Why DigitalRefugee?
The Future of TV Tech: Applications

The Future of TV Tech: Recommendations


Today, amongst TV service providers, there is an ever-increasing focus on content discovery and recommendations. Some say that content discovery is the new battle ground for TV service providers, both traditional and new entrant. I still strongly believe that the content itself is still king, but service providers need to work to keep consumers finding and viewing the content that differentiates them.

This article looks at TV recommendations, do they work, what is the future for them and how this affects the TV service provider landscape.

Alert: In this article a media executive looks at his own behaviour and uses to generalized about the industry. Read further at your own risk.

Do we use recommendations?

Stranger-thingsThe easy answer to the question is yes, all the time, but no necessarily in the way you might be thinking. For instance, my friend Hardeep recommended Netflix’s Stranger Things to me and now I’m really enjoying the show. Hardeep didn’t have deep insight into my viewing habits, he doesn’t know me that well, but I was won over by his passion for the show. Looking at the shows I’m currently watching on Netflix and those I’ve watched in the past, I haven’t discovered any of them through their recommendation service, it has either been specific friend recommendations or due to an inescapable buzz.

This doesn’t just apply to TV. My wife has an acquaintance but who she gets her book recommendations from, they don’t know each other that well but share a very similar taste in literature. The acquaintance posts her recommendations on social media which my wife picks up.

Recommendations certainly work, we look for them and use them in all aspects of our life including what to watch on TV. The real question is: do we use machine driven recommendations and do they work. To answer this question, it is worth realizing that when it comes to TV recommendation we each have multiple need states that need to be satisfied.

Recommendation Need States

I personally have three distinct need states for TV recommendations, yours may be similar. These are:

  • Watch something now
  • Watch something special
  • Offline discovery

These all overlap but do have distinct characteristics. It worth digging into these.

Watch Something Now

This is when I want to watch something straight away. The discovery/recommendation system I use for this is usually myself and my PVR planner. The PVR planner on my STB has a line-up of shows that I’m currently watching and I can scan down the relatively short list to find something that matches my mood, the people I’m with (my wife, my daughter or both) and the time available. These days I also have the Netflix resume watching list. It would be great if this was combined with my planner. I also sometimes have a couple of current shows that I know about but missed recording on my PVR and need to watch through a catch-up service. These are just in my head, so are easily lost.

For other people watch now is usual satisfied with whatever they find on linear TV. There relying on the “recommendations” of the channel scheduler.

TV service provider recommendation come into their own when people fail to find something to watch through their usual method, be that through channel surfing or going to a PVR planner. I personally try to avoid using any recommendation services when I want to watch something straight away, as I fear being sucked into a discovery process that will eat into my valuable TV viewing time.

Watch Something Special

This is when I want to watch something special. For me typically this is a film, as it is usual when I have more time to watch TV. If I’m lucky there might be something in my planner that I’ve recorded previously, but usually I’m off searching iTunes. Once I’ve found something to watch then I have a quick check on Netflix to see if they have it available, to save a few pennies.

Satisfying this requirement is when people are most reliant on search and discovery services provided by their TV provider.

When I want to watch something special I’m prepared to spend time in the discovery process, browsing lists, but to be honest the quicker this is the better.

Offline Discovery

Most of the time I watch TV I don’t want to waste any time finding something to watch, so nearly everything new that I start watching I discover offline away from the TV set. I might do this by chatting to people about what they are currently watching or their favourite shows. It might be when I’m reading reviews or even perusing a recommendation service.

A great service offered by some service and content providers is an email with details of new shows that are starting in coming weeks. These are particularly useful when they have a button to automatically have a show recorded on my PVR, to have a reminder set or to bookmark them for on-demand consumption. I recently discovered The Split from the BBC from an email they sent me, sadly I then had to manually search for this on my service provider’s mobile app to book the recording.

These emails are typically curated by an editorial team, but they can benefit from recommendation technology.

Managing my TV experience is becoming more and more complicated and I’m drawn to solution that will simplify it. Ensuring I don’t suffer from FOMO (fear of missing out or FONTW (fear of nothing to watch).

Where do recommendation go wrong?

LawnmowersThere is demand for machine recommendation but do they work? We have plenty examples of where recommendation don’t work. The classic is you search for a lawnmower online, you buy a lawnmower and then you constantly receive recommendations (be that in ads or in recommendation lists) for lawnmowers. What is going wrong?

"with perfect information a recommendation system can do a great job, but few recommendation systems, if any, have perfect information."

Some of this is just bad technology or bad product owners of good technology. But the real issue is knowledge. Somehow the knowledge that I have bought a lawnmower is lost by the systems and more importantly the systems don’t know I only buy one lawnmower every ten years or the fact that I bought the lawnmower for my father in law, who has a much larger garden than me and needs a more substantial lawnmower.

My argument is that with perfect information a recommendation system can do a great job, but few recommendation systems, if any, have perfect information.

Information Requirements

Moving on from lawnmowers, it is worth getting back to TV and what information is required to make great recommendations.

