Back in January, Facebook announced a new tool called Preferred Audience, which is designed to optimize and target organic posts to specific audiences. Unlike its predecessor, Interest Targeting —which did not show posts to profiles that did not meet the targeting criteria —Preferred Audience allows posts to appear anywhere on Facebook while prioritizing the information for people who meet the selected audience specifications. In Facebook’s words:
Interest tags help Facebook better match content with audiences, prioritizing posts on particular topics for the users who are most likely to be interested in those topics.
For more information about the tool, Simply Measured has conveniently provided more details on it and how it can be utilized.
Given the functionality of Interest Targeting and assuming a similarity in performance, Preferred Audience was something we largely overlooked at Harvard Medical School. However, after learning about how the social team at Harvard was using the tool, we decided to test it on half of our posts for the month of September.
Now that we have the preliminary data analyzed, here is how we set up the program, and what we found by comparing like posts after removing statistical anomalies.
Use 6-10 tags. You can use up to 16 topic and interest tags, but Facebook suggests using 6 to 10. This advice is quite sound, as we found engagement rates drop severely after roughly 6 topics — although it is always good to have a couple more for testing purposes.
Use a mix of broad and specific targeting. In our tests, we applied a combination of large broad topics and then interests that were specific to the content being posted. This combination will allow you to understand the broad construct of who engages with your posts, while also determining what works best on a topical level.
Facebook also allows you to limit the audience by age, language, and location, which wasn’t applicable to our use case.
Reach stayed the same. Regardless of being targeted or not, the reach of articles remained roughly consistent, with non-targeted posts averaging a reach of 14,621 and targeted posts a reach of 14,587 (a 0.23% difference). Impressions followed a similar trend, with non-targeted posts averaging 24,290 impressions and targeted posts averaging 24,394 (a 0.43% difference).
Targeting helps engagement (slightly). Engagement (without link clicks) on posts that were targeted averaged 250, while non-targeted posts averaged 239 engagements (a 4.6% difference). More importantly, the engagement rates differed by 3.03% (1.02% compared to 0.99%). The differences are rather negligible, but they offer a positive trend that may be improved with further refinement of our targeting.
Link clicks declined. Despite the increase in engagement, we did see a decrease in link clicks from an average of 347 on non-targeted posts to 312 on targeted posts (-11.22%).
We learned about our audiences. By reviewing the like rate on posts, we were able to see which targeting buckets our engaged audiences typically fall within, while also understanding the specific interests topic-level audiences have. We can also come to a better understanding of whom our content is actually reaching.
Furthermore, we’ve been able to drill down into how specific audiences engage with different types of content by looking at things like the link click rate, share rate, and Like rate.
Increased page level audience optimization. If your Page has more than 5,000 Likes, the Audience Optimization feature is automatically turned on, and with the understanding of the broad interests of our audiences, we’ve been able to strategically utilize this feature.
Does continued refinement and testing show changes? As shown, the data has shown a slight difference in post performance between targeted and non-targeted posts. It will be important to understand if refinement shows any positive returns.
How does our content work with our target audiences? So far, we’ve examined who is engaging with and consuming our content from the macro level and topical micro levels to obtain base level data. However, the next step would be to review and understand if and how our content is reaching our pre-determined target audiences (and how they interact with our content).
Is there a difference between engagement rates of relevant content vs. paid content? Another next step is to compare how consumption rates compare between targeted, relevant content and paid content. By understanding this difference, we can better utilize our paid social.
Is there a difference in real and quality traffic to the HMS site? We need to add proper UTM tracking to links to compare the difference in real traffic to the HMS site beyond Facebook data and how these two cohorts act once on the HMS site.
Does page level optimization make a difference? Now that we’ve implemented page level audience preferences, we will need to understand what impact this change is having on the growth of our page likes, the quality of these likes, and the retention level of them.
We’ve only been experimenting with the Preferred Audience for a short while, so more data and further studying is necessary to understand its efficacy, but the initial results have provided some encouraging data for discovering more about the Harvard Medical School Facebook audience.