Staking your fortune on FAST? Don’t neglect the data!

Linear TV isn’t dead, but the internet has changed it forever. At a time of global downturn, an increasing number of content owners are pinning their growth hopes on ad-funded live or  VOD-to-live FAST channels. But drawing eyeballs and maximising ad revenues in this increasingly crowded marketplace will rely on more than just an EPG and good promos.

VOD personalization is now increasingly popular and as users we’re familiar with data-driven concepts like “recommended for you,” “trending for you,” “because you watched The Crown…” and “more like this”. At 24i, we’ve been working with a number of clients to translate these concepts into recommendations for what’s on LIVE now in linear and FAST channels. Here are three key lessons we’ve learned, along with a reminder that the data that’s already held in every streaming service can be put to better use with contextual advertising and win a higher CPM on FAST channels.

1. Data can turbo-charge content discovery for FAST

Content discovery has long been one of streaming’s toughest challenges. Personalised recommendations short-cut the search for something interesting to watch in a world of ballooning choice. They represent a great user experience and consequently help to drive user satisfaction. Some of our customers have seen recommendations drive VOD engagement by as much as 133%.

As the range of FAST channels grows, the data available in streaming services about user behaviour and personal preferences make it easier to enhance the traditional multi-channel electronic programme guide (EPG). Personalisation models can put together the content a viewer is most likely to watch and the content that is live right now and displays them in real-time in a prominent “On Now for You” rail where it’s most likely to catch their eye.

Push notifications, emails or in-app banners can also be sent (subject to relevant permissions) to remind users that shows they’re likely to be interested in are starting, or will be live in a few minutes. For example, if data shows you’re a fan of a particular series, you’ll get a gentle hint in the UI at 7.55pm that an episode starts live at 8pm.

2. Live recommendations depend on different modelling

The secret to successful recommendations is getting the right mix of algorithmic models to pick the content you’re suggesting. With VOD recommendations we weight our models towards content matching, helping to connect users with more content that’s thematically linked to the programmes they’ve already viewed. We’ve found that for live recommendations, it’s effective to lean more heavily on neural networks and pattern recognition when building your algorithm.

Neural networks aim to find underlying relationships in a set of data. So they’ll see that a user, I’ll call him Jim, watched three episodes of a comedy TV series and then watched a dating reality TV show. At the simplest level (and obviously with algorithms it’s never simple) we can determine that it’s worth recommending an upcoming live stream of the same dating show to a similar user who also watched the same three comedy shows as Jim.

The addition of pattern recognition compares Jim to other users - much like the classic phrase we’ve all seen while online shopping “people who bought this also bought…” in order to find the programmes that are on live right now which are most likely to be attractive to a specific user.

3. User experience can be tricky

Although personalisation can be subtle and invisible to the average user, this poses a challenge for services that offer a combination of VOD and live content and want to start recommending both. How do you differentiate between them to avoid user confusion?

Imagine the classic “recommended for you” rail in a streaming service. If you don’t clearly signpost that some of these items are live, you run the risk of annoying users who click on an item and find they’ve missed the first 10 minutes of the show. Most EPGs show you how much of the current programme has already been shown so you’re aware of what you’re getting into. Live recommendations need a way to do the same.

Then comes the question of timing around live recommendations. Should you promote content that’s on live NOW or on live in the next 5 minutes? At what point does it become too late to promote something that’s live because the user has already missed so much of the programme that they’re better off catching it as VOD later on? Our customers are experimenting with various UX options - from a simple “LIVE” icon alongside the relevant items, to a form of countdown clock. The perfect UX will depend on your mix of VOD and live and the volume of live channels you have to offer.

4. The metadata challenge hasn’t gone away

For every single streaming service we talk to about personalisation, the main challenge is metadata; getting it, cleaning it, matching it. Live TV recommendations can complicate this even further, as one of our customers recently found out. Their VOD library and their Live playout system were using completely different asset IDs for the same series and episodes. This kind of mis-match makes it much harder to connect the dots between a user’s VOD viewing habits and the content that’s on live now.

The metadata used for VOD and catch-up is typically much richer - with keywords and categories - than the simple title and synopsis that comes with most Live TV EPGs. In many cases this richness has only come about at great cost of time and effort in enhancing the metadata for search and recommendations. While this shouldn’t be a problem in a VOD-to-live scenario, broadcasters may need to shift their metadata enhancement processes further up the content value chain if they want to surface the best of their live content to individual users.

5. Data could help fast-track higher CPMs for FAST channels

Every streaming service should be looking to make the most of the data that’s already being collected by their platforms. It’s a top priority for most of the companies we speak to about personalisation. In the world of FAST, that should include providing more of that data to advertising networks. The “context” in contextual advertising shouldn’t just mean data about what’s being shown on the screen at a particular point in the show. Alongside the demographic information that we’re passing about an individual viewer, services should also be looking to supply (anonymised) information about their viewing history.

If we know that Jim has watched 12 home improvement shows in the past few weeks, that demonstrates a potential intent-to-buy for homewares that makes him far more valuable to certain advertisers than a random male FAST viewer of his age and geographic location. Data on whether a household is watching lots of kids TV could also help determine their value to advertisers, even when the kids are safely in bed and the adults aren’t watching kids content. Streaming services need to harness the data available and find new ways to supply it in real-time to their ad networks to get the most value out of programmatic advertising.

At 24i, we’re working with a wide range of video service providers who are keen to make the very best of the data in their streaming platforms to maximise engagement and drive revenue. Contact us for more insights into the main considerations when offering your users a choice of models for accessing content, including AVOD, FAST, SVOD and TVOD.