As you know, personalization is vital to creating the best user experience for OTT service subscribers. Are you looking to improve your user engagement with personalization? You can do this by investing in a recommendation engine that powers personalized experiences. Personalization is not only keeping customers engaged but helping to improve revenues, a PWC study found that 43% of consumers are willing to pay more for a personalized OTT experience. It’s clear the importance of recommendation and personalization. Now organizations face a critical decision: should they build their own recommendation system in-house or purchase an existing solution? This analysis explores the key considerations, differences and implications of both approaches.
What is a recommendation engine?
A recommendation engine (RE) is a software system that uses data analysis, statistic and deterministic algorithms, machine learning algorithms and Artificial Intelligence (AI) to suggest relevant items to users. It processes data about users’ behaviors, preferences and characteristics to predict what they might be interested in. Recommendations are a proven driver for engagement for streaming and TV. Netflix claims that more than 80% of what people watch on their platform is from their recommendation system.
You can increase overall engagement by understanding your user’s behavior and watching preferences with a RE. If you have the right solution, it can power dynamic user interfaces, notifications, tailored ad experiences and relevant upsell offers for subscribers.
When deciding what steps to take when determining to either build or work with a third-party, product and marketing teams with all other stakeholders need to have a complete understanding of what the scope of work entails and the resources required.
The Build Approach
Building an in-house recommendation system offers several distinct advantages, but comes with significant challenges.
Advantages of Building
Complete Control and Customization: Internal teams can tailor the system precisely to their specific content library, user base and business objectives. This flexibility allows for unique features that differentiate the service from competitors and adapt quickly to changing market demands. It gives the opportunity for your team to build the algorithms and the system exactly how you want it, using the right data that you need.
Proprietary Intelligence: The algorithms and insights developed become intellectual property, potentially creating competitive advantages. The data collected, lessons learned and control over the solution remain within the organization, contributing to in-house know-how about the ins and outs of the system.
Integration Flexibility: Custom-built systems can be deeply integrated with existing infrastructure, allowing for seamless data flow between different parts of the platform and easy implementation of business rules. Within a platform ecosystem, there are many different parts that a recommendation engine needs to seamlessly integrate with, when building from scratch, so some teams believe it can be easier to build to fit into an existing ecosystem.
Challenges of Building
Resource Investment: Developing a recommendation system requires input from many different people within your team, such as:
- Data scientists and machine learning engineers
- Backend developers for infrastructure
- Frontend developers for user interface
- DevOps engineers for deployment and maintenance
- Data engineers for data management
- Product managers to manage releases and the roadmap
- Dedicated, agile software developers
- Tester and quality assurance experts
Not only can this take away team members from existing projects, but it also means that the full understanding of the system is with specific team members. And there is always a risk that the knowledge could leave with them.
According to two recent Garnet reports, 85% of AI and ML projects fail to deliver, this is high-risk for businesses to take on. As well as a risk of failure, these projects can take up a lot of time, for example, a study finds that 90% of the time on AI and ML projects is spent on solving data roadblocks.
Time to Market: Building from scratch typically takes 12-18 months to reach a point where it is working within the platform, albeit for the first set of business goals. It takes time to understand how the algorithms will work, how teams want them to work within platforms and how to ensure 24/7 operation and scalability of such a new system. This timeline can impact business objectives and market positioning. One study by Bain & Company found that companies that buy solutions achieved a 50% faster time-to-market compared to building from scratch.
Ongoing Maintenance: Internal teams must continuously, for example, update algorithms to prevent recommendation staleness, scale infrastructure as the subscriber base grows and implement new features and improvements. Not only can this be difficult to manage operationally, but it can become more expensive over time.
The Buy Approach
The alternative to building in-house, is to work with a vendor who specializes in recommendation engine technology. Purchasing a recommendation system provides immediate capabilities, but requires careful vendor selection and integration planning, yet brings expertise and a deep understanding of a whole project.
