Sky Media Executives Discuss Keys to Successes and Eye Challenges Ahead
Sky Media has leveraged proprietary technology and vendor platforms such as Imagine Communications’ Landmark Sales system to deliver targeted advertising to eleven million subscribing households. Now the U.K. MVPD is extending such capabilities across its multiple service outlets to all screens, utilizing metadata, big data analytics and algorithms that enable highly automated approaches to multi-platform ad campaigns.
Attendees at this year’s IBC Show in Amsterdam had a chance to get an inside look at these endeavors when Imagine hosted a discussion about Sky’s strategies led by Sarah Foss, vice president of product management for advertising management systems at Imagine, and featuring Jeff Eales, director of systems and development at Sky Media and Tim Taylor, Sky Media’s head of advertising and content supply technology. An edited transcript taken from the video of their meeting follows.Sarah Foss, VP, product management, advertising management systems, Imagine Communications – What are the systems you call ad management at Sky? What’s that foundation that has billions in revenues going through it?
Jeff Eales, director, systems and development, Sky Media – Not surprisingly, the first one is Landmark [Sales System], which you guys supply for us. There used to be just an air time sales system and our broadcast system. But now it’s about 20 times more complicated.So we have an integration system that sits around Landmark that handles all the interface files that we use. We transact about 200 channels a day. So that means there’s about six to seven thousand files that come in and out of the system every day. And they have to be fully managed, and that’s managed on another platform.
We have a data warehouse platform, a staging area to collect all the data. Obviously we have all the broadcast systems downstream, but we don’t necessary need to count those. We have a reporting system that tells us pretty much live what the hell is going on.
For AdSmart we’ve added another four extra systems. We now have a segmentation engine that tells us the information about every household in the Sky universe. We call it the flexible allocation engine, but it’s neither flexible and it doesn’t allocate anything. It’s not an engine. But that basically is the think box that works out and tells every set-top box what to do when it gets an ad spot activation code. So where there used to be one ten years ago, there’s now at least ten.Tim Taylor, head of advertising and content supply technology, Sky Media – The other thing to think about is it’s a big ecosystem of ad sales systems. If you think about the process of ad sales from campaign management all the way through to delivery through to measurement, because we’re doing lots of new generation advertising, we end up with an ecosystem to make all this stuff work.
We kind of say, don’t worry about your broadcast systems. But there is stuff done in our broadcast systems where we have logic, which is required to be able to serve the advertising properly.
Foss – What are the next-generation business models or systems that you’ve had to expand to that you want that broadcast logic to be part of?
Taylor – Let’s take AdSmart as an example.
Foss – There are probably a lot of people who don’t know AdSmart.
Eales – We now push [targeted ads] down to the 11 million boxes we have in the U.K. and Ireland. We push ads targeted to the household down to every box. We key that using an activation code in the live stream so that we can substitute a better ad or a more targeted ad into the household than the one that’s in the linear stream.
Foss – So it’s your proprietary targeting system.
Taylor – Because Sky has the platform, from a technology point of view you can actually view the set-top box as an ad server. Essentially what we’re [setting up the ad campaign] in our ad management systems. The flexible allocation engine is kind of creating some rules about options for advertising. The set-top box has logic on it which defines what ads to serve at what particular point in time. And so the business logic for the ad serving spans all the way from the Landmark system to the set-top box system. And it’s all got to work properly to be monetized properly.
Eales – And what we’ve never done before is we now take data off three million set-top boxes a day. We use a constant panel of half a million set-top boxes to manage all the AdSmart measurement systems. So we know pretty precisely the number of people that have seen any particular campaign at any time.
Taylor – The interesting thing about AdSmart is it looks quite easy, but under the covers it’s quite complicated. Prior to the set-top box actually substituting an ad, it’s got to know campaign details. It’s got to have the ad copy down to it. It’s got to be given a warning to do the substitution.
