Medium Earnings Analysis: Month Two
Using stats to figure out how Medium pays its writers
In part one, I analysed one month’s worth of data from writing almost daily articles. Whilst it was interesting, one month’s worth of data is not enough to make any strong conclusions. It has, however, paved the way for some changes in tactics, and some very cool additions to the analysis for this month.
Here, I’m recording stats about viewership, readership and income and trying to figure out what makes Medium tick.
Read part one if you haven't yet:
Are we there yet?
I previously mentioned a set of “total earnings milestones” that I wanted to hit to make it seem like progress is being made when I slowly step from one to the other. After the first month, I had reached $50 and was sitting at $54. After the second month, I’ve tripled that figure, earning around $100 in the second month.
Annoyingly, the steps between a “10” and a “25” always take the longest, so I’m stuck on 100 for a while.
Also, that $10,000 is looking a very long way off.
Cold, hard, unyielding data
Earnings have not been the only metric I’ve been tracking. To understand how Medium’s algorithm works, I wanted to find out what is most important. What bears more weight? Views? Reads? Claps? Average views/reads over 30 days?
The “Averages” stats proved to be pretty useless halfway through this second month, so they’ve been scrapped. I thought they might have some influence, but they produced only noise. Here’s an update of what my main spreadsheet looks like now.
Notice the three new fields to the right.
I wondered if the percentage of viewers that read the articles contributed to the income. It makes sense if you think about it. Clickbait titles work to get eyes on your articles, but a shoddy article will cause those viewers to click away before finishing. That’s not what Medium wants. They want titles that lure you in and content that keeps you hooked.
So instead of the averages, I’ve added Daily Read Percentage, Total Read Percentage and Daily Read Points (Daily Read Percentage / 10)
Avoid the top of the bell-curve, everyone there is mean
After the first month, the relationships between views and income, and reads and income were positive, but the data was too limited to conclude exactly how correlated each one was. Now we’re another month in, these graphs look much nicer.
In part one I suggested that views were more tightly correlated to income. After another month of data, it seems that reads are providing a tighter correlation with fewer outliers. This would suggest that quality articles are far more important than clickbait titles aimed to get clicks rather than readers. No one likes clicking on a great title only to find the actual article is rubbish.
This feeds into a really important trend I noticed whilst I was writing. After the first month where I tried to consistently pump out at least one story per day, I lost a bit of motivation. I started writing less frequently and more thoroughly. I was focussing on the things I wanted to discuss at length, rather than churning out small, less detailed articles. Strangely, I noticed that revenue for these sporadically published articles far outweighed revenue for smaller, more frequent articles.
Now, that could be a false equivocation since I now have more followers than month one. However, it does support the conclusion that reads are more important since the content is uncountably better in my longer form, less frequent articles.
My best performing article monetarily is not my most viewed article, but it does have a great read percentage:
Read percentages
If the assumption that reads are more important than views is true, then we should see a correlation between reading percentage, and income. It won’t be linear since I could release an article and have 5 views, 5 reads. That’s 100% reading percentage but for only 5 people. That’s not going to be very profitable.
But, if one article has 500 views and 200 reads, and another has 500 views and 300 reads, we should see the latter be more highly rated and receive higher income right?
This graph shows us that this assumption is likely correct. It’s not a tight dataset because some articles get 100 views and some get 1,000. 10% of 1,000 is still more than 90% of 100, so the graph will only ever be clearer if the data that fed the graph was confined between limits (for example only days which receive between 400 and 600 views).
More graphs, please!
I’m still not sure about this graph. It’s a daily representation of reads, views and income day-to-day. In the previous article, it was a bit of an afterthought, but I’ve added a metric I think important to the relationship between the three.
The blue, red and yellow lines were present in the previous article, the green line is new. This represents the Daily Read Points, calculated by dividing the daily read percentage by 10 so it could fit on the right y-axis without messing with the visualization.
The reason I added it is because of the new assumption that read percentage affects income. If the income was related directly to views, the yellow line would peak every time the blue line peaked. If income was directly related to reads, the yellow would peak every time the red peaked. However, that’s not the case. It’s sometimes hard to see why the yellow income line peaks and troughs so severely when the other metrics don’t seem to. That’s where the read percentage comes in in green.
Unexplained troughs in the yellow line are explained by troughs in the green line, where blue and red seem to not move drastically at all.
This proves that read percentage does contribute to the income calculations.
This is an important conclusion to make because it steers the way writers publish stories. Clickbait titles are only good if you can back them up with an extremely engaging article, that keeps the reader interested. Not only that but in some parts of this graph, it seems as though a bad read percentage is punished. So if your title is too clickbaity and the content is shoddy, income drops sharply.
Takeaways
The best tactic is to write long, engaging content with an article which draws people in. I write mostly about Blockchain Development, but I’ve noticed that general Blockchain and Cryptocurrency articles are doing better in this respect, because people find the price predictions of Bitcoin more engaging than how to load a wallet address in Javascript, for example. Not only that, but it’s much easier to write a clickbaity title on the former than it is for the latter.
I’m hitting the milestones one by one but I can’t help but wish the progress was a little bit more parabolic. The graph above shows the steady increase over time, but to make real progress I want this to start ticking upwards more sharply.
Some exciting days over the past month saw one day earn me almost $10 in one day, by far the largest I’ve had to date. Even the day after brought in $6 despite me not publishing anything new, which shows the power of a really good article.
Also, I had an article reach the second page of hacker news, which drove a lot of clicks, but not a fantastic amount of reads:
Progress is slow, but it is without a doubt steady. This game is like any other in life, it’s infinite. There’s no end, there’s no winner, it just goes on forever. Targets and milestones fall a little short in that respect, they’re finite measurements for an infinite game. I might one day get to a total of $10,000 earned from this platform. Will that be the end? Of course not.
Further Reading
If you’re interested in Blockchain Development, I write tutorials, walkthroughs, hints, and tips on how to get started and build a portfolio. Check out some of these resources:
