As part of our day job providing personalised customer experiences driven by data science, The Filter helps power and reviews many video-based user interfaces for many different content platforms. These range from emergent OTT services through to complete household, digital set-top-box services and across many genres including box sets, films, catch-up, sports, kids and music. These user interfaces have some clear consistency in layout and structure, but we also see some subtle differences. The Filter also has a significant volume of data to tell what is actually driving views and what is not. So, what is the best video-based user interface out there? Is there a perfect interface? Is Netflix as good as you might assume?
Before we get into the detailed debate, we need to start with some more lofty thinking. The most insightful approach to understanding any customer experience is thinking though the mindset of the customer – it is a bit like playing poker; you have to try and think like the other players not yourself.
We think there are three core mindsets that a customer has when coming to any video-based home page:
- DISCOVERY – I have some time free now and I want to fill it with something that I/ we will enjoy watching but I have no idea what that could be.
- CONTINUE – I am in the middle of watching something and want to easily continue watching it (most often the next episode of a TV series but also partial viewing of a movie. Kids also return to watch the same show which we label as continue).
- KNOW – I have heard of a great show or film on your service – normally from a friend, marketing or a review. This can also include live TV, especially sport.
According to our data, these are in the right order. Netflix claims that 20% of its views come from search which is linked mostly to KNOW.
Here is a strawman design for the perfect user interface for an OTT service. Please add a pinch of salt to this as there is clearly no one right answer here:
1. HERO RAIL
a. Just about every service that we see has a well-presented hero rail as the first area for potential viewers to interact with. The Hero Rail should major on new shows that have the potential to be big and get talked about; mostly aimed at the DISCOVERY mindset. But this area should also reflect the brand of the service i.e. irreverent, epic, reality, British, etc.
b. It is important to personalise this hero rail. This is most obvious when you think about content you don’t want to watch. No matter how good the horror movies are in that hero rail, I will never watch it as this is a genre I really avoid. Accounts that are accessed by the whole household have clearly different day parts for kids, family and parents and different devices can be used for different modes of viewing – e.g. sport on mobile. We highly recommend excluding any content that has already been viewed as that is generally better suited to a ‘Continue Watching’ rail. In general, avoid repetition of content across the totality of the user interface.
c. The HERO rail is often editorially driven which allows people who get content to predict what shows and movies will be big and also incorporate brand and business priorities. However, we at The Filter strongly believe that the best hero rails combine editorial and data science personalisation. Let them work together.
a. Now whilst this is not a rail as such, we here at The Filter feel that it is the ‘go to’ destination for the KNOW mindset. You need to make the search prominent, easy to navigate to and quick to see the results.
b. There are a number of important TV specific terms that your search needs to cope with; TV specific mistyping, searching for talent (e.g. Nicolas Cage, Nolan, Peppa Pig), searching for content brands (e.g. Disney, Discovery) and also including all sports results where relevant.
c. As many OTT apps need to work with remote controls inputs such as Roku TV, search needs to be suggestive. Even when customers are using a keyboard, a suggested set of results should appear after two letters have been entered.
d. Once you know a users’ viewing history, you can start to personalise these suggested search results. If a customer always watches engineering programmes, the suggested results from “MA” should focus on all the results from engineering programmes. Obviously, the relevance of any search should never be lost when personalising.
3. BECAUSE YOU WATCHED
a. Because You Watched takes some of the last viewed content for a customer and presents titles that are matched to that item of content. You have just watched ‘The Joker’ and so here are some more of the Batman franchise movies for you. Many services have MORE LIKE THIS which is fundamentally the same just not personalised to your viewing history.
b. Customers like Because You Watched as it does not feel as ‘all knowing’ or ‘creepy’ as Recommended For You. ‘Because You Watched’ is just based on knowing content, not people, users are more accepting of the predictions. We also find it gets the clicks.
4. CONTINUE WATCHING
a. For the CONTINUE mindset, there should be a prominent rail that has all the titles in that a viewer is part way through watching for both TV series, catch-up, and movies on-demand. Moreover, when a series is released as weekly episodes, any new releases should be added here.
b. Ideally this rail should be ‘above the fold’ for any OTT home page. It is an important user journey.
c. However, this rail should be hidden for new or logged out users as it wastes valuable real estate. It would be empty.
5. MOST POPULAR
a. This rail provide reassurance for a customer that they have seen all the big stuff you offer. It should be different from the hero rail in that it focuses less on the new, potentially popular content and more on the established crowd pleasers. We find that it does not change as often as you might expect.
b. This is the one area where you can repeat content in here from CONTINUE WATCHING as customers would like to know that they have covered the bases with their viewing.
a. This rail is centred on the DISCOVERY mindset and is actually a chance to highlight the fastest growing content on your service not just the big titles. It should be calculated as % daily growth in. views not total views. It should be varied by time of day. This rail is a chance to surprise and delight customers with hidden gems and will respond fastest to external events
b. Care should be taken that this rail is not the same as the hero banner. There is a risk of a feedback loop here in that the hero banner promotes something new and makes it grow and so it naturally appears in Most Popular. We recommend down-weighting any content in the hero rail in Trending.
7. RECOMMENDED FOR YOU
a. This is the classic personalisation rail. It uses all the data associated with the content on the service and combined this with each user’s viewing history to generate a set of content that is most likely to work for that user. It blends many different data science models and needs a deep understanding of the content. These rails can be very successful for the DISCOVERY mindset. Say I like slow burn, crime series and comedy themed movies, this rail should highlight more of these for me.
b. Recommended For You can be filtered into more customer friendly, bite-sized chunks. These chunks allow customers to make some initial decision about what works for them at that moment which no AI can know. ‘I feel like something light hearted in lockdown’. It could be that you split Recommended For You into half a dozen rails; “Car Chase Movies for You”, “Dark TV shows for you”, “Originals for you”, etc. Thematic pop-ups also work well “Sean Connery for you”, “Christmas for you”, etc. Moreover, channel, content and franchise brands work as filters here.
c. There is really no limit to how many of these filtered Recommend for You rails you can have. In fact, the more there are the more it shows off the depth of your content catalogue. Many services start to order these rails by their popularity for that customer. If you always click on “Reality TV For You” rail then this rail should move up the page.
So these are the general principles we here at The Filter think are important for a good content based home page. There is clearly no perfect answer to any user interface and in fact the key is having the machine learning and testing in place to continually optimise. Data from real customers is way more important than any subjective reasoning.
So how does Netflix’s home page rate?
Well this is harder to assess as they claim every screen is personalised for that individual. Here is my Netflix home page and it is this which I will critique but I encourage you to look at your home page with the above context.