We will explore the importance of Recommendation Engines and why it is important for retaining users, reducing churn and supporting monetisation.
There has been an explosion of available content online through a variety of different services in the last few years, HULU, one of the biggest names in the OTT industry, has an estimated 43,000 TV episodes from 1,650 shows, and over 2,500 films! This dramatic increase in content has given birth to a whole new era of content recommendations. It’s safe to say there is a lot of content out there but how do content providers suggest content effectively to their users?
The boom in Recommended Content
It’s the year 2006, libraries of movies and shows are growing rapidly. The engineers at Netflix are searching for a solution to reducing the amount of time a user spends searching for new content to watch.
One of the events that energised recommender engines was the Netflix Prize. From the years 2006 to 2009 Netflix sponsored a $1,000,000 prize competition to improve their recommendations engine. The winner improved recommendations by 10%. This energised other recommender engines for new and more accurate algorithms.
As you have probably guessed, Recommender Engines need a lot of Data to work effectively. Good high quality data though and data that has the correct balance between quantity and quality. If you have incorrect data or insufficient data, your guesses and recommendations could be way off and lead to a poor user experience.
There are two types of main data collection that are used when recommending content.
- Content Metadata
- User Metadata
Content metadata can be obtained from a variety of sources, such as studios or content creators and more often than not it is a one-time ingestion of data. Other sources of Metadata include sources such as IMDb or similar rating websites or agencies that specialize in data collection. A few examples of content metadata information include the following.
- Audio Language(s)
- Closed Captions(s)
- Actors, Directors (Cast & Crew)
- Year of Production / Release
- Synopsis (Description)
Data about the User
Data about the user is another way to collect data for recommending content.
Each company has slightly different rules on the data collection because of several laws that exist to safeguard a user’s right to privacy, such as the European GDPR and Brazilian LGPD laws.
Once you have the users permission to track their activity, you can successfully recommend content to them based on their previous viewing behavior and a similar cohort’s viewing behaviour. A few things to track may include;
- Content Title / ID
- Watch-time or watch duration
- Language Preferences
- Scroll Activity
- Click Activity
After Recommendation engines have collected this information they can start to effectively recommend content based on content metadata and user behaviour.
For New Users
Recommender Engines like long term users. The amount of data collected is directly related to how effective the Engine can recommend content.
For new users, engines will not have enough data to make confident content recommendations to the end-user. Some OTT Platforms suggest content based on their location. For example, a rail on the homepage could show ‘Top 10 Shows in Italy’.
Another way to avoid a bad user onboarding is to ask the user their movie preferences upon signing up. Asking them to choose a couple of shows or movies they have watched in the past or would like to watch in the future feeds the engine some simple metadata to get started with.
Other OTT platforms require the user to input their age, gender and language preferences during sign up to help fuel the recommender engine. Be aware though that with this method you do run the risk of a more complicated sign up flow.
Recommendation engines have been proven to be a great way to reduce churn rate and improve monetisation methods for OTT platforms.
But how do you know, as a content provider, that they work?
Click through rate is a great way to monitor the success of your recommendation engine. For example if you showcase three films to a user and they choose the one that was recommended you know the engine is doing its job. To go a level deeper you can also measure the amount of minutes the user has watched this chosen piece of content to again confirm the engine’s effectiveness.
The main reason why a good recommendation engine is important is to increase a positive user experience. On average Brits spend 187 hours choosing something to watch each year and spend an average of 20 mins to find new content before moving on to something else. With this in mind it is paramount to ensure your recommendations engine is effective and accessible to help reduce these numbers. This will help reduce churn rate and increase your OTT platforms monetization method.
Accessibility obviously plays a huge role in ensuring the engine works. A popular way to showcase recommended content is through a dedicated ‘recommended for you’ rail on the homepage.
DIAGNAL understands the importance of recommended content. We have partnered with Recombee, an AI powered machine learning recommender platform.
Content Providers can explore performance metrics and configure recommendations to reflect your personalization needs using a simple and user-friendly interface designed for all your team members.
Recombee’s recommendation engine analyzes each content property such as title, date of publication, author, tags, expiration date, lead paragraphs or paid/unpaid content applying NLP in 80 languages. Recombee reacts to each new click or newly released content immediately, analyzing interactions like content detail views, bookmarks, scroll down rates or likes.
There are many things to consider when choosing the right OTT solution for your company. To compete with so many emerging services out there a good recommendation engine is paramount to the success of a new or existing OTT business.
ENLIGHT is DIAGNAL’s application solution for premium OTT service providers looking to publish and monetize content quickly.
ENLIGHT provides a collection of feature rich premium application experiences across a wide range of device platforms. Well tested and deployed with global customers, ENLIGHT apps provide high performance and a high end experience to end-users.