Is This Post Optimized For Engagement?
I've always been a fan of data in interesting contexts.<1> When much of my leisurely reading material is content from Wonkblog and FiveThirtyEight, I can't deny that their work partly inspired my move into the digital analytics space.
With so many potential topics to cover in this field, it was a struggle to decide the focus of my premiere blog post at Ambition Data LLC. The stakes seem much higher when you're writing professionally versus, say, writing a Facebook post where the worst thing can happen is someone unfriending you.
In an effort to make the most out of my first post, I looked for analyses that determined optimal or standard metrics for blog posts, specifically article title length and word total. A light bulb went off: writing the most optimized blog post about a blog post is so "meta", it would be the perfect topic for me!
Supposed "optimal lengths" have been measured for various social media platforms, all of which are covered in a thorough infographic by Kevan Lee. Whether it is Twitter, Facebook, YouTube, or even the forlorn Google Plus, the length (or brevity) of your titles and content correlates to the amount of audience engagement.
Some of these findings are backed by technical nuances (e.g., when your title gets too long it gets cut off by the platform), and some are psychological or behavioral (e.g., people only read the first and last 3 words of a title). Overall it's a fun read, though I think much of the presentation is either not as well visualized as the source material, or missing the real grit of the topic: why does the data show these results, and how can we test it?
In addition to being highly relevant to this post, tracking reader engagement by word count in blog posts was the most fascinating of all the topics covered to me. The original data analysis came from Medium, a user-driven, story-writing website.<2> The article thoroughly covers the analysis Medium reader engagement using pretty standard web metrics.
via Medium
I do applaud their aversion to using page views as a solely key performance index (KPI), both on an analytics and business level. The analysis also weaved the time spent and post length to the engagement calculation and found that 7-minute-long posts yield the most engagement (average total seconds spent reading per post). It is an interesting find to be sure, though ultimately specific to Medium users. I have a few observations and criticisms, all of which should be taken with a grain of salt:
Where are the error bars? Eye-balling the data, it looks like their local maxima of 7 minutes is surrounded by a fairly volatile engagement range. If I was in finance or science/engineering, 4-5 minute post length appears a lot less variable/risky for only about 10% to 20% loss in engagement (do note the log scale).
Time spent isn't necessarily engagement, as the metric should be combined with scroll depth to determine those who actually finished reading the entire post, versus the scenario of readers getting distracted or busy and eventually closing the tab.
Related to scroll depth, bounce rate (or disengagement) by post length may be an important factor to analyze, especially when the volatility of engagement is high for a particular range. Using an average time spent effectively lumps everyone in the same category, and I imagine Medium has a very diverse audience with varying interests and reading behaviors, including speed readers!
On that note, this data could be viewed from a different angle, specifically the customer/consumer treatment. Combined with scroll depth, Medium could categorize their readers by post length and engagement (scroll depth + time spent), then integrate content consumption to gain perspective on different groupings of readers and their behavior: this could lead to better ad targeting, assuming that is a primary source of revenue for Medium.
In the end, any single optimized finding is not a rule, theory, or law, especially in a rich landscape characterized by human behavior, trends, and imagination. Medium concludes that their findings are "just math" and I agree for the most part: data shouldn't dictate absolute limits of creativity. Otherwise you may end up with A/B testing to no end, likely at the loss of unique and groundbreaking products and content. However, as soft guidelines for writing articles, it's an informative slice of data for new and veteran writers alike.
Oh, and for those who are still curious and made it this far: the blog title is optimize at 6 words, and I aimed for about 800 words since the gains in reader engagement past 4 minutes of reading start getting volatile. I'll likely treat future blog posts much like Youtube videos, where 3 minute are considered optimized. Keep in mind I haven't even researched the optimal number of visuals, hyperlinks, and footnotes to include in each post.
Just kidding.
(Sort of.)
<1>: Evident by my Ph.D. in Physical Chemistry, but I'm not here to boast my time in a dark basement, shooting lasers at glass.
<2>: Medium has an interesting business model for a blog website. It doesn't need to produce much of its own content due to its user base, similar to how Tumblr, Yelp, and Facebook operate. I am interested to explore how they monetize and leverage their user base to increase revenue.