Understanding ad serving
Go back to basics to appreciate the possibilities of header bidding
April 11, 2016
A header bidding setup lets publishers that use traditional ad servers transact more effectively with buyers that use digital advertising technologies those ad servers were not originally built to accommodate.
In order to understand what header bidding is and how to take advantage of the technology, it’s important to first understand how ad serving itself evolved. Here’s a greatly simplified history of ad serving that explains the need for header bidding in today’s landscape. If you understand ad serving, feel free to skip to the next section. If you don’t quite, read on to learn how this foundational technology makes header bidding a powerful tool in publishers’ technology arsenal.
Phase 1: Ad server as delivery and decisioning technology
Let’s rewind the clock about twenty years. The web was a much simpler place then, and so was digital advertising. Individual web publishers began monetizing their sites by placing ads on them. In its simplest form, an ad could be a static picture with a link attached to it, but there quickly became a need to create a separate technology to handle digital advertising for publishers.
One reason for this was because publishers wanted to sell ad slots to more than one advertiser. In order to handle this, publishers needed a way to rotate several advertisers into a single slot. The solution: the publisher would put a piece of code on their site that called another technology where multiple ad creatives for different brands and campaigns had been implemented. This was the ad server. The ad server would respond with the respective ads in a rotation. Thus, a publisher could make deals with more than one advertiser to run in a single slot.
But there was a catch: all advertising deals weren’t the same, so an even rotation wouldn’t do. Publishers needed a way to prioritize ads so their ad servers could more ably balance the multitude of objectives and impression volumes that a publisher’s multiple ad partners wanted to achieve.
To deal with this, ad servers introduced a ranking system whereby publishers could assign an overall priority to different advertisers and campaigns, and could also assign goals to those campaigns. In addition to being a repository and delivery system for advertising creative, the ad server’s primary role was now also decisioning: figuring out what ads to run, on which publisher pages, for which users, and when, based on those publisher-assigned priorities and goals for different advertisers and campaigns.
Phase 2: Third-party buying sources hook into ad servers
Then the digital advertising landscape became more complicated. In addition to the deals they made directly with brands, publishers began selling the leftover ad impressions they hadn’t been able to sell directly to brands to ad networks at a deep discount. Networks got access to the publisher’s inventory through a single line item in the ad server, the same way a direct-sold advertiser would. Because the ad network model expressly addressed remnant inventory at first, it made sense to slot networks in at the lowest rung of the priority ranking in the ad server, because these advertisers were the lowest paying and lowest priority by definition.
Then things got more interesting. Instead of selling publishers’ leftover inventory at a static rate, technologies came along that created an auction marketplace for those leftover impressions, and allowed brands to bid for them. To determine what bids brands should submit for a given impression, bidding technologies evaluated the attributes of individual impressions in real-time and submitted bids based on the impressions’ predicted value to the brand. Because pricing was now dynamic, an impression determined to be highly valuable might command a price on par with the prices that a publisher’s premium, direct-sold advertisers were paying.
Phase 3: Real-time opportunity suppressed by traditional ad serving
This exposed one key constraint of the traditional ad server setup that header bidding was created to address. Ad servers were set up to prioritize and serve ads based on known, static prices and campaign goals. They were not built to handle dynamic pricing, real-time decisioning, or single “pipes” to demand sources containing a widely diverse set of potential buyers.
This meant that even though digital advertising as a whole evolved to incorporate dynamic pricing, publishers were stuck having to prioritize campaigns in their ad servers based on a single estimated static price. To account for RTB sources, they used their best guess of what the average price of all demand coming from dynamic sources looked like, even though the prices that created those averages could be highly stratified.
For buyers, that meant being mostly locked out of seeing higher-value inventory. For sellers, it meant denying impression opportunities to brands that were willing to pay high prices for individual impressions.
Phase 4: Denying parity to buyers that used new advertising technologies to find premium audiences
But however much traditional ad server setups depress potential revenue coming from RTB sources, another, even more consequential shift happened in digital advertising that represents an even more significant upside of implementing header bidding.
That shift is that brands and advertising technologies realized that the same technologies they’d built to transact unsold inventory could also be used to discover, setup, and transact private marketplace campaigns. Now the types of high-dollar, custom campaigns that were originally implemented by the publisher directly in their ad server were being implemented behind RTB line items that were still relegated to the bottom rungs of the ad server’s priority list, unable to compete with “traditionally” direct-sold campaigns even though their pricing and prioritization was on par.
A smart new way to transact direct-sold campaigns was being constrained by a legacy technology.
This is the key to understanding header bidding: it isn’t the solution our industry would have created to address digital ads the way they are bought and sold today. It’s a way to make the legacy ad serving construct we’ve inherited address the advertising automation paradigm that has evolved in our industry.