10 Do's & Don'ts of Great Content Recommendation Engines
Content recommendation was back in the news recently, for all the wrong reasons. A survey by Arkadium highlighted how much people dislike branded “content recommendation” widgets (the ones that tend to push ad links rather than, you know, recommend real content). More worryingly, that readers predominantly blame publishers for the negative experiences they have with them.
There is a way to do content recommendation so that it both boosts the bottom-line and makes your visitors happy, but it’s something that a lot of publishers struggle to do well, so here are some pointers to get you moving in the right direction.
Make Sure You’re Doing Real Content Recommendation
Real content recommendation -- suggesting relevant content from your site to your readers as they browse -- is a very powerful and simple way to get your audience to stay longer and engage more deeply with your content. This means happier users and increased revenue: publishers we’ve worked with have seen pageviews, dwell time, and bounce rate all improve on average about 10%. There aren’t many easy wins when it comes to improving your audience engagement, and a surprising number of publishers are missing out here.
Think Carefully Whether You Should Do it Yourself
There are two ways of doing content recommendation yourself. The old-school way is manual -- editors or authors find relevant content from the site and it gets added into the recommended content section of the page. This is time consuming, it’s hard to make sure you’ve got the best content (and there’re no algorithms to test and optimize for that) and when newer, better content gets added to the site, the recommendations don’t update.
The other way is to build an algorithmic system yourself. This is expensive, both to build and to maintain, and as the technology gets commoditized and is available as a service it makes less and less sense, unless your key differentiator is recommendation, like Spotify or Netflix. Make sure you’ve weighed up the pros and cons properly before you do it yourself.
Pay Attention to Your Page Design
One of the things we’ve observed is how much of an impact page design makes. Crowded pages; a lack of white space; not adapting the module to your site’s stylesheet: they will all make content recommendation widgets significantly less effective. Audiences want your site to look consistent and be structured in a way that helps them understand and navigate the content.
Think About How to Connect Your Content
Content recommendation isn’t just about improving your basic engagement metrics: done correctly it can also be a powerful way to drive traffic to your most valuable content. Do you have premium content that’s behind a paywall? Do you have certain content that generates a much higher RPM? Do you have sponsored or native content that you want to drive users to? Set up correctly, content recommendation can help users to find this content in a way that’s relevant and contextual.
Have Good Photos for Your Content if You Can
High quality images in the preview make a huge difference on how much engagement recommendations will get. Links without images are far less appealing, and poor quality images will also hurt performance. Not all content comes with a suitable image, but if you feature one it will definitely help.
Everyone these days is looking to follow the data, and content recommendations are a great place to do this -- our clients are often surprised by what proves most effective. The best way to find out is to A/B test your recommendation setup -- that’s not just the recommendations themselves (a good algorithmic system will do that anyway), but their place on the page (sidebar or bottom of the article?), what you’re optimizing for (recency, popularity, relevance?), and maybe even how many suggestions you offer (one row of four or two rows of three?)
Don’t Confuse “More from the Web” Links With Content Recommendation
Lots of third party widgets come with links to sites advertising insurance, skincare miracles and articles on what your favorite celebrities look like now, but don’t confuse these with genuine recommendations. We know some people need these links to pay the bills, but that’s not a reason to neglect real content recommendations that connect your site, improve your user experience, and also improve monetization of your own content. Featuring branded “more from the web” links doesn’t rule out doing real content recommendation too -- in fact they’re more powerful together.
Don’t Forget that Content Recommendation Can Help You Sell
There are lots of solutions for product recommendation, but as anyone who’s been on Amazon will tell you, even the best can get it wrong. When what you’re selling is “content” (e.g. a business intelligence report) or has a thorough description, content recommendation systems can be a great addition to your toolkit. They can often recommend products based on relevance when a normal system might only operate on e.g. popularity or customer similarity.
Don’t Make Assumptions About What Your Readers Want Without Data
One conundrum with content recommendation is how do you actually know what will make an engaging recommendation for any particular piece of content? Editors usually have a good understanding of their content and their audience, but an algorithmic system will test what’s working constantly. If you’re not relying on the algorithms to some degree here then you’re flying blind.
Don’t Expect that You Need to Fill Your Page with Widgets to Have a Positive Effect
There’s sometimes the perception that if one revenue generating widget makes you X, then ten widgets will make you 10X. There’s increasing evidence that that’s not true, and it’s certainly something that we find with clients. A page that’s stuffed with elements is not only slower to load, which you should watch out for, but also the individual elements can become dramatically less effective. Diminishing returns sets in pretty quickly.
Mads Holmen is the Danish founder of Bibblio, a pubtech/mediatech startup which raised $1.6 million in its seed round last year. It helped digital media increase RPM and audience engagement with high-quality, ad-free internal recommendations driven by AI. Prior to starting Bibblio, he was the Planning Director at AOL Europe.