Your browser is not supported. Please update it.

22 October 2015

Apple throws down the gauntlet to analytics

Although not directly, Apple CEO Tim Cook has challenged TV analytics to raise its game and enable the same sort of revolution in experience that his company brought with the iPhone in 2007. As it happens that was the year Apple TV also came along, so Cook’s recent utterances bring into focus the starkly contrasting fortunes of the two product lines. One transformed the mobile field and led to Apple become the world’s most valuable company while the other has been little more than a damp squib. When Cook lamented recently that the “television experience has been virtually standing still while innovation has been thriving in the mobile space,” he might as well have been talking about Apple TV.

It is not quite fair to tar the rest of TV with that same brush, for there has been some significant innovation during those 8 years since the iPhone’s arrival. Those years have brought us Netflix and catch up portals such as BBC iPlayer that have, if not transformed, then certainly given a huge boost to the viewing experiences for many.

Another point to remember is that TV is a monster with more than one head. There is little indication of a fundamental decline in lean back viewing of blockbuster movies and major sporting events, but at the same time there is a big shift to mobile devices such as tablets for a lot of secondary viewing and some of the premium material. There is also a growing appetite for engagement through social media and second screens that is not being satisfied fully by broadcasters and operators at present. Cook is targeting both those areas with his promise to transform the whole experience, simply by migrating from channels to apps.

It is tempting to be cynical about Cook’s claim that now finally Apple is about to revolutionise the TV industry, having been just a bit part player so far. The premise of Cook’s claim is that the app industry – or at least the branch of it dominated by Apple – is now ready to deliver. Cook has highlighted the failure of smart TV makers such as Samsung and LG to gain traction as a result of their “very cumbersome” platforms for developers.

We are inclined to side with the sceptics as far as the impact the revamped Apple TV will have, but at least Cook has outlined the direction OTT is heading as a whole. We can note for example US government figures indicating that the number minutes Americans are watching video apps every day rose to 198 in Q2 2015, easily overtaking the typical 168 minutes watching TV. If this trend continues at the same rate, so that much more premium content moves to apps and is streamed rather than broadcast traditionally, pay TV operators will be squeezed unless they too can capitalize on the same trend, as Comcast and TimeWarner in particular are trying to do in the US.

This is where analytics comes in and must itself rise to the challenge of convergence with online, along with fragmentation of content and audiences across multiple platforms. Apple TV is positioning itself at the crossover between broadcast, online and mobile and has been recruiting analytics to the cause from third parties on all sides. For example Adjust has enhanced its mobile ad analytics and attribution tool to support apps for the Apple TV. This will allow app developers to analyse the aggregate behaviour of users targeted by ad campaigns, using Adjust technology embedded in their Apple TV apps.

The aim here is to enable existing online marketing campaigns to target the big screen via Apple TV. For many pay TV operators the challenge is to go the other way and expand into the online and mobile worlds and support their advertising clients across multiple screens.

The analytics challenge facing pay TV as it embraces online can then be summarized in one phrase, Quality of Experience (QoE), but this spans multiple dimensions that have added greatly to the complexity involved. There is more opportunity to give users a superior QoE spanning multiple screens, bringing together all aspects of their online activities including interactivity, eCommerce and social media.

There is more data than ever before, but it is much harder to make sense of it, with the cleansing or pre-processing operation now far more protracted. It is also becoming harder to separate out the different aspects of QoE, with those relating to content consumption increasingly entwined with audio/video quality. An obvious example concerns the viewer drop-out rate after starting to watch an online item, which can be caused both by video quality problems such as buffering and also content related issues such as insertion of a preroll ad.

These in turn feed the data about content consumption. In the case of buffering there is a very clear relationship between duration of the problem and dropout rate, according to a recent survey from Akamai, which found that 15% of viewers abandon a stream after two seconds of buffering and 50% after 10 seconds. There is a similar relationship between duration and dropout for preroll ads, although there the nature of the ad also comes into play. If the ad is fun or well targeted in terms of relevance, drop out will be far less and it may also trigger a desirable action from the user.

We can now see why online video analytics platforms are increasingly covering all the bases and not just focusing on one aspect such as video quality. This is especially case for the analytics incorporated in leading OTT platforms themselves, from the likes of Ooyala or Kaltura.

Ooyala in May 2015 released Ooyala IQ, a revamped version of its previous analytics product, introducing multi-dimensional analysis without need to pre-configure reports. This does enable operators to measure preroll dropout rates by comparing the number of Plays requested with that of actual video starts counted only when all pre-roll ads are completed. It can track such metrics by content or by syndication partner. At the same time by enabling a separate player to be configured for different content types or sections of a website, it can measure the performance of each.