So getting all that working seamlessly is quite a challenge – to make it an experience where the viewer doesn’t actually notice because they had the ad targeted appropriately to the audience segment they’re in.
Foss – In terms of the business models that drove AdSmart, what are some of those next-generation advertising models that you think have actually either fueled it or you were able to lead the industry by having that type of platform?
Eales – The fundamental thing is it changes the business model from being able to put an ad to the whole of the U.K. and Ireland to being able to push it down to any segment of about 5,000 households that have something in common. So we’re talking now to large small businesses, even plumbers, local car dealers, house agents that never have been on TV before that often spend their money with local press or radio.
Now they’re looking at TV as a really, really good option, because you can make a pretty good ad without that much money now. Provided that we help them make that ad, we can pretty much get anybody on TV.
And the other end of that is Maserati. We’re talking to really high-end car manufacturers that would never be on TV before. And now you’re able to put on TV a £250,000 car ad to the right people. And it works.
Taylor – What it has opened up is more audience, the ability to take our customers and segment the audience to people who can afford a car or people who want a plumber in their local area. From a technology point we need to manage the segmentation of all those audiences.
Foss – Okay. So we’re covering the segmentation. We’re talking about audience now instead of just ratings. We’re talking about targeting and hyper-local targeting. What has it done for you two with something like programmatic, or the automation of the buy and sell process? Are you seeing that type of targeting be part of a programmatic pitch? And, if so, how are ad systems helping you with something like programmatic?
Eales – On mobile programmatic is absolutely part of the chain, because it’s centrally ad served or controlled by the ad servers. So you can do programmatic.
But it depends on what you mean by programmatic. Programmatic in the U.K., I think, means something very different from programmatic in the U.S. With programmatic in the U.S. people are talking about order management optimization and getting agencies to put an order in that’s then set from order management systems down to ad zone sales systems to give them a campaign, and that campaign then gets back.
We’ve been doing that since 2003. If I sound smart, I mean to be, because we’ve been doing it. Ninety percent of all the business we trade in Landmark is put there effectively by agencies, by agency buyers.
We have an industry system in the U.K. called Carrier, and all of the major broadcasters have shareholdings in that platform. We put it live in 2003. Basically it’s not only controlling sales orders; it’s controlling copy rotation as well. We take those copy rotations and campaign orders, put them directly in Landmark, and most of the sales people do not type in orders anymore.
If that part is programmatic, we do that. The piece we’re all now working on is whether or not you can actually deliver that campaign programmatically and target it to that particular person when they’re watching mobile. We’re nearly there.
We’re working with a number of companies trying to work out how we would take some inventory and place it out into SSPs (supply-side platforms) and DSPs (demand-side platforms). We would hope to have something running pretty quickly.
Taylor – Some of the concepts around programmatic we’re talking about, around audiences and managing audiences and making them visible and available, and some of the stuff around automation and the automation of booking, those are probably two of the big concepts we’re trying to work with at the moment.
The other concept we haven’t talked about too much is, if you look at the ad inventory – we talk about linear ad sales; there’s lots of on demand – we’ve now got content across a multitude of different platforms. So for us one of the challenges is how do we manage that in the best possible way, optimizing it from the advertiser’s perspective and optimizing it from Sky’s perspective, from a revenue perspective. Those are some of the challenges we’re grappling with at the moment on the programmatic stuff.
Eales – Simply, where does the ad get stored? We have three places the ads are stored – whether it’s going out linear; whether it’s going to be ingested into the set-top box, or whether it goes out in mobile. Currently they are three different stores.
But where the ad is on programmatic is another issue. There is no way currently we would key a commercial to be put into the center of Game of Thrones when you’re watching live on SkyGo [Sky’s OTT service] where the ad is sitting in Mumbai somewhere. That cannot happen.
