Me: “What exactly am I looking at and what are you actually averaging?”

I’m such a kill joy.

Averages are the strangest of metrics in my opinion because they’re often quite opinionated. And opinions are decisions.

Someone somewhere chose which population to blend together before the math ever happened - and that choice right there is doing a ton more work than the resulting number ever will.

To really illustrate how ridiculous just saying an average is:

Over 99% of all humans have more than the average number of human legs (2).

The average temperature in the United States in March is 41.5F.

Imagine just leaving those there. No explanation. No deeper context.

I mean - sure, yeah both averages are very real and very accurate. But the numbers are actually kind of useless. Do we design the majority of pants and bicycles for less than 2 legs? Do we need a heavy jacket for Phoenix, Arizona?

With what I’ve given you above, it’s a big ol’ shrug.

And needless to say, communities are no different. Blending Alaska and Arizona’s late winter weather is the same as doing an average across your entire community. The averaged number tells you the temperature of a place or time that doesn’t exist.

Take the metric “average posts per member.” Sounds like a health metric. But it’s actually two very different populations all mashed into one. It’s new members still finding their footing - but also the small maintenance crew who’ve been here since the beginning and post constantly because the place would go quiet without them. But average those together gets a number that doesn’t describe either group. And maybe it looks fine. But in reality, it’s describing absolutely nobody.

Same with “average time to resolution” on a support community. Blend the veteran members who know exactly where to look for answers with the brand new members who don’t even know what to search for yet. Get a resolution time that flatters the community’s onboarding while it’s actually terrible or makes the community’s speed look like absolute garbage while it’s actually fine - all of course depending on which cohort happens to be larger that month. You can’t tell which. That’s the whole problem. The average doesn’t know which cohort it’s hiding.

So what do you actually do differently?

First, before you average anything, name the darn unit. Not the metric… the population underneath it. If you can’t say out loud what group of members this number is describing, you’re not ready to average it yet. You’re just doing fancy math and hoping it counts as insight.

Second, segment before you aggregate - not after! Cohort by tenure at minimum: week one, month one, month six, year one plus. A single “engagement score” split into those four buckets will tell you more in five minutes than the blended number told you in a year. Why? Now you’re looking at four small, coherent temperatures instead of one big meaningless one.

Third, treat any metric someone hands you pre-averaged with some serious suspicion. Ask what’s inside it. If the answer is “everybody,” ask again.

Of course, none of this is really about statistics. It’s all about admitting that “the community” isn’t one big population. It’s cohorts, arriving at different times, learning at different speeds, needing different things from the same room. Averaging across all of that and you won’t find the center of anything. You’ve just found the spot exactly between two places, neither of which anyone actually lives.

Oh and that long awaited explanation: Some humans have 1 or no legs, which therefore lowers the average below 2, despite the majority of humans having all their allocated legs.

Mostly just feeling the leftover side effects of anesthesia (which I’ll have you know I somehow just spelled that right without spellcheck… whhhhaaat!?)

Anyways. So this week is different. As you all probably know at this point, I’ve been busy standing up a new community. We’re about 3 weeks into the soft launch at this point and it’s that awkward point where you’re nervously kind of staring at your early hypotheses and wondering “Are these things going to hold?”

So instead of some grand article telling you exactly what to do, I’m just going to give you the messy run down of what my brain decides to ponder at 4am.

BET ONE: Belonging isn’t the job anymore, at least not for this kind of community.

As recent as Wednesday of last week, I had a peer suggest I set membership as an objective for the next few months. How about….

Needless to say, “building for belonging” is great for a lot of communities. But the community I’m building is a knowledge and support community, not a social one, and I fervently believe membership as an objective has no home here. Knowledge transfer is going to be my bread and butter. Transactional, in-and-out, solved-my-problem-and-left behavior is going to be my world and my strategy, metrics, effort, and reporting are all going to reflect that… for good or bad.

BET TWO: The traffic that actually matters to me in this era hasn’t started counting yet.

If you asked me 5 years ago about metrics, you’d have heard me tell you all about how page views, unique visitors, and the like were pretty darn important. And they were. But then our AI overlords showed up and I’m betting on new metrics now.

Direct traffic just ain’t what it was.

I don’t think this community’s page views, unique visitors, or impressions become objectively meaningful until the content starts showing up inside LLM-generated answers, not just usual search results. And if that’s true, then the launch campaign, the engagement push, the whole freaking arc of “get people here, get them excited” is mostly serving hype (for me TBH; I like any excuse for cake) and peripheral awareness right now. And that means, none of that will be the primary growth driver. Heck, it might not even be a growth driver at all in the way we’ve traditionally meant it.

So instead of pouring effort into the perfect launch, I’m spending more of my time on what I’m calling concurrence efforts. This is the stuff like making sure the content is structured, cited, and cross-referenced in a way that gives it a shot at being pulled into generated answers across domains. And that’s kind of a funky thing. 5 years ago I was doing customer marketing stuff to launch. Now? AI massaging. Cool cool.

BET THREE: Localization isn’t worth the investment anymore.

Someone out there is clutching their pearls and gasping at ^^ statement, but I said what I said. For a long time, localization was the thing for advanced communities with platforms and vendors going so far as to offer in-platform localization and other vendors charging exorbitant sums to translate knowledge bases.