The audience – who is going to watch, this can be inferred from the time of day and past viewing at that time but these days in my house it is pretty random if the audience will be just me, my wife and me or the whole family.

The audience’s mood – what kind of thing they want to watch. My wife and I love a good gritty police drama, but sometimes we just want something to cheer us up.

Time available – how long the audience has to watch TV. This may not be that common, but I’m often triaging choices based on length to fit in the time available. I for one need my beauty sleep but also am very bad at stopping watching something half way through, however late it is getting.

What’s currently been watch – most of us watch TV series and we need to keep track of where we. One issue my wife and I have is we can’t watch too many series at once, we have a limit before we start to confuse the plot lines.

What’s been seen before – this is at the heart of the lawnmower issue, if I saw a film at the cinema or the back of an airplane seat then I won’t want to watch it again, well at least for a while.

The rise of Digital Assistants

How can recommendation providers gain this knowledge? The ultimate answer is a digital personal assistant, always active and attentive personal digital assistant.  They can know who is in the room, what mood you are in, what time you have available and importantly what you have watched before.

Imagine the following output form a personal assistant:

“Hello Matthew, I believe you would like to watch some TV. I recommended finishing watching an episode of McMafia, yesterday you fell asleep at 26 mins into the last episode and you can resume watching from there. You have enough time to either watch the following episode or watch the next episode of Homeland, you haven’t watched this for two weeks. You could start watching the City & the City if you want to try something new”. The_city_and_the_city_header_david_morrissey

Can such technology exist? We are not far away from devices that monitor everything we do. My watch knows my heartrate and where I am, amongst other things, almost all the time. It won’t be long until we have devices that can hear all that we hear. Though I think we are a way off devices seeing all we see. Hearing may be good enough to know most what we have seen, especially television, after all TV audience measurement system use sound recognition/watermarks today.

The data from an always listening, always aware digital assistant would also be very valuable to advertisers, a key revenue source for television.

The issue for TV service providers is that personal assistance to help us find TV to watch are going to be used for a lot of other activities and are going to come from smart device/home/life providers. The leaders of course being Apple, Google and Amazon, though Samsung is trying to be part of that crowd. But these smart device providers are also becoming TV service providers, so we have to ask the question if recommendation are a real differentiator, do the traditional TV providers have a place in the future?

If content discover and recommendations and going to be even more critical in the future then the smart device providers are going to have an edge. But it will take time for such technology to advance. I would not say that in the six years since to launch of Siri, it has not made any great advances towards being a true digital assistant. Also, the required data sharing requires a user confidence that has recently been dented by Facebook/Cambridge Analytica and other stories.

There is another reason our smart phones and smart watches are not already always listening and that is battery life. Battery life is a huge compromise in smart devices and the increased load of an always on microphone and the data processing required to make sense of what is being heard pushes the concept out further in time.

What can TV Service Providers do?

I believe the TV service providers have a window of opportunity to make their services better before they are overtaken by the smart device providers. A step they need to take is to properly aggregate content services together. This requires data sharing with their content providers who these days also have a direct to consumer offering. This includes catch-up services like BBC iPlayer and SVOD services like Netflix.  This means if I watch a Netflix show on my tablet, my TV service provider will know about it and take it into account in the content discovery service.

Having worked with these content providers, asking for such data sharing is difficult, but with everything there is a negotiation to be had. The inclusion of Netflix in the Sky service, bundled in the Sky subscription is a good example that deals can always be done. Also, some providers are already sharing this information with Apple when the content is consumed on an Apple TV, as details in my previous article.


Effective content discovery and recommendation is important to TV viewers and a differentiator between platforms. The bundling of Netflix with a Sky subscription is only going to be good for Sky if the SkyQ box is used to discover Netflix content instead of consumer accessing Netflix through another devices (e.g. Chromecast).

"there is really no difference between recommendations and advertising."

The final point I would like to make is there is really no difference between recommendations and advertising. A recommendation is simply an ad for a piece of content that a consumer is going to want to watch.  Maximizing revenue is not always the objective, though maximizing viewing time nearly always is. Recommendations and well targeted ads both need data to be successful. In fact, the data required for recommendation is a subset of the data required by advertisers.

In many ways, what recommendations are trying to do today are ahead of where advertising is in really understand an individual’s propensity towards an action.

Success of both recommendations and advertising on TV will, now and in the future, be tightly linked.

More on the future of advertising in future articles.

Matthew Huntington

DigitalRefugee Consulting


Feed You can follow this conversation by subscribing to the comment feed for this post.

Verify your Comment

Previewing your Comment

This is only a preview. Your comment has not yet been posted.

Your comment could not be posted. Error type:
Your comment has been posted. Post another comment

The letters and numbers you entered did not match the image. Please try again.

As a final step before posting your comment, enter the letters and numbers you see in the image below. This prevents automated programs from posting comments.

Having trouble reading this image? View an alternate.


Post a comment

Your Information

(Name and email address are required. Email address will not be displayed with the comment.)