Coalhouse Forge makes garden tools by hand in Carmarthen and has 11,400 followers. In the first half of 2026 it posted thirty times. The median post was seen by 640 people. One post was seen by 214,000, and nine out of ten of those people had never heard of the forge before that morning.
Both of those numbers came off the same account in the same six months, and the distance between them is the thing this topic is actually about. The follower count did not produce either of them.
The unit is the post, not the account
Every major platform now decides what to show from what a piece of content does in its first hour, rather than from who has subscribed to it. A five year old account and a two week old account start each post from roughly the same place, and the follower graph is consulted only for the first small batch of people who get shown anything at all.
That changes what a follower is. A follower is permission to be considered, not a delivery. Coalhouse's 11,400 are the reason the first two hundred views happen quickly, and after that the post is competing against everything else on its own merits. This is why an account can grow for three years and stay invisible, and why a first post can reach a quarter of a million people.
Which number to keep
The mean of those thirty posts is 8,200 views. The median is 640. One of those describes what happens when Coalhouse posts something and the other describes what happened once in March. Report the mean and the account looks like a going concern; report the median and it looks like what it is, which is a small workshop that got lucky in the spring.
Take the median for anything you intend to plan with. Keep the outlier separately, because it is a different kind of event and it deserves its own question, which is whether anything about it can be repeated on purpose. At Coalhouse the answer so far is no, and four attempts to repeat it landed at 310, 890, 1,100 and 480.
What 214,000 views were actually worth
The forge sells a garden knife at £64. The best post of 2026 produced 88 clicks to the site and eleven orders. That is £704 of revenue from a post seen by 214,000 people, and it is not a disappointing result, it is what this channel is: the number at the top of a social funnel is an audience measurement and the number at the bottom is a business one, and they are separated by four steps that each remove most of what came into them.
The collapse everybody blames on short attention spans sits at the top of the video, and on this post it is not where the audience went. Of the 214,000, around 61,000 watched past three seconds and 9,800 stayed to the end, which is nine thousand people who chose to watch a blacksmith finish a knife. Eighty eight of them then went looking for the price of that knife. The loss worth arguing about is the one between the end of the video and the shop, and reading it properly is the whole of what to count when reach is not sales.
The first hour decides the rest
A post has a life and it is short at the front. Distribution is decided in the first hour, from what the first few hundred people did, and everything after that is the platform acting on a judgement it has already made. A post that was going to travel is usually visibly travelling before lunch, and a post that was not cannot be rescued by sharing it again in the evening.
The practical consequence is that posting time matters for a reason nobody explains properly: it is not that a magic hour exists, it is that the post needs its first few hundred viewers to be people likely to watch it to the end, and those people are not awake at the same time everywhere. What to do about that is the routine in the routine that makes an account work.
The life of a post
Four states, and the first one is over before lunch
Frames rather than steps: these are things that happen to a post, not tasks anybody performs on it.
- 01The first few hundredDrawn mostly from followers. The only moment the follower count does any work.
- 02The first hourWatch time and replies are read, and the size of the audience is set.
- 03Days two to fiveThe judgement already made, being carried out. Nothing done now reverses it.
- 04The long tail, or noneSearch and recommendation keep a few posts alive for months. Most stop dead.
It does not say which posts get a tail. Nobody outside the platform knows, and at this workshop four deliberate attempts to repeat one landed between 310 and 1,100 views.
Where this stalls: the schedule that cannot be stopped
The stall in social is not running out of ideas. It is that posting becomes an obligation with a frequency attached, and once it has one, the quality of what goes out falls to whatever can be produced on that frequency. The platform reads a run of thin posts as a reason to show the account less, which makes the next post land worse, which is read as needing to post more.
Coalhouse went to daily posting in January 2026 and the median view count fell for eleven straight weeks. Stopping felt like giving up and it was the only correct move available. The signal that this has happened is easy to check and nobody checks it: compare the median of the last twenty posts against the median of the twenty before them, and if it is falling while the count is rising, the schedule is the problem.
What to open in 2026
Every platform's own analytics is free, better than any third party tool at describing that platform, and useless for comparing across them; that is not an accident. Meta Business Suite covers Facebook and Instagram together and replaced the separate page insights screens that used to live on each. Buffer schedules across several platforms and replaced sitting with a phone at the chosen hour, which is what it is worth paying for and the only thing.
The honest position on paid tools here is that below about ten thousand followers there is nothing to buy. The free floor does everything, and a scheduler is a convenience rather than a capability. What is genuinely new since 2024 is that every platform now pushes the same short vertical video format, which means one production can feed several of them, and that is a real change in the arithmetic of the format every platform now pushes.
Five clusters, and the one that is not broadcasting
Start at what each one is for, and who is there, because choosing one and refusing the other four is the decision that makes the rest possible for a workshop with two people in it. Then the format, then the routine.
The part that is not broadcasting is the cluster most likely to be skipped and the one where a small account has an advantage a large one cannot buy: replying to forty people is a Tuesday for Coalhouse and an impossibility for a national brand. Measuring comes last because it is only meaningful once there is a routine to measure.
Sideways, the material being posted is usually made once and cut up, which is writing the things that get found and read, and the reason to do any of this rather than buy the attention outright is the comparison at buying attention, and not overpaying for it. The place to move the audience once it exists is the channel you own, and nobody can take away. All four sit under getting what you make in front of the people who want it.
Are 11,400 followers worth anything to this workshop now. They cost three years to collect, they no longer decide distribution, and the only test anybody at the forge has proposed is to stop posting for a month and watch what the shop takes. Nobody has been willing to run it.