i read 2,652 posts about distribution. here's what the numbers say
i scrape a lot. every day a daemon pulls reddit, hacker news, rss, youtube, github and x into a folder of markdown on my mac. as of this morning it holds 70,396 documents.
i went looking for distribution advice in it, because that is the part i am worst at. what i found is that almost nobody who writes about distribution shows you a number you can check, and the few who do disagree with each other in ways that are more useful than the agreement.
here is the filter, the survivors, and the part where the corpus lies to you.
the filter
start with everything, cut until only checkable things remain.
70,396 documents in the lake
2,652 mention distribution
166 carry a magnitude number (impressions, users, clicks, visitors, mrr)
82 are playbook shaped, meaning they claim a method and not just a result
73 also name a failure alongside the win
the third step is one grep:
[0-9][0-9,]{2,}( |,)?(impressions|users|clicks|visitors|views)|\$[0-9,]+ ?(mrr|arr|revenue|/mo)
two honest notes about that.
166 is a fact about my regex, not about the world. i dropped the |/mo alternation on one run and the count fell from 166 to 127. thirty nine posts appeared and disappeared based on nothing but whether i remembered that people write "$3k/mo" instead of "$3k mrr". any time someone shows you a funnel like this, the number is downstream of a decision they probably did not show you.
the last cut is me reading, not a grep. 73 down to the handful below is judgment. a post that implies the whole thing was mechanical would be doing the exact thing i am about to criticize.
also the counts move while you work. this started at 69,741 documents and 2,649 distribution mentions a few hours ago. the scraper does not stop because i am writing.
the reframe that made the rest make sense
the most useful thing i found was not a tactic. it was nevo david, who runs postiz, in a 22 minute video posted yesterday. postiz just crossed $2,000,000 in arr, which makes him the only person in this entire set operating at real scale.
his argument: saas is not dead, it got harder, because everyone can build software now. so the software stopped being the interesting part.
then he draws a line i had never seen drawn this cleanly:
- the problem is "i have trouble scheduling social media posts"
- the solution is "a platform that schedules your posts"
- the result is "having money, or getting views"
we are told to sell the problem. we mostly sell the solution. he argues you have to sell the result, and that everything else follows from getting that wrong.
that reframe is why his channel list splits the way it does.
what the numbers say is decaying
ads, unless your retention is already solved
nevo's math: if an ad costs $50 to acquire and the customer is worth $9 a month, you need them to stay a long time. with typical b2c churn you take, in his words, a minus in the bank on every purchase. the workaround is selling annual plans so the first payment covers acquisition.
the corpus has the retail version of that lesson. a dev building a life tracker called loggd spent €1,400 on meta and google ads. it produced about 150 signups and 2 paying customers on a $4.99 subscription. that is roughly €700 per paying user.
his conclusion is the same as the $2m founder's, arrived at from the opposite end: you need a funnel that converts before you buy traffic, not after.
seo, and the tension worth sitting with
this is where the corpus fights itself, and the fight is the interesting part.
nevo pushed hard on seo early at postiz. traffic went up and up. revenue did not follow. he built free tools that pull traffic that will not convert, and he now thinks traditional search is dying.
the small end agrees, harder. a founder building crawlmouse, which is an seo tool, did everything correctly for his own site: 12 blog posts, schema, clean technical setup, everything indexed. seven weeks in, his google organic clicks were literally zero. domain rating 0, no backlinks, ranking pages 4 through 9.
and someone running seven side projects exported 16 months of search console data into one table: 218,836 impressions, 1,584 clicks, a 0.72 percent click through rate. one project was 80 percent of all impressions. two of the seven produced zero clicks in 16 months. his read: "it is really one project plus six experiments."
so seo is dead. except it is not, because agensi, a marketplace for agent skills, went from 5 clicks per week to 900 clicks per week in 10 weeks off 86 articles, and watched impressions go from 300 a day to over 20,000.
what reconciles them is not effort. it is how old the niche is. agensi wrote things like "where are claude skills stored" for a category that barely existed, where nobody was competing. nevo is writing about social media scheduling against buffer and hootsuite. crawlmouse is writing about internal linking against every seo blog on earth.
seo is not a channel. it is a bet that there is an unoccupied question you can answer. when there is, it compounds. when there is not, you get 218,836 impressions and a thousand visits.
building in public, as an acquisition channel
nevo says this plainly and it stung a little, since it describes what i am doing right now:
even the biggest indie hackers building in public are getting hundreds of thousands of views, but if you are looking at their mrr, it is not growing that much.
he is not saying do not do it. he is saying stop counting it as customer acquisition. the corpus mostly backs him. the build-in-public posts with real numbers attached tend to have large view counts and small revenue lines sitting right next to each other.
what the numbers say is working
proof, published on a schedule
nevo posts his revenue publicly almost every day, with a stripe link and a trustmrr listing so it is checkable. the point is not the money. the point is that once people can verify one hard thing you said, they extend credit to everything else you say.
the part i did not expect: he explicitly says the proof does not have to be revenue. it can be traffic, or a public analytics dashboard, or any real number you are willing to keep publishing. the mechanic is verifiability, not size.
being genuinely useful on somebody else's platform
this is the one everything converges on, and the convergence is the finding.
the crawlmouse founder, at zero authority with one paying customer, started answering questions on reddit properly. that got him 25 users over a few weeks. then he left one comment in a "show me your project" thread that named his own failures, including how few customers he had, and woke up to 50+ free users in 14 hours, about a third of his all time volume. every polished pitch in that same thread got scrolled past.
his line is the best sentence in the whole corpus:
at zero authority, content is a savings account and communities are your paycheck.
at the other end, a founder running salesrobot at roughly $1m arr had linkedin working well: six people posting daily, 3.3 million impressions, one post with 3,000 comments that closed $2,000 mrr, contributing about 30 percent of the growth from $550k to $1m. he moved to reddit anyway, and said why:
linkedin posts don't show up when someone asks chatgpt "best linkedin automation tools." reddit threads do.
nevo calls the same thing aeo, and names the asymmetry that makes it work: the content does not have to live on your site. a reddit answer, a linkedin article, an x article, a listing in an "awesome" github repo. all of it is readable by the models, and all of it sits on a domain with authority you did not have to build.
a founder with one customer and a founder with $2m in arr landed on the same tactic for completely different stated reasons. that is the strongest signal in this post.
the mechanics, from the people doing it:
- account age matters. comment for a few weeks before you post anything.
- the first 30 minutes decide it. two posts published and abandoned got 600 views combined.
- the same piece across 5 subreddits got 2 of them removed. cap it at 3.
- niche down. "social media scheduler" competes with buffer. "best social media tool with agents" surfaces postiz.
x articles, the channel most people are not using
nevo calls this one of the biggest growth hacks available right now. x pushes long form because it keeps people on the platform, and the ranking signal is not what you think: saves are weighted over likes and comments, along with time spent reading.
his own comparison: a video of his got around 650,000 views with far more likes and comments than one of his articles. the article had fewer of both and many more saves, and did better by the thing x actually optimizes for.
the tactic he pays for: quote repost the article with a roughly 30 second video, so the long form and the short form amplify each other. he pays creators to quote repost with a paid tag. the version without a budget is reposting your own article repeatedly.
doing the unscalable thing first
my favorite entry in the whole set, because it is so small.
someone built a certificate maker for people running small online courses. the launch went the normal way: "i made a thing" posted in a couple of big founder subs, a few upvotes, zero signups.
what worked was embarrassing. he found three small communities, two discords and a niche subreddit, and when somebody mentioned dreading 200 certificates after a cohort, he just made them that night and sent the files back. free. about half of those people asked how he did it so fast, and that question was the opening.