We have very high Ofcom regulations. Obviously we want the viewing experience to be perfect. We cannot rely on not getting the ad or the ad not being compliant. So that has to go into a safe store. So it’s not programmatic fatigue for video. But it’s a hundred miles away from programmatic for display.
Taylor – The funny thing is, with my other hat on about content supply, we have challenges about creating content in a multitude of different transcode formats to display on devices. It’s exactly the same challenge in advertising.
The other challenge we have on content supply is there’s a whole bunch of stuff going around personalization. There’s a big theme about creating content for you. It’s exactly analogous to audiences and trying to create a personalized ad experience for audiences. So the challenge about three different ad stores and 23 different transcode versions is what we’re grappling with to try and do it in an automated, efficient way and to make sure at the end of the day that everything is compliant, because if we’re not compliant then we run into a lot of trouble.
Foss – Let’s follow that theme for a second. Are you seeing the convergence of that personalization of content and the targeting of advertising creating convergence of systems within Sky? Is that something you’re starting to see?
Taylor – Not yet. They’re being driven by two different drivers. From an ad sales perspective it’s around audiences for revenue. From content personalization it’s around driving consumption for subscribers for Sky.
Foss – When we talk about all the systems you have and the different next-generation business models, how are the systems driving top-line growth for operational efficiencies? It sounds like you have lots of different initiatives going. But are there ones that you’ve really seen a material difference in terms of either your top line or your bottom line?
Eales – Well, here’s the thing. In Sky Media we have some of the highest quality TV or required TV by buyers in the Premier [League] football and, let’s say, some of the not-quite-so-interesting, what we call the long tail. For instance, we still have in the BARB (Broadcasters Audience Research Board) measurement world one of the best measurement systems in the world, [but it] still provides that 40 percent of our spots get nothing.
So how do you work between an environment where you might get nothing to an environment where you get two to three million viewers watching football that is immensely and hugely required? And the answer is, what we’ve been able to do is come up with an agreeable algorithm with the buyers that enables us to put that altogether. If we didn’t we’d just be here for years.
So we do it all on a ratings buy. That ratings buy effectively behind the scenes values everything and enables us to pull it altogether. We wouldn’t be able to expand the way we have in the last five or six years if we hadn’t come up with an agreement with the agencies as to how to buy different sorts of TV at the same time and value it at the price that it should be to the target audience that they require and at the time it’s watched.
By doing that we can at least move forward and automate around the big buys. There are very, very few of our spots in the system that either have a rate card or have value against a single spot. The future cannot be about selling single spots for a rate. Not all of them.
It just can’t be that way, particularly if you’re putting in impressions and you’re trying to look to see how this new area of viewing, where people are viewing on SkyGo, how that inventory can be merged with the linear inventory. That’s the challenge for certainly the next 18 months. We might be kind of getting there. I’m not sure I answered your question.
Foss – It was interesting, though.
Taylor – I’m not sure I’m going to answer your question either. But one of the challenges I want to raise is we’re collecting lots of data. AdSmart we talked about; data is coming back from set-top boxes about what’s happening. Where we’re serving on demand, we’re seeing lots of data come back.
One of the pieces we’re doing is around the analytics of that and the decision making which that drives to [determine] what sort of campaigns to serve where. I just think the optimization challenge is going to be really difficult going forward, because we’ll have lots of audiences, lots of platforms, lots of different services to serve ads to people.
How do we decide what ads to serve where? If you get a request for a certain audience and a certain volume of eyeballs for that audience, where do we do it that’s best for Sky but still meets all the advertisers’ needs?
Foss – It’s interesting, because, although the question was about systems, you both answered what are enablers that are helping you solve the problem of growth across the multi-platform environment, whether it’s optimization or it’s analytics. In that model, it almost seems like next-generation advertising systems are less about the distribution of what spot for what targeted ad goes where so much as what’s all the data and the optimization technology that allows a sales person to create an effective campaign and differentiate Sky from other media companies. Is that a fair statement?
