i posted 1,728 times in 3 months. one post was 57% of my best day.
between april and july i posted 1,728 times. a scraper on my own account took 17,315 snapshots of how those posts performed.
i finally sat down with the data. some of it backed up what i believed. one number took the whole strategy apart.
TL;DR
- posting more does not cost you reach. from 6 posts a day to 30, i averaged about 170 impressions per post either way. no decay.
- but reach is a lottery. the top 1% of my posts earned 21% of all my impressions. the top 10% earned 55%.
- my best day was one post. 57% of that day's reach came from a single tweet.
- a link costs 41% of your reach. posts without a link averaged 183 impressions. posts with one averaged 108.
- 87% of my reach had no path to my product. only 340 of 1,728 posts contained a link at all.
first: volume actually works
i post a lot. i wanted to know if that was doing anything, or if X was quietly throttling me for flooding the timeline.
it is not. here is every tracked day, grouped by how much i posted:
| posts that day | days | impressions that day | impressions per post |
|---|---|---|---|
| 1 to 5 | 6 | 321 | 121 |
| 6 to 10 | 10 | 1,447 | 183 |
| 11 to 20 | 31 | 2,808 | 179 |
| 21 or more | 38 | 4,911 | 168 |
read the last column. from 6 posts a day up to 30, each post pulls roughly the same reach. there is no volume penalty. posting 30 times gets you about five times the reach of posting 6 times, because every post starts fresh.
so volume is real leverage, and it is the part a machine can genuinely do for you.
but posting a little is worse than posting a lot in a way i did not expect. on my quiet days, each post pulled 121 impressions. on busy days, 168 to 183. showing up rarely does not just get you less reach in total. it gets you less per post.
then: the number that broke my model
if volume drives reach, my best days should be my busiest days.
they are not.
| day | posts | impressions | per post |
|---|---|---|---|
| may 1 | 25 | 23,898 | 956 |
| may 27 | 36 | 20,383 | 566 |
| a typical busy day | 30 | 4,911 | 168 |
may 1 was not a heavy day. 25 posts is below my average. but each post did nearly six times its normal work.
so i looked at what was in those days. on may 27, one post was 11,590 impressions. that is 56.9% of the entire day. on may 1, the top post was 6,954, and the top two together were 54% of the day.
here is what those posts were:
I did it! from 0 -> $1. Around August 2025 I said I wanted to work...
I FINALLY MADE MY FIRST SALE!!!!! WOOOOOOO!!!
my two best days in three months were not strategy. they were the day i got my first dollar and the day i said so out loud.
you cannot schedule that. you cannot prompt an agent into having a first sale.
what that means for an automated marketer
i build a tool that posts for me, so this matters more to me than it might to you. the data changed what i think the automation is for.
the naive version of an agentic marketer is "write more posts, post them more often." my own numbers say that works, and also that it is not where the wins come from.
the wins are a power law. the top 1% of my posts carried 21% of my reach. the top 10% carried 55%. more than half of everything i earned came from a tenth of what i wrote.
i cannot pick that tenth in advance. neither can a model. i went back through my top posts looking for the pattern and the honest answer is that the biggest ones were true things that happened to me, written plainly, at the moment they happened.
so the job of the automation is not to produce good posts. it is to produce enough attempts that the lottery has somewhere to land, and to keep doing it on the weeks when nothing exciting is happening, which is most weeks.
that reframes the whole thing:
- volume is the machine's job. it does not get tired, and the data says volume does not decay.
- the milestone is your job. when something real happens, say it immediately and say it plainly. that is the post that outperforms a hundred scheduled ones.
the number nobody wants to look at
now the uncomfortable part.
only 340 of my 1,728 posts contained a link. here is where the reach landed:
| impressions | share | |
|---|---|---|
| posts with no link | 253,329 | 87.3% |
| posts with a link | 36,746 | 12.7% |
87% of everything i earned was on a post with no way to reach my product. not throttled, not wasted exactly, just pointed nowhere.
the 316 clicks those link posts produced work out to 0.86% of the impressions they got. that is an ordinary click rate. the click rate was never the problem.
the problem is the split. i spent three months getting good at the half of the job that does not convert.
and there is a real tension here, because the platform prices links:
| post type | count | avg impressions |
|---|---|---|
| no link | 1,388 | 183 |
| contains a link | 340 | 108 |
posts with a link averaged 41% less reach than posts without. i cannot prove from my own data that X is penalising the link rather than my link posts simply being worse posts, and it is worth being honest that both stories fit these numbers. but the practical answer is the same either way: do not put the link where it competes with the reach.
what i changed
five things, in order of how much they mattered.
1. links leave the main post. the post earns reach. the link goes in the first reply, where it costs nothing.
2. i write longer. this surprised me:
| length | posts | avg impressions | avg bookmarks |
|---|---|---|---|
| under 100 chars | 296 | 79 | 0.06 |
| 100 to 199 | 533 | 215 | 0.25 |
| 200 to 279 | 839 | 156 | 0.14 |
| 280 or more | 60 | 349 | 0.78 |
my longest posts pulled 349 impressions and thirteen times the bookmark rate of my shortest ones. i had been writing short because short felt punchy. short is just less.
3. the agent runs the floor, not the ceiling. its target is a consistent 15 to 30 posts a day so the lottery always has tickets in it. it is not trying to write a hit.
4. milestones get written by me, same day, unedited. those are the posts that actually move the number, and polish appears to be beside the point. "I FINALLY MADE MY FIRST SALE!!!!!" is not good copy. it was my second-best post in three months.
5. i stopped reporting impressions to myself. the metric on my dashboard is link clicks. it is a much smaller and much more honest number.
FAQ
does posting 30 times a day annoy your followers? the data says each post still pulls its normal reach at 30 a day, so the algorithm is not punishing it. whether humans mind is a different question, and mine is a small account, so treat the ceiling as unknown.
is 0.86% a bad click rate? no, it is ordinary. that is the point. the click rate was fine and the growth still was not there, because 87% of the reach was on posts with no link to click.
should i just wait for milestones then? no. the milestone posts only worked because there was an audience already there, built by the boring daily volume. the volume is what makes the lottery win worth anything.
what if i have nothing to announce? then you are in the normal state, which is most of the time. that is exactly the period to automate, so that when you do have something to announce, someone is listening.
the takeaway
i went looking for a growth strategy and found a division of labour.
the machine posts every day, because volume does not decay and consistency is not something i am good at. i write the three or four posts a year that actually matter, on the day they happen, badly and in a hurry.
and i stopped calling impressions growth. 290,075 of them bought 316 clicks. the number that matters is always the smaller one.
corrected 2026-08-18: this post first reported a 0.109% click rate, dividing 316 clicks by all 290,075 impressions. that denominator included 1,388 posts with no link in them, which cannot produce a click. the real rate on link-bearing posts is 0.86%. the corrected split is the more useful finding and the section above now reflects it.