After Twitter open-sourced part of their internal source code, I finally learned what may have been plaguing my Twitter Art account (I post my new art there almost every day). A year ago, I averaged about 3,000 impressions per post, adding 700 or so monthly followers. Then, in mid-year or so, with over 5,000 followers, my impressions dried up to an average of 300. After that, if I was lucky and got a couple of early retweets, I might get to 500. Otherwise, Twitter started showing my content to fewer people, including most followers.
After scanning through the Twitter source code, I found a calculation of "Mass," which penalized the account if it had more than 500 followers and followed 60% of that number or more. I reciprocated the people who followed me, so I easily had too many. I still don't know how this translates into how often Twitter shows my content (the source-code drop is incomplete, and there is no Git history to see when and why this was included). There might be additional calculations behind the massive drop in impressions. My account should have had consistent impressions since (1) most posts had media (art), (2) I posted consistently (daily), and (3) likes were consistent, so the content was popular enough.
The solution for the imbalance of follows versus followers seemed a relatively easy use for writing code since I had to eliminate almost 2,000 people I followed. Looking up the documentation, I needed to make an API call to delete a follow, which was shown to be rate-limited to 50 per 15 minutes, which was slow but not a deal-breaker.
Sadly I would have to pay $100 to get the API for a month since Twitter had wholly changed the API terms and price simultaneously, sunsetting all old plans. At 200 per hour, it would take 10 hours, long but not terrible, assuming everything worked correctly. I hated to pay money to fix something Twitter should have documented long ago. Still, I can't even imagine how many of my followers have never even seen anything I posted after they followed me. This has been frustrating me for at least ten months now. So I figured I could at least improve my "Mass" score, and maybe that would help.
What a waste of money that was.
It turns out that the documentation had never been updated with the new rate limits that rolled out with the revised API service. A single page listed them, but I never saw it since I was looking at the API documentation and other notes that still referenced the old API. The actual rate limit for the call was now a glacial 20 calls per hour! Despite paying $100 for what used to be free, it would take 100 hours to complete, assuming no issues. Ensuring the code worked correctly kept running into the limits, necessitating frequent halts. I finally gave up.
I found I could unfollow people by hand with four keypresses (tab-tab-return-return). Today I finally got the number below 60% (assuming it hasn't been changed again). Still, I have no idea how frequently "Mass" is recalculated, and I also do not know for what else Twitter might be penalizing me.
Looking at the other rate limits, the $100 per month "Basic" plan is primarily useless since the rate limits are per-user. The "Free" plan only allows posting (handy for bots, I guess). All other plans cost a fortune, and I am unsure if the same limits apply. This pointless API will turn people back to scraping content from the Twitter website (which is why people offer APIs in the first place).
Fellow artists mainly follow my Art account; I don't use Twitter for anything else besides posting links to new Codist posts.
So I spent $100 to get precisely nothing. Still, it's cheaper than spending $44 Billion buying Twitter!