Give the first arrivals something to react to rather than an empty page: a real submission, a specific open question, a decision not yet made. The first ten members change a community structurally rather than numerically — they make a thread possible without the founder in it, and they turn a vote from a formality into a result.
Flow diagram of 5 steps: Publish something with a position before you invite anyone, Ask ten specific people individually, and tell each one what to do, Give the first arrivals a decision that is genuinely still open, Reply to everything at first, then deliberately stop, Write the guidelines after the first real disagreement, not before.
The usual explanation for a community that never starts is that not enough people heard about it. That is occasionally true and mostly a distraction, because the failure happens to people who did hear about it and did turn up. Someone follows a link, reads two sentences of description, sees a page with nothing on it, and leaves — not because they disapproved, but because there was no action available to them. You cannot reply to nothing. You cannot vote on nothing. The only thing an empty community offers a visitor is the invitation to be the first person to talk into a room with no one in it, which is a large thing to ask of someone who arrived thirty seconds ago and owes the project nothing.
This is worth stating precisely, because it changes what you do about it. The problem is not awareness, so more promotion does not fix it — it accelerates it, by spending your one-time supply of goodwill and personal invitations on a page that converts none of them. The problem is that the page has no content for a visitor to act on, and the fix is to put something there before anyone is invited to look. A community with four members and a live argument about something specific holds a newcomer far better than one with four hundred registered accounts and an empty front page, because the first one shows a visitor what participating here would consist of and the second one only asserts that participation exists.
The first ten are not a tenth of a hundred. Each of the early ones unlocks something specific that was structurally impossible before them, and the unlocks are not evenly spaced — going from one member to three changes more than going from thirty to ninety does. It is worth knowing which threshold you are actually at, because the thing that is broken at four members is a different thing from what is broken at nine.
What becomes possible at each early size, and what is still broken
| Size | What becomes possible | What is still broken |
|---|---|---|
| 1 — the founder alone | A stated position, and something concrete for a visitor to agree or disagree with | No conversation exists. Every thread is one person talking, and it reads that way however well it is written |
| 2 to 3 | A real disagreement, visible in public, which is the first proof that this is a discussion rather than a newsletter | Two people arguing in an otherwise empty forum reads as a private exchange that a stranger would be interrupting |
| 4 to 6 | A thread that continues after the founder stops replying — the first evidence the place is not one person’s project | A quiet day is indistinguishable from abandonment, because there is no normal rate of activity to compare it against yet |
| 7 to 10 | A vote whose outcome is not obvious before it is held, and a second opinion on a submission the founder likes | The margin is one or two people, so the result records who happened to be online rather than what the community prefers |
This gap is where most founders lose their nerve. Suppose a third of members see a given thread on the day it goes up, and a third of those reply — fractions that are illustrative rather than measured, but not generous. Ten members then produce roughly one reply per thread, and some threads get none at all, purely from rounding. Nothing has gone wrong at that point. It looks exactly like failure, though, and it looks like failure for weeks, which is why the second month is where most attempts stop rather than the first.
Votes are supposed to decide things, which makes this the more serious version of the same problem. With ten members, the gap between first place and second sits comfortably inside the noise of who checked the site that evening, so announcing that the community chose an idea attributes a preference the community did not express. The honest handling is to publish what the numbers were and how few they were, treating an early vote as a straw poll that informs a decision rather than a mandate that settles one.
Coverage is the arithmetic people forget. A single-city membership leaves the community silent for eight hours, so a question posted late in the evening sits unanswered until morning and its author sees an unanswered question. Ten members across three continents fixes the silence and creates the opposite problem: no two are ever reading at once, so a live discussion never quite happens. Neither is wrong, but knowing which one you have stops you reading your own community’s normal behaviour as decline.
Almost every step below is commonly done in the wrong sequence, and the wrong sequence is what wastes the invitations. You get one first impression per person you personally ask, and there are not many of those.
Not a description of what the community will be about — an actual piece of work with a view in it that someone could disagree with. A visitor needs to see the kind of thing that happens here, and a mission statement does not show them that. This also forces a useful decision early: if you cannot produce one concrete example of the thing you want others to produce, the community does not have a subject yet, it has a category.
A broadcast post asking anyone interested to join converts almost nobody, because it assigns no task to any particular person. A message to one person saying you want their view on a specific submission, with the link, converts at a rate that looks implausible by comparison. It also does not scale, which is fine — this stage is not meant to scale, and treating it as a growth channel is how founders skip it.
