Every community that actually works has a room that makes things happen. In the best one I ever helped lead, that room was a WhatsApp group.
It held thirty-eight people at its peak. The number moved around: new people came, a few got removed - we learned some lessons the hard way. They were the group the community trusted to move it, and move it they did. The official community wasn't dead. The forums were bustling, packed with hopefuls asking questions and building reputations, and every public metric looked healthy. But the forum was the stage; the group chat was the real engine room.
I got lucky. They invited me in, and they let me help write the code that ran the place: nobody joined without a nomination. Not everyone had to vote yes, but no one could vote no. A single veto and it didn't happen. We had plenty of disagreements. We handled them, sometimes badly, but got better at it in time.
The best part of the code was that the door stayed open. Hopefuls from those bustling forums got nominated in, which was always the intention. We grew it and diversified it until the group spanned multiple countries, cultures, religions, and races.
Between 2017 and 2021, our corner of the MVP program (the recognition track everyone in that ecosystem was chasing) went from roughly ten people to 250.
Thirty-eight people in a chat decided it should happen and then coordinated to make it so. Somebody would spot a name in the forums, somebody with promise and gumption, and would drop it in the room. Then the machine would start: nominations written, evidence gathered. From the forums, it looked like the program was finally waking up and noticing people. It was our room, running the whole time.
That's the version of the story I bring up at conferences. The other version starts back at "we handled them, sometimes badly."
We removed three or four people over the years. A couple just went quiet and drifted off. One got removed because he would not stop hammering the room with politics and religion after being asked, more than once, to give it a rest. And two got removed for having a problem with me. A husband and wife, both good contributors, and neither of them did a single thing to the group. Their beef was personal and it was with me, but the room decided that keeping the peace and keeping me was worth more than them.
Maybe that was the right call - it kept the room working, but I wasn't comfortable with it. Afterall, a private group with no oversight, no appeal, and a one-person veto removed two people over a conflict that had nothing to do with the work. Nobody in the forums ever knew a decision happened. There was no announcement, because there was nothing to announce. The same room that lifted 240 people up, pushed two others out.
And none of that showed up in any of our dashboards. If you had asked our tooling where the power in that community lived, it would have pointed at the wrong room.
Alexis Madrigal coined the term dark social back in 2012: all the sharing that happens in email, texts, and DMs where analytics can't follow. By 2016, RadiumOne found 84% of consumer sharing was already happening in those untrackable channels.
That stat is ten years old and it has only gotten truer. In 2023, SparkToro and Really Good Data went after it from the plumbing side: 1,113 visits pushed through eleven networks with tracked URLs, then a look at what the analytics actually reported. TikTok, Slack, Discord, WhatsApp and Mastodon all came back at one hundred percent "direct." Facebook Messenger lost three quarters of its referrals. Instagram DMs lost a third.
So when your dashboard says "direct traffic," some real chunk of that is somebody's group chat. Your analytics file it under "typed the URL from memory" because there's no other box to put it in. Or skip the studies and check your own phone: the last five links people sent you almost certainly landed in a text or a DM.
Instagram's own boss has been saying this since 2023. Adam Mosseri noted that teens spend more time in DMs than in Stories, and more time in Stories than in feed. By 2024 he was telling creators that sends (how often people forward your post to a friend) are one of the biggest ranking signals Instagram uses. And in his year-end memo this January, he went further: "The primary way people share now is in DMs."
The platform built on public performance now ranks content by what happens in private.
This is understandable when you consider everything we've covered in this series so far. Your members are running on empty. Posting publicly is labor, and increasingly it's a risk: every take is attached to your name and your next job search. Meanwhile the public spaces themselves are filled with synthetic participation, AI-generated replies wearing human names. So people did the rational thing and took the real conversations somewhere the performance pressure and the bots can't follow.
Nobody retreated into a group chat because they stopped caring. They went there to protect one thing: trust. They built their own little trust infrastructures, one invite at a time, exactly like my 38 did. And the decisions moved with them.
6sense's 2025 Buyer Experience Report (nearly 4,000 buyers) found that 61% of the B2B buying journey is complete before a buyer ever contacts a vendor, and 80% of deals go to the favorite chosen before that first contact. The evaluating, comparing, and recommending is happening in rooms you will never see. Somebody is pasting your community's thread into their exec team's Slack right now, and your dashboard will record nothing.
Jakob Nielsen wrote up the 90-9-1 rule in 2006: in any online community, roughly 90% read and never post, 9% contribute occasionally, 1% do most of the talking. Twenty years of community platforms and the ratio has barely changed. Nielsen's advice back then was to stop trying to force lurkers to perform, but we mostly ignore that advice. Instead, we build badges and streaks and "share your biggest challenge this week" prompts, and report the resulting noise as health.
Stop trying to convert the quiet 90%. They're already working: reading everything, carrying your best threads into group chats with the words "this is what I was talking about." A member who reads every word and says nothing looks identical, on your dashboard, to a member who left eight months ago. Only one of them is gone.
You can't instrument a group chat, and you shouldn't try. Bug the room and everyone will go elsewhere. What you can do:
Then watch that room for rot. Everything that makes a private room work is also what lets it go bad, and it goes bad with no visible sign. Three things to watch for:
Mine hit at least one of those and I was standing right there when it did. You can't fix this with oversight, because oversight kills the exact trust that made the room useful in the first place. What you can do is make growth a condition of existing. A room that hasn't nominated anyone in six months is a clique with a history.
Trust has an inconvenient property: it's mostly silent. Loud metrics measure performance; quiet ones measure trust. If you only fund what's loud, you will end up defunding the exact thing that's working. That means you'll sometimes need to defend the quiet with outcomes. When leadership points at a flat engagement graph, bring receipts of a different kind: the deal that cited the community, the member who implemented an answer, the write-in attributions.
My 38 didn't post the most or have the biggest followings. What they had was trust: theirs in me, mine in them, and all of ours in the platform we were part of, because they knew I was on the inside working for them.
This week, ask five of your quietest members one question: "What's the last thing from this community you shared with someone else?" Their answers are the dashboard that matters.
That's the whole damn job.
Jon is the community builder behind global ecosystems at OutSystems, Microsoft and DocuSign. Keynote speaker, YouTube host & creator. Motorcyclist, skier, and collector of moments.
