Home Opinions Personalisation is an art as well as a science. Here are five...

Personalisation is an art as well as a science. Here are five tips to getting it right

Improving content discovery with personalised recommendations reduces time-to-play and increases customer satisfaction. It can also help uncover the gems currently hidden in your content library. Increased engagement is crucial to subscriber retention or ad-funded revenue. But implementing personalisation is no small task. What strategies should streaming services be adopting to maximise their engagement?

Damien Read, SVP Data Products, 24i
Share on

Whether you want to tackle churn, grow your ad revenue, or increase the range of content your viewers are watching, personalisation works. The customers who use our managed personalisation service to drive content recommendations are testament to that. On average, users who interact with the sections of their apps that include recommendations watch for longer and watch a larger range of individual content items than those who only interact with parts of the app that aren’t data-driven. In fact, one of our customers saw the average number of content plays per user session rise by 133% when users were interacting with recommended content!


Five engagement-boosting strategies you should adopt right now

So, what can streaming services do to improve the effectiveness of their personalisation efforts? Here are five engagement-boosting strategies taken from our recent e-guide:


1/ Max-out your metadata

When you’re house-hunting they say it’s all about “location, location, location.” In personalisation and content discovery, your mantra should be “metadata, metadata, metadata!” The better shape your metadata is in, the better content-to-content recommendations you’ll be able to provide. Don’t be afraid to use machine learning (ML) to improve your keyword game. You may need expert help to review your existing metadata and define an appropriate and manageable set of keywords for your specific content. But then ML can help fill in the gaps and improve metadata quality in a cost-effective and timely way.


2/ Feed the desire to binge

Do you encourage users to binge-watch TV series by auto-playing the next episode? Are you missing the same trick in situations where there isn’t an obvious sequel?  Maybe you won’t want to use auto-play in these scenarios, but offering one or two highly-relevant content suggestions tailored to your user’s viewing history at the end of a one-off show or movie, or after the very final episode of a bingeable series, is a great way to keep viewers engaged for longer.


3/ Personalise email to your advantage

Well-timed personalised emails containing “hand-picked content for you” are a fantastic way to encourage repeat visits to your apps. They give users a concrete reason to return that’s more effective than a blanket email about your hot new show that’s sent regardless of whether they’ve shown any interest in similar content before. One of our customers saw a 121% uplift in visits to their app within three days after adding personalisation to their emails in A/B testing.  For users who regularly visit your apps from their mobile devices, a personalised content recommendation via push notification can also be highly effective.


4/ Match your message to your goals

There are plenty of different options for how you present data-driven recommendations, from “Recommended For You” to “Trending” and “More Like This.” The subliminal messaging of each is slightly different and requires different data models behind them. The first step is to identify your goals – reduce churn, increase video views, grow the range of content watched, promote your exclusive content etc. – and then work with data scientists to test and confirm which models and messages will deliver the best results for each of those specific goals.


5/ Layer algorithms on top of your editorial picks

Don’t be fooled into thinking that you can replace your editorial team with algorithms. You need both. But you can make them work smarter together. For every editorially selected section of your service (including the all-important “hero banner”) add a layer of intelligence to determine the order in which content should be shown to each individual user.


What’s next? Test, refine, learn, innovate!

Recommendations can’t be a simple “set it and forget it” process. Even the five highly effective strategies outlined above need constant work. Regular A/B testing will reveal what works for your specific content library, user base and market.


Share on