Should you be using interest tagging on Facebook?


In our latest data deep-dive we analysed whether adding interest tags to organic posts on Facebook creates more traffic and higher click-through rates. We found that, contrary to what one might expect, adding interest tags has no effect on traffic and does not increase click-through rates, regardless of the quality of the tags. Social media users who are trying to reach a specific section of their followers should instead focus on posting their best content at the optimal time to their entire audience.

When Facebook replaced its old Interest Targeting feature with Audience Optimisation in 2016, there was plenty of speculation about whether the new feature would improve on the limitations of its unpopular predecessor. After the release of Preferred Audience Optimisation, Facebook announced to publishers that the new interest tags “do not limit reach.” Moreover, the social network, which generates well over 10% of all organic traffic to online publishers’ websites, promised that the use of the new interest tags would increase click-through rates.

Facebook generates over 10% of organic traffic for publishers

However, following initial excitement, no definitive independent study ever managed to establish whether adding interest tags to organic posts had any positive impact on traffic or engagement. Using extensive data from publishers’ Pages with millions of followers, we are now able to provide clear evidence that adding interest tags does not lead to an increase in click-through rates (CTRs), suggesting that audiences reached by carefully tagged posts are no more likely to click on a link share than audiences reached by posts without tags.

This analysis is based on data describing the social traffic of a number of major news publishers in the US, which serve diverse audiences. These publishers were Echobox clients and agreed to participate in the study. In order protect the data privacy of these publishers, we are unable to provide their names, but as in earlier Echobox studies, we carefully selected the participants to ensure representative results.


In order to conduct this investigation, we first trained a machine learning algorithm using natural language processing techniques that could assign relevant interest tags to social media posts. The algorithm did this by analysing and understanding the content of the share message and the article being shared. We built this in order to include tags in thousands of organic posts, and automating this process was the only way to do so quickly and consistently.

To ensure that the algorithm picked high-quality interest tags, we asked the Page owners for feedback on a randomly selected sample of tagged posts. All Page owners agreed that they would have picked similar tags to the ones chosen by the algorithm, and that the posts were of sufficient quality to be shared on their Pages.

Using this custom-built algorithm, we published thousands of interest-tagged organic posts to several different publications’ Facebook pages. To ensure a valid dataset not biased by trending news events or other factors, we alternated days. After each day of sharing tagged posts, we shared posts without tags for one day, before switching the interest-tagging back on.

After a full month, we compared the median impressions achieved by posts without interest tags to median impressions achieved by posts with interest tags. As in previous studies, we chose to use median values in order to avoid too much distortion from viral articles.

The results were both clear and consistent. None of the publishers’ Pages saw a statistically significant increase in CTRs when interest tags were added to their posts, suggesting that tagging did not achieve its main aim of directing posts to an audience more likely to be interested in the publisher’s content. One of the Pages even saw a statistically significant decline in its CTR, as impressions increased while traffic held steady.


Surprised by this result, which confounded our hypothesis that there should be an increase in CTRs and traffic for tagged posts, we hypothesised that our algorithm might not have generated high-quality interest tags.

We therefore decided that we should repeat our experiment, but this time with randomly generated tags. Using the same experimental design as before, we shared posts with random tags and posts without interest tags on alternating days. After 4 weeks, we compared median impressions, clicks and CTRs achieved by posts tagged with random interest tags to the figures for untagged posts.

The result showed that randomly tagged posts performed no different than untagged posts in terms of clicks and CTRs, except in one case where the CTR declined for randomly tagged posts. Like our first experiment, this again suggested that interest tags have no consistently beneficial or detrimental impact on organic traffic. We therefore conclude that preferred Audience Optimisation does not offer a viable way to reach specific subgroups within a Page’s audience with the content that is most relevant to them.


Our insights show that publishers should focus their time on crafting high-quality share messages rather than trying to reach a specific audience among their followers. Instead of adding interest tags, many of the publishers we work with create many different, topic-specific Pages on Facebook which they curate using intelligent automation. This allows users who are only interested in one sports team or only want to see economics-related content from a national news outlet to self-select into the content stream. Additional Pages contribute to traffic and reach high engagement rates by speaking directly to users with niche interests, who are likely to engage and appreciate customised content feeds.