This in turn can allow operators to assess the impact on overall performance or QoS of factors such as changes in ad loads, the player’s UI or switching on of a content recommendation tool. This latter example highlights the overlap between quality and consumption quite neatly, since operators might be equally interested in measuring the impact of a new search or discovery tool on traffic levels from a QoS and operational perspective as they would be to identify whether it is driving more content consumption and even creating revenue.

Ooyala IQ exemplifies a key trend among OTT analytics platforms towards being able to drill down by user, content type, platform and geography, to field queries such as “how many unique viewers do I have for wildlife videos in Canada on iPads and how much of this content do they consume a day on average?” It may then be possible to proceed to a finer grained level down to individual locality and user demographic. Geography is in fact one of the key challenges and opportunities associated with OTT analytics, whether we are talking about consumption via fixed or mobile networks. In the past, operators knew their customers were at home when watching their channels, which meant the information was invariant and therefore of no value. Now people are watching from all over the place and that information can be interesting for historical analysis, but particularly of course for location based advertising.

One inevitable aspect of online analytics, given the need to improve both presentation and quality of the insights is the fact that the information obtained is not only inconsistent across platforms, but between successive versions of a single vendor’s package. This too is exemplified in Ooyala IQ, which the firm admits can give different answers to the previous version. Data is aggregated differently, geographical reporting is more fine grained and the way metrics are defined has been revised, which can lead to discrepancies in calculations of playthrough and unique users, for example.

Like its rival Ooyala, Kaltura as an OTT platform vendor has been raising its game over analytics and also bringing together quality and consumption aspects, with a strong focus on user engagement. Content drop-off is a bigger issue for OTT than conventional channels and Kaltura can produce reports tabulating videos against drop off and completion rates, showing the distribution of drop off times in each case. The other side of that coin is whether content elicits positive actions and Kaltura follows that too by tracking behaviour or interactions such as editing, sharing and downloading. An obvious extension would be to add tweeting, as some analytics product already offer. We may recall Twitter itself acquiring Bluefin Labs for $40 million back in May 2011, having noted the strong correlation between numbers of tweets while a program is showing and viewing levels.

Kaltura can also produce reports on both bandwidth and storage consumption broken down by user, content and timeframe to help with network capacity planning and to an extent with live operational management. It can also now measure transcoding by the volume in MB of assets output in a given time.

Some OTT platform providers are buying in analytics components. For example New York based Vubiquity, one of the world’s largest multiplatform video services platform providers, selected UK firm Mediamorph for managing content rights and BeBanjo, a subsidiary of video hosting and asset management group Arkena, for scheduling content delivery to video service provider clients.

This is a rather more upstream platform. Mediamorph Connect transforms for Vubiquity the previously cumbersome process of managing content rights and avails (ad slots) from studios, via spreadsheets and emails, to an automated flow of data through APIs. Vubiquity is using Connect to consolidate rights and avails data flowing from hundreds of content providers, with BeBanjo’s Movida product then kicking in to schedule content for delivery to these video service provider clients around the world.

Mediamorph is also being used by other upstream distributors, such as RLJ Entertainment to gain insights into digital sales. RLJ is using the Mediamorph cloud-based platform to track license usage and revenue from Internet-based platforms as well as cable VoD licensees.

Amid all these developments we should remember that there is plenty of scope for traditional analytics firms focused purely on service performance, network monitoring and QoS to play in the OTT world. A good example is Paris based Witbe, famed for its robots, which are essentially software simulating user behaviour, including transactions like renting a movie as well as just watching TV. The robots can report on user experience as far as they are able, based on Mean Opinion Scores (MOS), or perhaps the video specific successor PEVQ (Perceptual Evaluation of Video Quality). The robot can be implemented on a dedicated appliance or increasingly as a virtual machine.

Witbe has just made a key announcement for its OTT strategy of making its Smartping algorithm, designed to pinpoint network links or nodes causing a performance degradation or blip, compatible with its robots. If this works as promised it will address a particular pain point for OTT operators, of being able to associate the quality perceived by end users with specific events or locations within an unmanaged delivery network. It sounds like there is still work to be done making this work transparently and to automate some of the troubleshooting, but it is a step in the right direction. The idea is that operators can now combine Smartping with other Witbe tests. They can now exploit its ability to locate and troubleshoot bottlenecks in delivery chains and correlate that with the existing Witbe OTT, streaming and application tests, as well as with the Witbe MOS technology.

This then is a step towards having a single view combining the quality being delivered to end users on any device at a given time with the instant localization of any network issue that might impact that quality. Witbe says this was developed in response to requests from customers. Quality measures might relate to jerkiness, blurs, blocks, frozen or black screens, as well as content delivery errors. The real time detection of these can now be correlated immediately with diagnosis and location of delivery issues such as incorrect routing, bandwidth saturation, CDN congestion, routing asymmetry (where packets taking different routes to and from destinations) and packet loss.

Of course correlation is not proof and there is still the danger of failing to pinpoint real causes of problems, but there is at least scope for further advances in troubleshooting and automation.