Ask them to settle something you have not settled: which of two formats a first segment should take, whether a submission is too broad, what the rule should be for something with no rule yet. People stay where their contribution changed an outcome. Consulting them on a decision you have already made is transparent within about two replies and costs more trust than not asking at all.
For the opening weeks, answer every post, because an unanswered first post from a new member usually means no second post. Then stop on purpose, and let a thread sit without your reply in it. This is uncomfortable and it is the point: a community where the founder answers everything trains members to address the founder rather than each other, and that dependency gets harder to unwind the longer it runs.
Rules written in advance address imagined problems, are longer than anyone will read, and carry no authority because nothing has happened to justify them. Rules written the week after an actual argument are short, specific, obviously necessary to everyone who watched the argument, and enforceable for exactly that reason. A one-line rule everybody remembers beats a page of policy nobody has read.
Each of the following produces a number that goes up, which is why they are tempting and why they survive so long before anyone admits they are not working.
Followers and members are different things, and a launch that optimises for signups collects the first while reporting the second. Someone who joined because a post was interesting is not thereby a person who will reply to a submission, and a community of five hundred readers with four contributors behaves in every practical respect like a community of four — except that it is now much harder to see that, because the membership figure suggests otherwise. The measure worth watching early is how many distinct people posted something this week, which is usually an uncomfortably small number and is at least a true one.
Second accounts, replies written to your own posts, votes cast by people asked to vote rather than to read — all of it is visible sooner than founders expect, because the fake material has a texture the real material does not. The deeper cost is that it destroys the only instrument you have. Early activity is your one signal about whether anything here interests anyone, and once part of it is synthetic you cannot read any of it, so you carry on for months without knowing whether the idea works. On most platforms it also breaks the terms you agreed to, and on this one it is what the vote-integrity checks exist to catch.
Categories, tags, roles, a moderation ladder, a code of conduct, a weekly schedule — all of this is real work that feels like founding a community and can absorb a month. With nine members it is also entirely premature: twelve empty categories make the place look abandoned rather than organised, and a moderation hierarchy with no one to moderate is a diagram. Structure earns its keep when the volume it manages exists, and adding it later is easy in a way that removing it is not.
This assumes you are genuinely starting from nothing. Anyone launching a community into an existing audience — a mailing list, a following, a class, a workplace team — is solving a different problem and should ignore most of the above: the cold start is already paid for, and the real risk is that a hundred people arrive at once, nobody knows what the norms are, and the tone is set in the first day by whoever posts most. That situation needs the structure early, which is the opposite of the advice given here. Communities where participation is part of someone’s job are different again, because attendance is not evidence of interest and the quiet-week signal means nothing.
There is also a limit worth naming rather than dressing up. Kind Channel is new, its community is small, and no programme has aired yet, so nothing above is drawn from our own membership data — there is not enough of it to draw anything from, and any specific figure we published about early growth here would be invented. The arithmetic in this guide is arithmetic, offered with its assumptions stated so you can substitute your own; the rest is the ordinary observable behaviour of small forums, editorial groups and volunteer projects. Treat it as a practitioner’s account of how cold starts usually go, not as a measured result, and be sceptical of anyone starting from zero who offers you the latter.
There is no fixed number, but the meaningful threshold is not a headcount — it is the first thread that continues without the founder replying to it. Below that point the community is one person’s project with observers; above it, conversation happens between members. In practice that tends to become possible somewhere around four to six active participants, and active is the operative word.
Writing real content under your own name is not only acceptable, it is necessary — a visitor needs something to react to, and someone has to produce the first example. What is not acceptable is fabricating other participants: second accounts, self-replies presented as other people, or votes solicited from people who never read the submission. The first builds something to join; the second destroys the only signal you have about whether anyone is interested.
Most often because their first post went unanswered, or because they could not see what to do next. Both are fixable and neither is about interest levels. Answering every early post matters more than the answer’s quality, and giving new arrivals one specific open question beats a general invitation to contribute, which offers no obvious action to take.
Publish the counts, and publish how small they are. A four-to-three result decided by one person is real information about a straw poll and misleading if presented as what the community chose. Stating the raw number lets readers judge the weight themselves, and a community that hides small counts early has no credible way to start showing them later.