How Do Containers Enable Custom Decisioning In Programmatic Advertising?

In terms of containerization, you know, one of the questions that comes up is, "Well, why now?" And I think there are a couple of different forces at play that are making this, uh, you know, really compelling answer at this time. The first one is that historically we've just optimized for how do we make our products better? We've looked at this buy-side, sell-side division, and the sell-side has worked on making their tools better, the buy-side has worked on making their tools better, and we've been sending data back and forth. And we're in a world now where there is so much data, so many ad requests, that the efficiency of that is breaking down. And that, combined with the fact that containers have gained popularity, it's a technology that's been around for probably a decade, and we've been using it for 6 to 7 years, but it lets you portably deploy logic, uh, bespoke models, you know, your IP, and, uh, deploy it in the cloud or or on prem if you want, uh, but it really enables the scalable, um, and kind of composable building of different technology solutions. But as the volume of traffic goes up, what you see is that that old model of let's send everything across and make these decisions here and these decisions here starts to break down. And so, uh, as ad tech always does, people come in and they say, "Well, I think if we innovate, we can package some of the functionality that's worked in one place or another, and we'll get better performance, better results. It'll be more sustainable if we execute that in the right place." And containers make that feasible. Uh, you don't have to worry about security, you don't have to worry about your IP, but what you have is a tool for packaging a discrete piece of logic or optimization and running it where it makes the most sense. So with containers, it gives you the ability to do things where they make the most sense. I can now package a piece of logic and move it to the place that is most effective to run that logic. So, for example, if I want to run a performance model, I I can do it now at the source in the SSP where I see all the supply, all the signals, and I can shape the the traffic that gets sent downstream to me, which gives me much better performance and the op- opportunity to optimize what I'm doing. Uh historically, that's something I could only do on the DSP side, and I had to do it, you know, in the way that each DSP enabled me to do it. But now I can do it in a generalized way at at the source, which gives me phenomenal results. We've been cloud-native for 7 years at this point, and our whole platform is built on containers. So, we've been using containers for quite a while, and we understand how to do it well. And not only that, we also understand how to scale things up and down. And because we've been managing containers for a number of years, we have a lot of expertise there. That allows us to manage our partners' containers well. We We understand what is it that they need when they're running a piece of their application in someone else's infrastructure. And then there's also the efficiency and the scalability piece. We have access to just a tremendous amount of compute, and on top of that, we're constantly optimizing what compute do we use to run our containers because we want the most efficient options possible. And our partners get the benefit of all of that, the scalability, uh the management, the efficiency, and because it's integrated directly with our platform, it's incredibly uh performant from a latency standpoint. We are, you know, sub-millisecond for a number of things, and that leads to great performance for our customers. Custom decisioning gives me the option of packaging logic uh or models or whatever it is that I want to use to influence the outcome that happens. I'm no longer limited to the tools built into a given platform. I can do whatever I want.