ent recommendations: How to increase website engagement

Content providers want more people to engage with their content online. This can be achieved by using content recommendations. But there is no one-size fits all approach.

This is how you can create the best recommendation strategy to your website, content, and audience. Although most content recommendation engines that come off the shelf won’t have the same capabilities, knowing what is possible can help you to find the right solution for your company.

What is content recommendation?

Based on their likely interest, content recommendation systems recommend additional content. Take, for example:

• YouTube and Netflix use content recommendation to recommend additional TV shows and videos based on users’ viewing history.
• Spotify looks for patterns in music and suggests similar songs.
• My daughter claims that TikTok finds relevant content for her amazingly well.

In all cases, the goal is to keep visitors engaged with compelling content on your platform. This raises two important questions.

• How does the system decide what to recommend?
• What is the context for the recommendation?

How content recommendation works

Data analysis is used to predict the content that a user will engage with. It collects information about user behavior such as which pages they visited, what they clicked on, and how long they spent on each page. The system can then generate various types of recommendations, such as:

• You can find the most popular articles right now on this site.
• Popular articles within a particular category
• Popular articles by a particular author
• Visitors who have read the article are able to view it.
• Visitors with similar browsing history have read articles.
• Popular articles that are relevant to a particular job title.
• Articles are read by readers who are similar to the author.
• Articles that are read in a particular area.

Each option may have a different purpose for different content or areas of your website. Some of these options (such as “most popular on this site right now”) are based only on analytics. Others (“people like these articles”) are based upon look-alike modeling.

Regulars vs. drive-bys

Your site may have many visitors who only read one article, then go elsewhere. It can make a big difference in traffic to your site if you get some of these “drive-bys”, who will stay longer on each page. One way to solve this problem is with content recommendation.

You don’t know much about drive-bys. It’s much harder to do look-alike modeling because they don’t have a history on your site. There are still options.

• Third-party cookies/audience data can be used for as long as they are still available.
• Data from the HTTP header can be used, such as geolocation and referrer.
• You can also rely on the general statistics of your readers.

With your regular visitors, you have many more options. You can also make predictions based upon their browsing history.

• Display content similar to what they have seen (in the same category, same author, same tags or keywords etc.).
• Comparing their browsing history with others with similar browsing histories will show you the most popular articles in that group.
• You can display the most popular articles to people who have that job title if you have demographic data.

Multiple audiences

There are many sites that have different audiences. For example, there may be free users and paid users. If this is the case, it’s important to keep them apart so you can make the best content recommendations.

Here’s why. Take a website about medication with content for doctors and consumers. These stats should be separated to provide recommendations for doctor content for doctors, and consumer content to consumers.

How to categorize content

Content recommendation is all about identifying the right content and classifying it to meet your goals. There are many ways to categorize content, including:

• Title words
• Keywords and tags
• The article’s word density.
• Categories.
• Author.
• Long vs. short articles.

Your use case may be affected by how the content is classified. You might not recommend long articles on your site to readers who prefer to read short articles.

Different types of content recommendation algorithms

You can use a variety of AI-based content recommendations algorithms to improve your website. These are the most popular.

Collaborative filteringrecommends contents based on similar user behavior. It analyses the history of users and recommends content users who have interacted with.

Content-based filtering suggests content similar to previously consumed content. It analyses the current page and suggests similar content based upon keywords, tags, and other relevant information.

Hybrid recommendation uses both content-based and collaborative filtering to give more precise and varied recommendations. To make better recommendations, it considers user preferences as well as the characteristics of the content to be viewed.

Popularity-based filtering suggests content based on its popularity. It will recommend the most popular content that has been shared, interacted with and viewed by many users. This tool is very powerful because it combines popularity-based filtering and other types, such as this content being most popular with people who have this job title.

Knowledge-based filtering recommends content according to user preferences and profiles. It uses feedback and user data to recommend content that matches the user’s interests. This includes ratings, reviews, and previous purchases.

Reinforcement learning recommends content according to the user’s actions, feedback and input. It uses feedback and user interactions to improve its recommendations.

Go deeper: The return on investment of recommendation engines for marketing

Selecting a content recommendation engine

It’s unlikely that every vendor will offer all these options, as we have already stated. Consider how you would like to use content recommendations to your site. Take into account your audience, your content, and the options available. Then, decide which method is most suitable for your situation. You can then send that list to potential vendors to help you find the right match.

Always put the reader first

It is important to put the reader’s needs first when creating content recommendation strategies that work. It is easy to get lost in the trap of thinking about how you want your reader to promote your business model.

Instead, think about the reader and create your content recommendation strategy around the things that will help them find the content they are looking for. This will make your business more profitable in the long-term. Engagement will be increased if you address the needs of the reader.

Increase website engagement using content recommendations was first published on MarTech.