There’s a raised awareness of the value provided by user recommendations to drive engagement and reduce subscriber churn, but OTT providers are being faced with a catch-22 situation when attempting to implement the underlying technology. Of the two main options, off-the-shelf platforms are expensive, and custom deployments require specialist skills. The ideal solution needs to be both cost-effective and easy to use, as well as providing new and exciting recommendation strategies to keep eyeballs on content offerings and reduce churn.
The shortfalls of current go-to strategies
The dreaded combination of sustained economic pressure and a market approaching saturation has meant that OTT providers are keeping a close eye on their budgets. Off-the-shelf solutions can provide the foundation for personalised recommendations, but it doesn’t make financial sense to bring them in when they carry high capital and per subscriber costs. This leads to a poor return on investment and a line of questioning from business leaders looking for an explanation behind the expense.
OTT providers are therefore looking in-house for a cost-effective solution, but are falling at the first hurdle as employees lack the specialist skill sets to build a custom platform. To overcome these challenges, content providers require a flexible and budget-friendly solution that combines data, analytics and machine learning (ML) to deliver the most relevant recommendations to expectant viewers.
Powering personalisation in numerous ways
A server-less, scalable and unique personalisation platform, such as AWS Personalize, is cost-effective to run, requires little knowledge of ML and when implemented well, is robust and secure.
Such a solution ticks many boxes for OTT providers looking to tap into the recommendation opportunity. Once the solution is operating, organisations may find that the use cases extend far beyond what they originally thought was possible. A ‘recommended for you’ carousel on a home page, while valuable, is just the tip of the iceberg when it comes to consumer engagement.
By leveraging user interaction data effectively, users could be shown a ‘more like this’ or ‘other users also watched’ carousel. A viewer that’s just finished a documentary about the life and times of a prominent association football player might be interested in the career of a golf legend. Previous viewing habits can help inform the journey for others with similar interests. There’s the potential for OTT providers to re-rank search results or even entire lists of editorially curated content recommendations. This is a particularly exciting use case as each user could see something completely unique from a common search result, based on what they most want to watch. AWS Personalize was not specifically developed for video content, thus OTT providers can benefit from personalised recommendation algorithms that have been successfully used in other markets.
There are also some interesting and new directions that the technology can take. For instance, if the OTT provider wants to promote a particular content item in recommendation results. The recommender can be configured to show a certain percentage of content from a promotions category. Contextual recommendations are also becoming a reality. A user’s viewing habits are likely to vary depending on a number of factors, including the device used, the time of day or where they are when the content is consumed, meaning that recommendations can be tailored based on these variables.
Deploying AWS Personalize in a media context is more complex than many other AWS managed services. This is where toolkits such as the Merapar Development Kit (MDK) can help to take the hard work out of provisioning AWS Personalize, supporting the infrastructure needed to transform and ingest data at scale. By using a toolkit such as this to integrate AWS Personalize, an OTT provider can be up and running quickly, taking advantage of the modular approach where the provider can use as much or as little as is needed. This is a great solution for OTT providers who want to bring new features to market quickly, to remain competitive and allows features to be rapidly tried to determine what works the most effectively in terms of promoting content.
Continuous testing for future improvements
There are also some clever ways in which this technology can test, monitor and ultimately make improvements to recommendation services. Multiple recommendation models can be trained with user data, with parameters adjusted automatically and the best performer then used to go forward. This helps to improve the accuracy of recommendations.
Beyond accuracy, it’s important to know whether users engage with the suggestions that are put to them. Clickstream data can be sent to a corresponding analytics platform to enable providers to monitor vital KPIs. This might include the click through rate (CTR), conversion rate of journeys that began with a recommendation click, or even the uplift in overall engagement of customers that click recommendations or watch recommended content. All of these insights can enable a constant evaluation and optimisation of recommendation services to increase user engagement and reduce churn.
Driving competitive differentiation
Differentiation has never been more vital in a market that’s adding new providers on a regular basis. But the need to keep users engaged in content has been made all the more urgent by the cost-of-living crisis, where more people are questioning the value of their current subscriptions. A range of varying recommendation services, from basic carousels to contextual suggestions are valuable tools in driving competitive advantage. The question for many providers is how these services can be implemented in a cost-effective way and without relying on the limited skillsets of in-house expertise. Server-less, scalable personalisation platforms, deployed with a corresponding toolkit, can drive recommendation services that resonate with viewers.