40 paying users later, every single one traces back to doing the work for free first. not one came from a launch post.
the first users did not want to hear about my product, they wanted the problem gone tonight.
knowing which platform your thing works on
loggd, the one that lost €1,400 on ads, gets about 70 percent of all its users from threads. top post over 300,000 views. the same posts on x get 10 to 15 views.
meanwhile his reddit posts hit 80,000, 40,000 and 30,000 views with 200+ comments and produced under 50 users.
three platforms, the same content, three completely different outcomes. reach and fit are not the same thing and reach is the one that is easy to mistake for progress.
the metrics that actually route you
the trap running through the whole corpus is counting the thing that feels like progress.
impressions lie. in that 16 month dataset, the best click through rate in the set, 2.04 percent, belongs to the project with 833 total impressions. small numbers make ratios look good. the project with real reach had one of the worst rates.
views lie too. see loggd's 80,000 view reddit post and its fewer than 50 users.
the one metric i am stealing comes from indexerhub, which posted a 30 day breakdown: 1,766 visitors, $1,294.55 in revenue, $0.73 per visitor, at a 0.69 percent conversion rate. their framing:
when your revenue per visitor is healthy the job becomes distribution, not conversion.
that single number tells you which problem you actually have. i had not been computing it. i am now.
the three ways this corpus lies to you
i went in looking for playbooks. i came out with a filter for detecting sales copy, which might be more valuable.
one. the playbook that is an advertisement. one document in my lake reads as a complete, credible product hunt launch playbook. list four weeks early, form reciprocal upvote commitments, tuesday through thursday is the most competitive window. all reasonable. it closes with a call to action for the author's own launch coordination product. the advice is not wrong. it is a funnel, and it is shaped exactly like advice.
two. the honest breakdown with a plug inside it. that indexerhub post i just quoted, the one with the metric i am stealing, also routes readers to an analytics product partway through. both things are true at once. the numbers still look real. you just have to hold both.
three. elastic timelines. this is the one that changed how i read all of it. the agensi founder posted "8,000 active users in 8 weeks" on april 23rd. two weeks later, on may 7th, he posted "0 to 15k users in 8 weeks."
same founder. same "8 weeks." two weeks apart.
the growth is probably real. the denominator moved. and i only caught it because i had both posts sitting in the same folder, which is the only real advantage a corpus gives you over reading things one at a time.
worth noting the corpus knows this about itself. there is a post in there titled "most of the posts on this subreddit are masqueraded advertising." it was submitted twice, two weeks apart.
the eighth data point, which is mine
it would be dishonest to write all this and not check it against my own numbers, so here they are, live off my homepage as i write this:
2,098 followers on x. 3 newsletter subscribers. 5 posts, the first one 14 days ago. $19 in mrr.
that $19 is my own subscription to my own product. external revenue is $0. vibedraft peaked at $171 across 9 subscribers at the end of may and the last paying customer churned on july 20th.
so what does the corpus predict for me? almost exactly what i am seeing. a 14 day old publishing habit on a domain with no authority is a savings account, and it has not paid out. the corpus says my paycheck should be coming from communities, and i have not really been showing up in any.
the one thing i am doing right, i was doing by accident. my homepage publishes live stripe revenue, a live follower count and a live subscriber count, with a verification badge next to the product. that is nevo's authority mechanic, and he says the proof does not have to be money. mine currently proves i am not making any, which is at least a real number.
whether that converts, i genuinely do not know yet. ask me in a month. i would rather write "untested" than pretend.
what i actually took from this
- sell the result, not the software. the software is not scarce anymore.
- seo is a bet on an unoccupied question, not a channel. check whether the question is taken before you write 86 articles.
- at zero authority, go where the authority already is. reddit, other people's platforms, answer engines.
- do the unscalable thing first. it converts at a rate no launch post will ever match.
- compute revenue per visitor. it tells you whether you have a traffic problem or a product problem.
- when someone shows you a number, look for their other numbers.
and the caveat that undercuts everything above, which you should apply to this post too: every figure here is self reported by a founder with a reason to round up. i filtered for the ones who named a failure next to the win, because that is the only signal i have that someone is not selling me something. i cannot verify any of it. neither can you.
that is the actual state of distribution advice. the best i can do is show you the filter and the counts, and tell you where i am standing.
more next week. real numbers, win or loss.
ricky