor health on the open web are real: the ad tech issues, the shift from desktop to mobile apps, and the accordant rise of those walled gardens where Google products aren’t used and Google doesn’t make money. Through this lens, and alongside initiatives like AMP, disrupting their own ad tech stack makes perfect sense. It remains to be seen whether this is indeed their strategy, and whether, with all the calcified interests involved, the strategy all work. But for the sake of media publishers, media buyers, and users of the open web, I hope so.
*Disclosure – I work with some of these companies, but we’re small potatoes in their world and they certainly don’t give me inside access to anything. Everything in this post is based on publicly available information and my own theories and strategies.
 I’m using the term “programmatic” to mean “exchange-based”. Technically, any algorithm or software-driven buying is “programmatic”. However, the vast majority of this today is exchange-based, and exchange-based is what matters in the context of Google’s announcement, so please forgive me the move for ease-of-use. I also use “websites” and “publishers” interchangeably throughout.