Who’d be an analytics company? It has become the new grudge purchase, right up there with security and encryption – it is only noticed and of any use, when something goes wrong.
The thing about OTT video is that not only are things going wrong all the time – buffering, jitter, black screens, App stacked clients, weakening portable batteries drained, inappropriate screen sizes, malformed manifests, CDN delay – but there is often little you can do about it at the time.
Operators are looking for a new kind of AI intelligence which is expert at not only identifying which clients are currently unable to play their video, but which are able to adopt strategies to minimize the problem or eliminate it, right now.
All of which is why analytics firms tend to be small or small departments of larger businesses. The market is supposed to grow, according an ABI Report out this Summer, to over $3.7 billion, from $1.8 billion now, but analytics under that report’s definitions was a broad church indeed – taking in audience engagement measures, audience measurement and network optimization – most of which are really not analytics at all.
We accept that Network Optimization can be part of the picture, because the problem may be in the configuration of the network, although network Optimization tools alone should be a bigger market than that in 5G for instance.
But how engaged people are is really just an advertising tool, and audience measurement is… well it’s audience measurement – you only do it to convince advertisers to spend money because you have such a good audience, so it’s another advertising or monetization tool. Those, because of their relationship to money tend to yield higher prices. Everyone else in analytics is a grudge purchase.
To us video analytics, as we are over-fond of saying, is quite simple, and breaks into 3 parts. What condition was the video when I sent it, what did the network deliver, and what did the device receive? There is of course a fourth part, which is what was the device able to do with it once it arrived?
We interviewed some four specialists in analytics at IBC – Agama, a pure play video analytics firm that has been around for 13 years, a Swedish firm which is managing the transition from hardware probes to software services and OTT; Tektronix which is part of Fortive (previously Danaher), a $17 billion business, and embedded within its $5.8 billion diagnostics division which services everything from automotive to pathology; Verimatrix, a newcomer to analytics in that its Verspective Intelligence center was floated as an idea in just April 2015, but for which no clients have yet been announced although it has been continuously developed, and Accenture Digital Video.
Funnily enough our chat with Accenture, which acquired Irish testing and analytics firm S3, did not involve analytics at all. The company seems to feel it has analytics worked out – instead it is obsessed with seeing what the next big thing is in video, and we agreed between is that it is voice (see other story on Google and Amazon this week, and our IBC special last week).
Others at the show we have talked to in the past, notably Nice People at Work, Conviva and Cedexis, which has come out of CDN analytics, but may be poised to partner its way to a full analytics stack.
In June Conviva secured a $40 million funding round, so there are investors out there who are chasing video analytics as a fundable subject issue and we think there are multiple drivers. First off, ABI may not be that far off in suggesting that the core video analytics business will more than double over the coming 5 years.
But secondly we would make a couple of key additional points. Broadcast video remains, for the time being, the main activity of pay TV providers and free to air publishers. To some extent we should look at Free to air and TV Everywhere and SVoD as requiring very different types of analytics.
Free to air, for the most part, wants to sell adverts on the back of proving that the advert was viewed all the way through. So to some extent it could care less if the video is poor quality – it’s the ads they care about. TV Everywhere continues to be a free extra that pay TV companies “give away” while the main paid for product is really the broadcast stream. It is only SVoD (and live sport and other specialist forms of SVoD), that cannot live without knowing that its video quality is up to scratch, and which cares about every stream. SVoD for the most part falls into 3 main categories globally, the US skinny bundles, Netflix style generic VoD services and pay TV providers who are just entering SVoD, usually as a defense against Netflix.
The US fresh love of skinny bundles, another specialist form of SVoD, is what is driving the need for US investors to be onboard with analytics. Hence the money that’s available to Conviva. Netflix and most large Pay TV operators who are moving out of TV Everywhere into SVoD, are developing their own analytics systems or are the single contracts keeping smaller analytics players alive. It is only when smaller operators have to join them with an SVoD offering that the requirement for analytics will become truly urgent.
We see a key inflection point in 2 to 3 years, when suddenly instead of SVoD being responsible for 10% of revenues, and TV Everywhere responsible for 20% of viewing, those numbers double or treble, and every pay TV operator will be desperate to secure its video chain. Prices will rise, revenues will quadruple, for the right players. But who are they?
When total revenues from OTT delivery start to approach 30% of an operators revenue, then all pay TV players will want chapter and verse on every stream they send, and will pay more per household or device to ensure a) that it is delivered in good shape and b) when it isn’t some automated system will fix it and report back. Hence Conviva’s mission to invest in AI, primarily to provide that type of service to US skinny bundle suppliers, but it is also focused on the larger players in Europe.
