September 21, 2020
By Nadine Krefetz Contributing Editor, Streaming Media Editorial Staff
Short Cuts

Best Practices for Server-Side Ad Insertion

See more videos from Streaming Media East Connect on the Streaming Media YouTube channel.

Learn more about SSAI at Streaming Media West 2020.

Read the complete transcript of this clip:

Geir Magnusson: What are some of the best server-side ad insertion practices? It's a complicated set of interactions between multiple systems and it's important that it not go wrong, because--at least in the case of linear--once your ad avails are gone, they're gone. You'll never see that inventory again.

Michael Bouchard: We've run into the issue because we have so many different audiences on different devices and different advertisers that want to advertise on our networks that we can't just take an SSAI approach. We actually support both client- and server-side depending on the situation. And we actually have some streams that do both interchangeably, which gets very interesting. And if you tie that into something live like a news broadcast ... You've all seen live news broadcasts just appear. They don't listen for the ad pod to finish, so it's truly live. And so our ads, whether they're server-side or client-side, have to adapt to that. And that was a big challenge in the early days, working with Darren and his group.

Nadine Krefetz: Were you building it? Were you buying it?

Michael Bouchard: We thought we could buy it because ... Everyone's showing ads. You watch the Super Bowl, you see ads, but it turns out that's not truly live. They understand their pod size. The other thing that surprised me was finding out that--for certain advertisers and certain programmatic ads--because of the analytics we get back, they pay a higher CPM for client-side. So we're all proud to be rolling out this new platform that's all server-side ads, and we have to turn half of it off because of the revenue. Very unexpected. And we're still running some of those today because we got a higher CPM. So it really depends.

Nadine Krefetz: The money talks. Magnus, how about you? What are your comments on this area?

Magnus Svensson: I would usually say that for server-side ad insertion preparation is key. You need to prepare the content coming in with markers and everything like that. You need to have slates and fillers prepared in case of backup. You need properly encoded ads coming in so you don't have a glitch in quality for the ads themselves. If you're going live, I would say, be prepared for high peaks. Everybody's asking for ads at the same time. So using preload, and header bidding, and these type of techniques to flatten the curve is pretty popular. Be prepared, because if it's a live event, a lot of viewers will press ads at the same time. It will be high peaks for the system. So with preparation, you should do what gets quality in the end.

Darren Lepke: It's very true that server-side ad insertion and the ad ecosystem online are very, very complicated. That leads to a couple different challenges. I think when we talk about CPMs, today server-side ad insertion supports VAST advertising and VAST protocols; client-side still supports VPAID and some of the interactivity there--that's where you see some of the higher CPMs. But I think with some of the advancements and protocols around VAST 4.0, some of the open measurement standards, some of the things that are coming online there, I think we're starting to see a little bit of rebalancing of those CPMs, because you're starting to get more analytics from the server side.

So that's certainly something to think about. Ultimately, you have to understand the balance of inventory in your programmatic ad system, or in a presold ad system. If you only have paid ads and not VAST ads, that presents a challenge. So you have to look to your ad partner there as well.

Now, the second part of complexity you want to think about is, in the television ecosystem, it's very controlled in terms of how ad breaks are scheduled, the exact durations of ads inside of an ad pod on the internet. And especially when you're going after open exchanges or programmatic exchanges, if you request 30 seconds of ads, you don't always get exactly 30 seconds back. Sometimes you get 28, sometimes you get 32, and you have to have a flexible system that can say, "I'm going to allow my ad break to go 30 seconds," and maybe my live stream is a little bit further behind live. So you can kind of flex your ad pods. You may make a decision to say, "I'm going to chop the last ad. Right when we get to the 30-second break, I'm going to cut out of the ad no matter what, and go back to live." And that may affect your completion rates and your revenue. And then you may also decide, "If I get 32 seconds worth of ads, I'm just not going to show the last ad. Or I'm only going to do 20 seconds of ad and then show 10 seconds of slate."

So, it's a little more complicated. I think that that will become a little less complicated over time as the industry meets some of the requirements for SSAI. Ultimately, it's about planning and understanding what's most important to you. Are you optimizing for revenue? Are you optimizing for latency? Are you optimizing for production? These these are all the things you have to consider when you're looking at best practices for SSAI.