Eales – Absolutely. But on the other end of it, you have to work out whether we measure stuff or whether we use things for a currency. It just so happens that normal TV people meter panels that we have all around the world measure a program, and as you measure a program, you can use that measurement to measure the ads.
So you can use that to measure the ad, but the basic algorithms behind both give you the same number, and it’s stable. And even if you don’t agree with it, it’s still stable, which is the strength of TV around the country.
The problem now is potentially that we can do census-based impressions ratings, because you can collect all the viewing data off the mobile universe or you can get to very large samples. Interestingly enough, if you get census data, what do I want? Do I want to know everything about everybody who logs on? Or do I want to know about a subset of that data so I can get that quickly to make a decision about a trend?
I have to tell you, different people have very different views on it. If I was able to make that decision from all the information all the time and make that very good decision straight away, then I think I probably want all of it. But we can’t disagree there’s a tradeoff a little bit at the moment. Can you actually collect 60 million set-top box data in two or three hours to provide next morning’s revised advertising schedules? If you can, great. But I’m not sure you’d want to do that.
Taylor – You talked about systems. I kind of ended up talking about capabilities and saying there are capabilities and there are ways to exploit those capabilities. From a systems point of view one of the things from my perspective is just the rate of change. Purely from a system point of view, with the number of platforms, the rate of innovation in broadcast, the new services, we’ve got to be able to innovate quickly. And that means whatever you’ve got in terms of systems, we need to be able to adapt quickly to keep up with rate of change and to exploit it to the best of our abilities.
Eales – If we have another way to be able to see great content, it means we have another stream to be able to sell. What seems to happen is we get ourselves so that we can sell it, and we create another currency in doing so. So we currently are up to about ten, ten different currencies that we trade with agencies. If you look back five or six years, there were probably two.
Taylor – The other example coming up is VR (virtual reality). If VR takes off, we have to create a new ad form for VR. What kind of audience watches VR, and how do we target stuff best to them to be able to exploit that audience? That’s what I mean about the rate of change. As innovation happens we need to think about how our advertising systems are adapting to be able to do the advertising in the first place, but to do it in the best possible way.
Foss – So, two follow-up questions: Is technology leading the sales team right now? And how much is the sales team coming back and saying, well, wait, if you can give me that, Tim, do I get this data, too? How much is that conversation happening?
There used to be this firewall between the sales team and we never wanted them to know anything about how we were actually making things happen in the back room, or God forbid, we’d sell something we didn’t deliver.
Eales – We talk to the sales team probably more than they would like us to talk to them. All of them, fortunately, don’t have any restrictions about how long it might take us to do something, which is good and bad, because they think that rather than taking six months, it should take six hours. Why is it so difficult, because aren’t you great technologist?
So we’re definitely being asked to do things very much more complex, very much more quickly, but in an environment that’s one of the most highly regulated environments in the world. That’s a question. Technology is clearly leading our ability to sell extra stuff. But interestingly enough, if that technology is coming from the digital end, what’s coming with the digital end is some very un-grownup business processes.
Foss – What does that mean?
Eales – It means you’ll take any creative of any size because you need the money. If you’re in digital, that’s kind of what you do. On TV we’ll only sell you a ten [second slot], a 20, a 30, a 40, a 50 or a 60. So you can’t make an 11-second ad, or if you do, we won’t transmit it because we’ll ask you to take that down to ten seconds.
There are things on the TV side where there are some very mature processes, because we’ve been around 45 plus years. Those mature business processes are meeting the World Wide West, where we have to do everything really quickly. We don’t really care if it works. And we want to put those together.
That actually is an enormous amount of fun, but we do need to find out how we do it right, how we do it quickly and how we can make money. That’s our problem at the moment.
Taylor – The other challenge from a technology point of view, everything is so interconnected and complicated. Go back to the AdSmart example. What we were doing on linear TV was substituting a new ad over an existing linear TV ad. And the thing we really, really couldn’t do was mess up our linear TV.
