Background
I am interested in looking at how the changes to publishing stories, specifically how imposing that collection editors approve submissions might affect writers. Collections have a few interesting metrics that may impact writers, namely: number of followers, number of posts in the collection, and when it was last updated — and I wanted to see how these vary across a non-trivial number of collections.
I wrote a small piece of code that mimics how writers would search for collections to host their posts. I collected data from ~1,000 Medium collections by searching for ~100 nonrandom keywords.
A few caveats: First, as I mentioned above these results are not from a truly random sample. Second, the findings pertain only to English posts. Results might differ across languages. Third, the “Last update” data provided only tells us when the last article in the collection was added, not how frequently a collection is updated. The update frequency information can be inferred by crawling each collection, however that would abuse the Medium servers since they don’t have a public API. Finally, there are loads of interesting questions to study and data to gather on Medium, but I would wait until a public API is available.
Press enter or click to view image in full size
Collection statistics: Followers and posts
Of the 1,000 collections analyzed, thirty-five percent had either just one follower or none at all. Ninety percent had fewer than fifty-five followers. This means that, in the current submission format where you must randomly search for collections, the odds are such that you will mostly find collections with a very low number of followers.
Like most user-generated content on the web, Medium collections follow a power-law distribution where a few items have a large number of followers, while most have virtually none. This was true for both the number of followers and the number of posts in each collection.
The collections that had the most followers also tend to have the most posts, therefore publishing there is not only competitive, it is also unclear how many subscribers would even see your post. The best alternative is to find collections with a lot of followers, yet a low number of posts. Such collections are highlighted in the green rectangle in the figure below.
What the figure shows is how a collection’s number of followers and the number of posts relate to each other. The middle panel with dots (known as a scatter plot) shows the number of followers and posts that each collection analyzed has. The bar charts (known as histograms) show how many collections have a certain number of followers (the horizontal histogram) or a certain number of posts (the vertical histogram). You may notice that the range of the data is between (0-14), this is because the original data has a few collections with a huge number of followers or posts, while the overwhelming rest had barely any. Such datasets are known as skewed. To conserve the relative difference in the number of posts/followers, while being able to see what is happening in the smaller ranges, we take the logarithm of the data. That is what’s displayed throughout this analysis.
What both histograms tell you is that nearly 90% of collections have a tiny amount of followers and posts (between 0-50), while the high follower/posts collections are a minority (~2%). The red shaded area in the scatter plot is the set of collections that have both a high follower and high post count. Those collections might be very competitive to publish in. However, the collections with high followers and low posts are highlighted in green. Those collections could be a great place for you to (manually) submit your work.
Collection statistics: Activity
Let’s assume you find a few collections that are relevant to your post and have enough readers to give it a chance to be widely read. Now you need to submit your writing to each collection. Once you’ve submitted your story, the waiting begins. There is no way for you to conveniently know what’s the status of your submission, and you can’t even keep track of which collections rejected you or not.
Unfortunately, it seems that you might get stuck in limbo forever! The average collection was last updated eighty-five days ago! About thirty percent of collections got updated within a few days. Nearly fifty percent of collections haven’t been updated in a month, and twenty percent of collections haven’t been updated in a year.
Summary of analysis
So what does the current collections model look like? Most collections have low followers and low activity (upper left corner of figure below). A few collections have many followers and are active (lower right corner). However, in addition to finding collections that have high follower / low posts, you need to identify collections that are updated regularly. It seems right now Medium has worked very hard to create collections that make it worthwhile to submit to (Editor’s Picks, Design + Startups, On Startups, Design/UX, etc.) However many are artificially engineered given that (i) the editors are Medium employees and are paid to keep Medium going and (ii) they had a head start since they were the only collections around when Medium was closed to the general public.