Key Developments and Contributions of Programmatic Advertising

Until the late 2010s, real-time bidding (RTB) purchases operated in a "waterfall" fashion. When a user opened a publisher's webpage, the supply-side platform (SSP) offered ad space ready for monetization for sale on an ad exchange. The auctions worked as follows: the demand-side platform (DSP) with the highest historical cost per mille (CPM) was invited to bid first. If interested, it placed its bid and automatically won the auction. If not interested, the SSP offered the sale to DSP number 2, whose historical CPM was higher than that of DSP number 3, and so on until one of these DSPs purchased the ad space. Second-price bidding also dominated these exchanges. According to this principle, if DSP number 1 had a bid of, for example, 10 euros per mille, DSP number 2 had a bid of 8 euros, and DSP number 1 wished to buy the ad space, it ultimately only paid 8.01 euros. This encouraged bluffing, as the DSP aiming to purchase ad space was tempted to place a higher bid to position themselves better in the SSP's historical sales data while being assured of winning the auction and paying only one cent more than the next DSP in that list.

This mode of operation had some undesirable effects on publishers. Firstly, if they wanted to maximize their chances of monetizing their ad space, they had to connect their ad server to multiple SSPs simultaneously (sometimes two or three, but often about a hundred), which overloaded the ad server. Additionally, publishers were never certain they were getting the best price for a given ad space. After all, there was no guarantee that a DSP wouldn't place a higher bid than all the DSPs higher up in the bidding list. Moreover, there was no guarantee that an advertiser couldn't generate more impressions than another, despite being served by a DSP better placed in that list. In concrete terms, DSP number 1 offered a CPM of 10 euros, won the auction, paid only 8.01 euros, and their advertiser partner generated only 10,000 impressions, while DSP number 2 might have been willing to pay 11 euros and their advertiser, perhaps, generated ten times more impressions.

Header bidding is the solution found to address the pitfalls of the aforementioned waterfall auctions. Barely ten years ago, it re-empowered publishers who often received less than 15% of actual ad spend from advertisers due to the significant share taken by technology partners (ad server, DSP, DMP, visibility, ad fraud tools, ad exchange, etc.). It facilitates real-time auctions among multiple sources of demand for ad space. Publishers now directly call their partners' offers (ad exchanges, SSPs, ad networks) from their webpage instead of going through the ad server. The role of the ad server is still to choose the highest offer, but it now compares the best offers from each source of demand against each other, whereas previously it was confined to managing and accepting the relatively best bids that each SSP reported to it on a case-by-case basis in a waterfall model.

This new competition could have naturally driven up prices. However, for publishers to handle the technical capabilities of installing tags from different sources of demand – around thirty SSPs simultaneously – directly on their webpage, header bidding became a breath of fresh air. 

A solution rarely pleases everyone at the same time; what benefits publishers may not be advantageous for the various technology partners involved in the programmatic chain, whether on the demand or supply side. SSPs, in particular, now that they are called directly from the publisher's webpage, have complete visibility over their inventory. They also handle requests from thousands of pages simultaneously, significantly expanding the total inventory (and thus audience) they must try to fulfill. This increase requires more computing power and thus higher costs. Additionally, their ability to win auctions inevitably drops as they are now in direct competition with all the other sources of demand that the publisher's page directly solicits at the same time as them. To regain competitiveness, SSPs opt for a first-price auction model, which means that when they win, a DSP or an ad exchange, for example, pays the price at which they bid (rather than the price at which the second bidder was bidding plus one cent in a waterfall model). As a result of their desire to be more competitive, SSPs also need to offer a greater supply of ads for auction. Thus, SSPs start complementing the supply of their DSPs by connecting to more sources of demand, such as ad networks.

This solution found by SSPs to stay afloat becomes a thorn in the side of DSPs, who are solicited by these SSPs and also by the ad networks that these same SSPs now directly call. DSPs run the risk of placing bids against themselves. The cost of their operation increases due to the significantly higher number of requests to process, and their competitiveness drops.

As DSPs receive repeated requests from the same sources, they start filtering their partners (SSPs). In the waterfall model, they only saw the inventory of the SSPs to which they were connected, their added value for an advertiser residing in their ability to connect to as many SSPs as possible to increase their chances of reaching their target audience. Now that all SSPs have complete visibility over the inventories available on a page, DSPs no longer need to be connected to all these SSPs simultaneously. Consequently, they start filtering them. They systematically attempt to eliminate duplicate paths, focus on bids with a higher relevance/probability of winning ratio, and may even decide to exclude SSPs that no longer use second-price auctions or do not provide unique inventory. A well-executed SPO allows a DSP to reduce its connections to supply-side partners by a factor of ten.

Perceiving the unintended effects of these new filters on their activities, SSPs counterattack and refocus on direct purchases, neglecting auctions in which DSPs participate. This is referred to as Demand Path Optimization (DPO). Is there no end to this back-and-forth race for tricks? Yes and no. Overall, the ecosystem evolves based on the temporary (commercial or technical) advantage that one partner may have over another before the balance swings back in the other direction. Any loophole in the ecosystem, presenting an opportunity to provide more visibility, and transparency, or facilitate integration, is exploited by ad tech companies at a certain cost. This is how meta-SSPs and meta-DSPs (aggregators) were born in the past, along with other solutions repositioning header bidding on the ad server side rather than the publisher's webpage. It's how new technological building blocks will emerge in the value chain, as opposed to the finite number of ad spaces in a print newspaper or magazine, the inventory of online advertising is limitless.

But... the stakes of Programmatic also continue to evolve. In 2023, it's no longer just about optimizing publishers' yield and advertisers' ad performance. The carbon footprint of campaigns must be considered and reduced. In this regard, DK has developed a Responsible Digital Guide (Online Advertising Option) aimed at different profiles involved in campaign development, particularly ad ops. Our experts are available to our clients at any level of programmatic understanding to refine their programmatic buying strategies in the most environmentally responsible manner.