For the myriad pure play SVoD services in Europe and Asia, this is already true, but for the most part they have little money, although they are growing and will see have more.
So what are the steps? Funnily enough both Agama and Tektronix echoed one another. Ian Valentine of Tektronix, Business Director, Video Products at Tektronix, told us, “It has to be end to end and the trend is to use OTT for delivery. We have systems which measure the quality of the video as it goes into a head end, or into a just in time packager and then measures it again on the way out and compares them.
“We answer questions like what does your video look like, and how long does it take to start playing. Our clients need to be sure the video is not stuck in black and we have to do this on 100s of channel simultaneously. For ABR video we have to check that the manifest is not malformed, that the video is not blocky, and that the right profile is being delivered to the right device.” He did give us the impression that the Tektronix value add was to focus on picture quality, always acting as a client and comparing the video to what it knows was sent earlier and also where there was no source to check it against, checking that there are no tiling errors in the MPEG. “We can do a check post CDN too,” he added.
But to us this is really just the technological equivalent of driving round in a car with a TV and seeing if broadcast reception is good. Agama is at the other end of the process and has this week cut a deal with the dominant independent media player provider, VisualOn.
VisualOn has around 70% of the non-native media player market and as such can help players like Agama get a spy in the device. As OTT video develops we think that’s all-important. VisualOn collects multiple type of information
It has device information which can tell if the battery is about to run out, if there is enough memory to run this app, where is it geographically, and what is the correct screen size. The player software itself can tracking the operational state of the player, the video startup time as well as buffering, and can check the application logic. It also collects customer id, information on the title and asset number being played. It is uniquely placed to report back to any analytics console and Agama really needs this information, because rivals like Conviva already have it.
It’s all about getting a piece of software inside the App, or better still in every device, beneath the app, which can report on the apps.
Cedexis turned this problem on its head when it wanted to build something for reporting on the state of CDNs. It offered all the heavyweight (read video) websites, free analytics if they would put code onto their website. This is turn reports on their attempts to deliver data across CDNs, which makes the business a load balancer for CDNs. It really wants the opposite side of the coin and needs a piece of software on-device. It could do worse than to go to VisualOn too.
It is this ability to have something on-device that takes an analytics player the next step from telling an operator what happened afterwards, to telling it what is happening now. It is a short step from there to telling it what will happen next and reconfiguring its network or CDN, to prevent it.
Verimatrix got itself into this game by being able to upgrade its content protection software on the device with the Mirimon software it bought from Genius Digital, which gives similar capabilities to the VisualOn toolset. Of course it needs the permission of the operator it is working with to do this, but it has all the necessary ingredients lined up in a row. Its initial Verspective Intelligence Center focused on calls to the DRM server, and interpreting that into which content had been called for. It was never really in the game until it could come up with a plan to show what condition that content arrived at in the client. Its next step needs to be to intercept problems and fix them on the fly and provide a predictive element.
Agama previously had most of the same ingredients of Tektronix – being able to compare copies of the video by behaving as a client, sitting either side of the origin server, and the other side of the CDN, but until it did that deal with VisualOn, it had little insight into how video plays on the device. It may be 13 years old, but it has just got itself into the game.
Agama had always had the basic table stakes, which is a console that shows which part of the video chain has developed problems. But without being able to drive data to the cloud for huge populations of devices and tell how they are performing, it was still flying blind.
It now offers IP metrics, such as packet loss and jitter, in the network to validate the stream health and it has always had a view on the ABR stack, the availability of the service, manifest correctness, segment availability, as well as measures on the video quality. But again that’s as it leaves the Origin server. It even has syntax analysis on formats such as HEVC and MPEG. But it needed that final piece on the client itself.
Today it is also talking about AI and is currently being used by Sky and TalkTalk in the UK, KPN in the Netherlands, Vodafone, Telia, Telenor and A1 in Austria.
It is now offering a QoS prediction which is really an aggregation across multiple device types. This creates a QoE score on its control console. When that score is not high enough the customer care team can drill down into service paraments and device types and find the root cause of the problem. But right now the telco has to fix it, the process is not automated.
Per Unell, Business Development at Agama told us, “In a telco every week there are changes in software and this can screw up services. So we support the operators at the customer care level, and can do things like compare video feeds with those going to the neighbors and see if there is anything wrong and if what you are getting is a customer unique problem or a system one.”
“Engineering calls are 70% no fault found,” he said, “And it is so re-assuring if the customer care consultant can drill down and say ‘We can see you have a problem’ “
Just as soon as one of these analytics players can develop the strategy that then proceeds to put it right, or can predict in advance that this was going to happen, that’s when they can pick up perhaps half of that $2 billion of business that is waiting to be grabbed over the coming 5 years.