Implementing a ‘Recommended Content’ Feature on Your Website
A key part of your digital success lies in how well your website and applications offer engaging experiences to your audience. To ensure you are getting relevant and timely content to your users when they need it most, consider incorporating the ‘recommended content’ feature on your site.
At Forum One, we specialize in building content management platforms to help mission-driven organizations communicate their ideas and thought leadership. One strategy we often incorporate is a “recommended content” feature, which generally sits in the sidebar or underneath the main body content on a detail page and contains a list of other content that the user might find useful and interesting.
The feature is a win-win: your audiences get easy access to content that is relevant to them when they need it most, and your organization gets users spending more time on your site, visiting more pages, and consuming priority content that enhances your mission.
There are a number of tactics you can use to populate the recommended content list, and the tools we use to do so range from manual to machine. Here are some approaches you can take to implement this on your site.
Please note: the following suggestions assume the site in question does not have a mechanism for authenticating end users and collecting data about their preferences and interests.
Manual recommendations
Organizations that have a small team in charge of updating and maintaining content often opt to populate their content recommendation lists manually. Tactically, this means creating a relationship field in the content model so that the author can select other relevant content when posting a new page or article. While this requires dedicated, manual time to do, the benefit is that the author can easily relate content that might be relevant in a subtle or nuanced way. On the other site, this method requires a lot of historical knowledge of what could be thousands of pieces of content on a large site. It can also be labor-intensive to populate and maintain on a large site.
Related recommendations by taxonomy
A less manual approach is to allow your CMS platform to populate a recommended content list is based on shared taxonomy terms. This means creating a robust and flexible taxonomy structure and being disciplined about tagging content with relevant terms when they are added to your CMS. Drupal and WordPress are both capable of querying for content that shares one or more terms and returning that list sorted by “relatedness,” which shares the most relevant terms in common to the least relevant. If recency is more important than relatedness, then you can also sort by publication date. While this is an automated method, it does still require substantial labor to populate and maintain, and relevant content may be missed if the tagging strategy fails.
“More Like This” recommendations
Another way to populate recommended content is to rely on an external search application that is capable of performing a “More Like This” query. Many of our Drupal sites use Apache Solr, and it offers this functionality. It works by identifying keywords and phrases in a document and then querying for other content that shares them. We have the ability to configure the query to only apply to certain fields, like the body or taxonomy fields. We can also set a “minimum frequency,” which tells Solr to ignore content unless the identified words and phrases appear at least a set number of times. We can “boost,” or give more weight to, certain terms in determining relevance or give preference to content that was posted most recently. This method works well on sites with quality content that is optimized for search; however, it can also be unpredictable in terms of what content is considered “related,” as the algorithms behind it are complex.
Machine learning recommendations
A relatively new and interesting way to populate a content recommendation feature is with machine learning. In this method, we collect and store data about content and users, such as which pages users visit and attributes, keywords and phrases in the content on those pages. Using collaborative filtering, we can use the data to make predictions about what additional content a user might want to read. The dataset can be quite large, and the computational resources needed to make predictions is intensive. For that reason, this method requires third-party software that integrates with your CMS. It is the only method in this list capable of recommending personalized content to individual users based on their interactions with the site. This makes it a very powerful tool for engaging audiences.
Are you ready to create impact?
We'd love to connect and discuss your next project.