In this interview, Magali Quentel-Reme, CEO of Opti Digital, dives deep into the supply-path optimisation process. She examines whether publishers should favour a particular path, as well as looking at how they can protect their website's performance.

What is the environmental impact of the supply-path optimisation (SPO) process? 

The ad selection process is one of the most polluting processes in online advertising, as many experts have already pointed it out. The programmatic ad selection process requires a large number of ad calls, involving multiple parties such as DSPs and SSPs, which means a heavy impact on servers. 

One common mistake we often see is publishers connecting too many partners in header bidding. The usual setup includes both client-side with Prebid J.S. and server-side, with Amazon TAM, Open Bidding and sometimes a third-party server-side wrapper based on Prebid Server. For one ad single placement, there could be up to 40 connections, for example. This is not sustainable, nor good for revenue. 

In efforts to become more sustainable and avoid throttling, many publishers are reducing their number of resellers and only keeping direct connections. However, just because the path is direct, does not necessarily mean it’s more sustainable or profitable. 

So, which paths should publishers favour? 

Largely, the path a publisher should take is dependent on their strategy. But what we observe is that very few publishers actually harness traffic shaping, which involves reducing the number of ad calls sent to demand partners based on various criteria. Some intermediates – like us – are specialised in designing and implementing traffic shaping models, and selecting and curating inventory to send only specific requests to the SSPs that are most likely to respond. This goes far beyond sustainability. Too many partners and too heavy a stack have a negative impact on website performance, in particular on the loading speed of site elements, including ads.

Today, audiences are very mobile-heavy which means that SPO is essential to optimise user experience, ad loading and ad revenue: even if competition is lower, selecting a refined selection of partners improves both user experience and monetisation. By minimising server calls, publishers reduce site latency which may lead to higher SEO ranking. Additionally, loading ads faster improves viewability rate, leading to higher CPMs and overall ad revenue performance. Eventually, with less bidding partners sharing the cake, every SSP still connected to the stack gets more profits and this increased profitability gives the possibility to lower or totally waive any potential throttling the SSP used to apply to protect their income and the publisher will receive more programmatic demand.

What role do algorithms and machine learning play in the ad selection process in programmatic advertising? 

They play a powerful role. From the buy-side, powerful models throttle the ad requests that are not profitable for the demand-side platforms. Likewise, there are powerful models on the supply-side. We suggest publishers arm themselves with the right tools to take back control over the requests they send. At Opti Digital, we design and implement traffic shaping models. We use similarly powerful tech to segment the audience automatically according to multiple signals in real time and automatically filter their requests to limit any throttling applied by the buy-side. A different strategy is defined per segment. 

We also harness AI to dynamically change the information we send in the bidstream, depending on the demand predicted for the browser/location of end users, viewability, or time/weekday of the ad being displayed etc. Dynamic flooring information is sent in the bidstream to all of the partners connected through Google, Prebid and Amazon. This process is fully automated, powered by machine learning models trained over time on every publisher’s audience and revenue data. Due to its nature, it’s constantly changing. 

How can publishers protect their website's performance and loading time?

Generally, we are seeing that the average ad stack on the market is very heavy, up to around 550 kilobytes, running about 200 ad calls at page load. Publishers using these heavy ad stacks can experience negative effects such as decreased performance and increased loading times. 

 At Opti Digital, we’ve reengineered our ad stack into the lightest ad management tool, reducing its size to 300 kilobytes with only 15 calls. As a result, ads load in half the time on mobile devices. This is making a great difference. We believe this will become the new benchmark. By choosing a  lighter ad solution, publishers may rank much higher on Google SERPs and obtain more audience from organic search. Therefore, digital media publishers keeping a heavy adstack will be penalised in terms of audience. 

Another extremely important component to consider is the header bidding auctions setup. Server-side auctions offer a considerable advantage over client-side auctions, since calls are moved server-side, enabling faster data processing and content delivery, improving overall website performance. However, we often see publishers enabling server-side header bidding by connecting a server-side wrapper to their usual client-side one. This Prebid J.S. code still negatively affects the web performance and it’s not SEO friendly. It may be good for their revenue, but it’s certainly not good for their audience. 

At Opti Digital, auctions are run server-side only on all ATF ad units that must be loaded as fast as possible. On all the other ad units, when speed is not critical, a hybrid header bidding strategy seems the most profitable one that enables partners that still heavily rely on third-party cookies and are not server-side compatible yet to keep bidding.