I spent a month rebuilding my posting system and logging every post. I also removed large batches of obvious bots daily. The common wisdom (and the X algo code) says that hurts reach.
My logs show the opposite: profile visits per 1k views and average views per post trended up while follower count went down.
Here’s the system, the receipts, and what I’ll test next.
1) What I changed (week‑by‑week)
One daily “hero” post in my best hour (19–21 UTC), link‑free. Quotes the previous day’s hero to chain the posts and pass views along. Use Typefully to find out your best posting time, or just use a free logger like mine (open sourcing it when we hit 100 subs here!). This is what mine looks like:
When possible, two replies under it: Checklist and later Proof, encouraging bookmarking and dwell.
2–3 quote tweets that each teach something new and point back to the hero.
Daily clean‑up: remove obvious bots at human pace with a small browser helper (uses X’s own “Remove follower”).
Tracking by fetching all the info I can via X’s API:
2) The surprise
Popular belief (and some code reading) says “declining account” signals will punish you if your follower number drops quickly.
My data says otherwise: While I removed thousands of dead followers, my average reach rose and my Profile visits / 1k views improved.
Likely reasons:
A cleaner follower graph routes you to real people faster.
Removals via the native flow, at human pace, don’t look like spam.
Posts stayed tight: one screen, payoff first line, proof later.
3) Some data screenshots
4) What actually moved the needle
Originals on “follow days” → more profile visits per 1k (better for follows).
Quotes on “reach days” → higher median views (better for discovery).
One screen, one idea → higher bookmarks and more early likes.
Clean‑up → fewer dead edges, better early routing.
5) What I got wrong
I expected pruning to hurt. It didn’t.
I overvalued raw likes; Profile visits / 1k and Bookmarks / 1k were better early health signals.
6) What I’m testing next
Whether clean‑up changes who likes early (bridges).
Whether public Lists amplify the routing effect.
How much dwell formats (Before/After, If → Then → Because) compound the ladder.
I’ll keep publishing receipts in Shadowban Chronicles and ship the simple logger tool when we hit 100 subs.