Advantages of Buying
Rapid Deployment: Commercial solutions offer proven technologies that can be implemented within months rather than years, allowing faster time to market. Some solutions, like XroadMedia, can even be tested before you buy, so you can see firsthand the results of the personalization on your KPIs.
Proven Technology: Established vendors have refined their systems across multiple implementations. This includes tried and tested algorithms, scalable infrastructure, regular updates and also the expertise that comes with it, like guidance into best practices in recommendation and personalization strategies.
Predictable Costs: Licensing models typically provide clearer cost structures compared to the variable expenses of internal development and maintenance. When comparing the two options, this could be seen as more appealing, as it is easier to budget and prepare for, compared to when building it in-house.
Challenges of Buying
Limited Customization: While vendors offer configuration options, fundamental changes to the recommendation logic or unique features may be impossible or require vendor cooperation. This isn’t always the case, though, with our backend tools, our personalization is flexible to what features and algorithms can be utilized.
Vendor Dependencies: Organizations must consider:
- Long-term vendor viability
- Contract terms and pricing changes
- Integration support quality
- Data ownership and privacy implications
Integration Complexity: Despite being “ready-made,” solutions still require effort to integrate with existing ecosystems. To ensure that the recommendations are accurate there is some leg work to be done with the metadata of the catalogs. Similar to when you buy any piece of technology, there is the need to train teams on how they can work with them.
XroadMedia’s solution is delivered with one single API and can deliver recommendations and personalization beyond the UI
Decision Time – Build or Buy?
When comparing options, all stakeholders must be aware of all of the possible factors and outcomes that will be considered. Upgrading an existing solution or integrating a new provider is a project that can impact different areas of a business.
Hybrid Approaches
For some businesses, the best outcome is opting for a hybrid approach. Combining the two options for a period of time or indefinitely. The most effective way for the hybrid approach is to contract a specialist solution, combined with in-house know-how and expertise on how best to apply the chosen technology to make it work best for your team and your service.
1. Buy the Core Technology, Build Features: Invest in the technology for the recommendation engine then construct specific features or data processing or utilize the algorithms to build relevant use cases.
2. Multiple Vendor Strategy: Combine different vendor solutions for various aspects of the recommendation system, maintaining flexibility and reducing vendor lock-in. The downside of this approach is that it can become very cumbersome and expensive to control and maintain many different systems.
For the hybrid approach, it’s vital that the recommendation system selected is a flexible solution and not a black-box. Success often lies in starting with a clear understanding of requirements and constraints and then choosing an approach that balances immediate needs with long-term goals. As the streaming market continues to evolve, the ability to adapt and improve recommendation systems remains crucial for maintaining competitive advantage.
XroadMedia’s Approach
At XroadMedia, Everything is Personal. Even with how we work with service providers, our solution is a flexible toolbox offering multiple algorithms and use cases and can work around you and your KPIs. So even if you are buying from a third-party, you can have complete control over how it is implemented within your platform. From which rows are personalized, which features are used and which subscribers you can include within the tailored experiences.
We understand that it can sometimes be tricky to work with vendors. But we are different as we offer a free trial so you can test and understand how our solution would be implemented and how your teams could work with it, before committing resources. Our solution is delivered via one single API, so your team can benefit from many of our features from day 1, after a low-effort and seamless integration. The personalized experiences we deliver are all based on anonymized data, being GDPR compliant and reassuring for you and your end-users.
Working across different teams, different business models is something our experts are used to. The backend tools we provide will allow your team to ensure that the solution works for them and your KPIs.
The build vs. buy decision for recommendation systems requires careful evaluation of numerous factors. While building offers greater control and customization, buying provides faster implementation and proven technology. Organizations must align their choice with business objectives, available resources and long-term strategy. As the streaming market continues to evolve, the ability to adapt and improve recommendation systems remains crucial for maintaining competitive advantage.
If you want to know more about how our expertise could help you, get in touch with our team and we’d love to discuss what an implementation looks like with us!
