Sunday, March 13, 2016

What kind of book post does Facebook "like"?

One benefit of blogging virtual footnotes to The Billionth Monkey and posting links on the book's Facebook page is that it gives me data about what posts have more "reach" than others. Here's what I discovered:

Performing well: 

Links to illustrated blog posts.
On the Billionth Monkey's Facebook page, links to blog posts that contain an image had an average reach of 500 people. This number was slightly correlated to likes/shares/comments (r=.44). This makes sense, since likes potentially mean the post shows up in other users' feeds. The size of the correlation, however, isn't huge, accounting for only 20% of variation in a post's reach.

Performing poorly:

Links to blog posts with no images.
I just couldn't find an image for a handful of posts, and these performed significantly worse than those for illustrated entries. These posts had an average reach of 123. This number may be exaggerated (skewed) by one post that performed much better than all the others. The median reach for these posts was just 41, compared to a median reach of 524 for illustrated blog posts.

Non-blog posts that self-tag.
In various book-related news and announcements on the Billionth Monkey Facebook page, I also tagged the page itself. These posts also performed poorly, with an average reach of 93. As with the previos category, one high-performing post skews the distribution. The median reach for these posts was just 11.

(NB: Self-tagging the Billionth Monkey page did not affect the reach of posts that linked to illustrated blog content.)

Posts using the word "sale."There was an ebook flash-sale in February to celebrate the Year of the Monkey. All posts about the sale had an average reach of only 21. This does not include one promoted post which also appeared at this time.

Links to
Occasionally I linked to in order to let people know where to purchase the book, or to draw attention to a new review. These posts had an average reach of 24.

Post type Average reach Median reach
Illustrated blog posts 500 524
Blog posts without images 123 41
Self-tagged announcements 93 11
Using the word "sale" 21 22
Links to Amazon 24 24

This suggests that posts linking to illustrated blog content had a reach over ten times greater than non-illustrated blog content, and roughly twenty times greater than commercial posts (i.e., which used the word "sale" or linked to a commercial website).

A few caveats

I'm only counting the 61 posts since late December 2015. The way Facebook calculates reach apparently changed significantly around Christmas. Prior to that, the "reach" reported by Facebook for my posts is consistently around 1/20th of what is reported after that date. Rather than confound these results with two different methods of calculating reach, I stopped collecting data prior this point.

Since the number of post "likes" seems to affect reach, it's difficult to know whether posts are low-performing because Facebook's algorithms limit the reach of certain types of content, or if poor performance is related to people just not being interested in certain types of posts. However, I found that reach is somewhat correlated with likes (r=.44), and the number of likes on all of my posts was pretty consistent. So I'm inclined to think that the substantially limited reach of these posts has less to do with the number of likes and more to do with Facebook's algorithms for who sees what in their news feed.

In conclusion, here's an image from Giphy to help more people see this post on Facebook: