The Hidden costs of OTT TV

I spent an hour or so trying to get my son's iPad working with Sky Go. It worked for 11 months and then suddenly stopped for reasons that are not apparent to me or (so far) Sky's technical support people. While this is frustrating for me and my son (he has to resort to sharing the TV in the living room!) it is an example of how hard it is to reliably run TV services across the Internet to customer owned devices. (A bit like BYOD in the corporate world!)

To be fair to Sky, their customer service is no worse that their competitors and they do try to "push the envelope" on new technologies, so their will be some pain from time to time. However, they have bundled Sky Go into their service packages as they did in the past with Sky+ and entry level broadband in an attempt to reduce churn and now they have to support it!

OTT (or "over the top") TV is the idea of delivering (mainly) TV content via the Internet rather than paying for your own distribution system like satellite, cable or terrestrial transmitters. YouTube and BBC iPlayer where early (and free) entrant into this sector. Netflix and Hula are probably the best known pay providers in OTT and all the traditional operators are trying or thinking about getting into this market, either because they see it as "strategic" or as a defensive ploy.

OTT sounds like a free ride as you don't have to buy network capacity to the home? However, the real cost of OTT is the need to make the service work with all devices that you "support." This is not so bad if it's a "free" service and it's probably no coincidence that YouTube got going (and still is largely) free. People don't generally complain if the they are not paying in for first place. However, when you start selling a service and say it will work on a list of devices, customers expect it to work or they will churn out.

In the old world of closed networks and devices, broadcast platforms (e.g. BSkyB, Virgin, UPC, DirectTV, DISH, etc) tested each new set top box (STB), firmware download or network update to death. And even then, they rolled it out slowly to "trialists" or "friendly customers" (most customers are unfriendly, you see ) before unleashing it on the whole population. The reason they can do this process effectively is that everything is locked down - they know exactly what the hardware is, they know the versions of all the software components and they know the network environment. These factors are all reproduced as closely as possible in the test lab, so the test results are reasonably accurate. There are still surprises when new products are put into the real world - real people do things that were not anticipated or unexpected environmental factors cause problems e.g. certain fluorescent lights interfere with remote controls. These environmental factors are usually caught in trials, but scaling issues can still be missed until full deployment.

So, even with everything under control and nailed down, it's quite hard to catch all the bugs. In the OTT world, operators usually offer support for half a dozen devices. Doesn't sound like too many, right? Well, each device probably has half a dozen operating system versions in field .... and maybe several hardware versions. In the case of PCs in particular, there may be several web browsers and several user contexts. Apart from games consoles (which usually run one application/game at a time) there are an almost infinite combination of software that co-exists on the device and may interfere with your content delivery app. Now multiply these factors together and you get a huge multidimensional matrix of possible environments that could be tested. Oh yes, I also forgot the local network context, which will change depending on your ISP and home router, etc.

Delivering OTT to many devices is a hard problem, but not insoluble. We are seeing the following approaches -

1. Aggregation. Hula, Netflix, Lovefilm Instant are mainly content aggregators (House of Cards excepted!) and their value is in their ability to ingest content, transform it for target devices and manage that process. Apple is a special case of an aggregator as they own the delivery chain down to the device (and have the customers' credit card details!) Amazon are also trying to play here. So the cost of testing is spread over a large portfolio of content and therefore purchases.
2. Targeting of Devices This is the idea of focusing your effort on the top devices (by market share) until the cost of supporting a device (and/or web browser) is not justified by the likely revenues. Even the BBC, with their commitment to broad access, had to cull their supported devices.
3. Risk Based Testing we can't test the whole matrix of possible contexts, but we can test the most likely and/or representative. Good feedback from customer support is important here.
4 Customer Support There will be many more customers who can't get the service to work than in the traditional model of a set top box supplied by the operator and customer support needs to reflect this. We see the use of forums to encourage users to help of users either for kudos, discounts or just good will.

The increasing maturity of software stacks also helps in this process. APIs that really do abstract the hardware and lower layers and software that traps errors make life much more predictable for application developers and testers.

The final twist on this is DRM. It is hard enough to make an end to end system work "in the clear." DRM adds a parrallel delivery chain. Not only do I need the content, I also need the licence and decryption keys. These have to be delivered in sync to make it all work. That would be hard enough, but DRM is usually designed to work as a "black box." This is (understandably) to stop people (hackers) looking in and defeating it. However the effect is that we often end up without a meaningful error message when things break. (My problem with Sky appears to be DRM related. There is no error message and the "clear" promo video plays.) On top of this, DRM is designed to be sensitive to unexpected changes in the environment as these might be hacking attempts. For example, the system clock going out of tolerance or the presence of an unknown/suspicious app on your device. Again, it usually won't say what it doesn't like as this information can be useful to hackers.

So far, I haven't mentioned UltraViolet which is a Hollywood sponsored attempt to allow a "buy once, play anywhere" model of DRM. It also defines a common format for the content itself. I think that it's unlikely that we will make a common format stick as the format is driven by the constraints of the user device and even if we can agree now, new devices will come along with new constraints and capabilities. The other problem with UltraViolet is the absence of Apple, who would say that they have already solved this problem (if you buy Apple!) So UltraViolet becomes another row (or rows) in the matrix!

So, what are the conclusions? Basically, OTT is hard and we should expect to see a few aggregators emerge, either as brands or as white label providers (e.g. KIT Digital.) These aggregators will probably be the names mentioned already, as scale is the key to solving this problem. The more customers you have, the cheaper testing and development is per customer.

By the way, my spell checker keeps trying to change "aggregators" to "aggressors" which is probably how it feels if you are an incumbent operator