So we’re having to build that [new] technology on top of something which has got to work. And as we continue to innovate we’re continuing this complexity. Managing that complexity is one of my worries – making sure everything keeps working while you add the next thing on. That’s one bit about the technology.
The other thing – you asked a question about technology or business leading. I think we’re kind of getting into new territory, certainly with the programmatic stuff in video. We’re having to [ask], how is this going to work from a system point of view? How are you going to choose ads? How are you actually going to serve ads?
But the big point [is], how is this business going to sell in the best way and how are we going to present it to agencies? [With] these sorts of innovation we have to work quite closely together. We have to deliver a solution that works over all.
Foss – You almost just identified what keeps you up at night – with complexity. Is there anything else that keeps you up at night when you think about next-generation ad management?
Taylor – It’s complexity and keeping the service going. AdSmart is one of the things. [With] the logic running from our ad sales system to the set-top boxes, if we have a problem, where on earth is that problem? We don’t diagnose it quickly, and there are lots of components that are part of it.
That’s one thing that keeps me awake. The other thing that keeps me awake is this speed of change thing. How do we keep up with all this? How do we do programmatic really quickly?
Foss – How about you, Jeff? What keeps you up? Compliance?
Eales – Absolutely. If you describe the situation on one side, I’m completely with Tim. I think Confucius said when the pace of change outstrips the pace of learn, you get a gap. There’s no question we’re getting a bit of a gap. No question that we have to learn as quickly as the pace of change. That’s one of the things that will keep me somewhat awake at night.
But in an environment where we have more advertising, more places to put it, greater ability to target, many more compliance and regulation issues, if somebody says, do you think TV will become less regulated or more regulated, I think most of the room would think it will become more regulated.
So that means I just need metadata coming out of our ears to enable us to control that. And one of the things that’s very good about that, and one of the things we tried and tested was that, effectively, we’ve taken all the control data sitting in the good old fashioned workhorse called Landmark in the air time sales system, all of that compliance data, and pushed that into the eight million set-top boxes that are now working as mini ad servers.
We weren’t sure whether that would work, but it actually does. We put an awful lot of checks into making sure it does. It’s a really good stepping stone to say we definitely can do that. Now we can go to the next stage.
Taylor – The compliance thing is kind of exciting. We’ve got the Web and all the displays and programmatic stuff and the innovation that comes with that, and you’ve got the TV where we’re trying to create new services and platforms but still adhere to the rules of the game in terms of compliance, in terms of keeping advertisers happy. You’re right about that challenge, because our converging world means we need to manage video advert in a different way than you do with multi-display adverts on the Web.
Foss – I love this. You’ve given me the 4 C’s. We need to worry about control, compliance, complexity and business continuance.
Eales – And change.
Foss – And change. The 5 Cs of next-generation ad management. As a parting thing for you both, what do you think is the next big advertising technology that Sky is going to have to tackle in five years?
Taylor – Maybe not over five years, but programmatic is our current challenge. But you go beyond programmatic and you ask, how do we really exploit audiences across platforms and across our services quickly and easily? That’s the challenge. We’ll get down to real top personalization, real targeting of adverts, to be able to do that properly in a way which is automated.
Eales – We have a lot already. The biggest challenge for us and the community in the next five years is oddly pretty dull. It’s standardization – making sure we call the same stuff the same thing. If we can do that it will stop miscommunication.
I mean, we know the U.S. and the U.K are separated by a common language. You call it sidewalk; I call it a pavement. You go up in an elevator; I go up in a lift. We do need in metadata terms to try to make sure that everything is named as correctly as we can. It might sound a bit dull in a very exciting landscape, but those are the sorts of things we know work back in the linear world, and if we can begin to bring them in, we can shoot for the stars.
Foss – Thank you so much Tim and Jeff for sharing your insights on next-generation advertising management.