Columnist James Green takes a look at the explosive growth of the bidstream, the challenges of header bidding and how automation and data science will shape the future of ad buying.
Over the past two years, there has been an explosion in the bidstream — or for the layman, the number of ad opportunities available for auction in real time has dramatically increased.
This isn’t because there are suddenly more people on the planet, or users are visiting massively more pages, or the number of ads on pages is abruptly increasing (though it may appear that way if you’re not using an ad blocker).
So, why has the bidstream grown so enormously? Header bidding.
As many of you know, header bidding lets publishers offer inventory to multiple ad buyers (simultaneously) before making calls to their ad servers. If no one buys, then they allow the same ad to go for sale on a broader range of ad exchanges.
To make matters worse, many exchanges have started to bid on each other’s inventory and then resell the same ad impressions on their own exchange (for a profit/markup).
In other words, the bidstream has grown because companies are selling the same impression multiple times.
If we take a top-down approach, here’s the number of ad impressions (or in geek speak: “queries per second [QPS]”) you might expect to see on all of the ad exchanges:
But if you look at the ads available across ad exchanges, you’ll see something like 6 million QPS — about 500 percent more! That’s akin to saying the US adult population is bigger than that of India!
Listening to all of this (duplicative) traffic is so expensive that our entire industry’s cost structure has been upturned, causing some very real repercussions.
Before the invention of ad exchanges, the price of ad tech infrastructure would tend to scale alongside campaign volume, so the cost of entry was pretty low. But now, tracking ad exchanges means listening to millions of QPS, which is table stakes for any industry player. That means incredibly expensive tech fees, and a much higher cost of entry.
You may be thinking, what does it matter if you don’t see all the bidstream? Think of it this way: For most clients, only a very small percentage of the population will ever buy what they are selling.
For example, 7 to 8 million cars are sold in the US each year, and Toyota brings home about 15 percent (just over 1 million cars). If you’re only listening to 50 percent of the available audience, you’ll see about half of the Toyota buyers. You’ll need to perform twice as well as competitors that account for all 1 million buyers to get the same result.
As a result, the ad tech business has become a game for the big players. All of the smaller DSPs (demand-side platforms) have gone out of business because they couldn’t see enough of the bidstream to be competitive.
Why has this happened? The industry grew up in favor of buyers, and we’re in a “correctional phase.” Consider the auctions and ad serving (stay with me):
Many theories say second-price auctions are great because they encourage everyone to offer their best price — knowing they’ll only pay the price of the second-highest bidder (and this will cause overall prices at the auctions to be higher).
But these theories don’t apply to ad auctions that occur in 20 milliseconds and are totally blind; we don’t see who else is in the room, or know what they did or will do in the future. So all the second-price auction does is steal money from the publisher.
However, header bidding is mostly done with first-price auctions — you buy the highest you can, and if you win, you pay that price, and it all goes to the publisher. As a result, publishers have seen big gains because of header bidding.
Google poses a major problem. DoubleClick has almost all of the publisher side of the business. It’s not a particularly big business for Google (probably in the low hundreds of millions of dollars). Google could easily build an ad auction business directly into DoubleClick and solve the monstrosity of a problem that we are experiencing.
DoubleClick still works on a priority system where ad campaigns are set to certain priorities (not prices), and these decide which campaigns get served first. The biggest mystery is why Google doesn’t innovate this system. They have another business (Google AdX) worth several billion dollars that would be adversely affected if DoubleClick just allowed the ad server to become the publisher side auction, but that’s just an observation…
As we look into the future, SSPs (supply-side platforms) and ad exchanges will continue to come under pressure from the people integrating auctions (whether open source or otherwise) directly into the publishers’ ad serving and ad inventory systems.
The holy grail is 100 percent transparency, where the ad buyer cuts a payment and the publisher sees 100 percent of it. This has already started to happen and is heartily welcomed.
Sure, the buyer will also pay others who helped them make the decision: DSPs, DMPs and anyone else who can help an ad campaign perform. But increasingly, these will be all be paid separately and directly from the ad buyer or the DSP.
What else can you look forward to? Massive amounts of automation, driven by data science, that will alleviate marketers from executional and time-invasive tasks.
Marketers will be empowered to refocus their energy on key business drivers such as client services, campaign strategy and business development. And marketers and machines alike will be able to concentrate on their areas of expertise to improve marketing performance — a win for man, machine and marketing.
Opinions expressed in this article are those of the guest author and not necessarily MarTech. Staff authors are listed here.