But translation tools are good now. Like legit good, not conference-demo good. As long as I stay aware of cultural context and keep the content itself relatively neutral and adaptable rather than idiom-heavy or region-specific, I don’t think dedicated localization work is where my limited brain juice should go.

Note: I’m not saying culture doesn’t matter. It’s saying the effort has moved. I’m spending effort on making content translate cleanly instead of translating it myself.

What ties these together, if anything…

Mostly just that I’m questioning long inherited defaults. Belonging, traffic, and localization all made sense as priorities in the communities most of us cut our teeth on. But I’m not one bit convinced they automatically transfer to a knowledge and support community operating in a landscape where AI is starting to mediate discovery. So I’m making different bets. And hey, I could be wrong.

But that’s sort of the point of writing this one down now instead of waiting until I’ve got it all figured out. You get to watch me find out in real time.

If you’re running your own version of any of these bets, I want to hear about it. Reply and tell me where you’ve landed, and where you think I’ve got it backwards.

But don’t think that means we walk out the door everyday with things for every single weather type. Over the years, we all learn how to read the upcoming weather in other ways - take a little bit of what the meteorologist think, a few looks at the sky, observe what the leaves and birds are doing, and all that weird stuff.

So why am I starting off a nerd post with a chat about weather? It’s not small talk. It’s because I actually think it’s a great approach to community climate too. As I’m getting ready in the morning, I’m not looking at the humidity for just sh*ts and giggles. It’s a little metric that gives me big insight into what to expect.

Community metrics work exactly the same way - if you’ll let them.

But somewhere along the way of building communities, dealing with the analysis paralysis of looking at community dashboards, and trying to report on our observations - we seem to have stopped viewing engagement metrics as weather instruments and started treating them as the weather itself. A spike in comments doesn’t mean your community is thriving any more than a pressure drop means a hurricane has already made landfall. But it does means something is happening that you haven’t fully diagnosed yet.

The metrics we measure are almost always proxies. Treating them as endpoints is where community strategy goes wrong.

This distinction sounds pretty freaking obvious in the abstract. But it gets much harder to hold onto in a quarterly business review, where a stakeholder wants a number that goes up and to the right. And unfortunately for us community professionals, “Well, technically this is a leading indicator of an underlying condition we haven’t fully characterized” does not fit on a slide or satisfy the stakeholder.

Sigh.

So we simplify. Engagement becomes health. Post volume becomes vibrancy. Reply counts become belonging. None of these substitutions are unreasonable on their face (a community with zero activity is obviously not thriving), but the substitution itself is where the trouble starts. Once a proxy gets treated as the target, people start managing the proxy instead of the thing it was supposed to represent.

You have most certainly seen this play out - if not literally been the one played. A community manager under pressure to show engagement growth starts running more prompts, more polls, more “what’s everyone up to this week” threads. The numbers move. Leadership is happy. But nothing about the underlying health of the community has changed! You didn’t create more trust, more expertise-sharing, or more durable relationships between members. You created more weather readings. You made the barometer twitch without doing anything about the actual system generating the pressure.

And we’ve talked about this in other articles: some of what we can measure barely correlates with what we actually care about, and some of what we care about most can’t be measured at all.

More sigh.

Think about what actually predicts whether a community survives a rough year and what doesn’t. It’s almost certainly not last month’s post count. It’s almost certainly members trust each other enough to disagree in public without the whole thread turning into a psychological hazard. It’s whether institutional knowledge lives in more than one or two people’s heads. It’s whether someone would notice, and say something, if a longtime contributor went quiet for a month.

And of course - none of that shows up cleanly in a dashboard. All of it is more predictive of long-term health than anything your analytics platform will hand you on a Monday morning.

So what do you do with metrics that are proxies at best? You don’t throw them out with the barometer. A pressure drop is still useful information, even though it isn’t the storm. The move is to change what you’re asking the number to do for you.

A few shifts to consider:

1. Stop asking “is this number good or bad” and start asking “what underlying condition would produce a number like this.” A quiet month isn’t automatically bad news. It might be a maturity signal instead of a decline signal, and the only way to tell the difference is to look past the reading itself.

2. Pair every quantitative metric with a qualitative check before you report on it. If comment volume is up, go read the comments. Find out if you’re watching real exchange or a spike in low-effort reactions. The number alone won’t tell you which one you’ve got.

3. Get explicit with stakeholders about what a given metric can and can’t tell you, before they build a narrative around it themselves. It is far easier to set that expectation up front than to walk it back after someone’s already put “engagement is up 40%” in a board deck.

4. Build at least one measurement into your practice that captures something durable, even if it’s slower and harder to report on quarterly. Trust, knowledge distribution, and continuity of relationships move on a different clock than engagement metrics do, and they deserve their own instruments.

And if all else fails, do what us weird northern New Englanders do: walk outside, look at the sky, listen to the birds, and just run with what you see. Chances are it’s not far off from the right number.

Every community manager out there has some version of the sentence of “our members are incredibly knowledgeable; experts in fact”. And yeah, we totally mean it as a flex. Deep expertise reads as a success metric, proof that we built something serious enough to attract serious people.

Nobody - not even me - puts the second half of that sentence in the deck: the deeper the expertise, the quieter the room gets.

Not empty. Quieter. There’s a difference and it takes awhile to be able to see it and spot the difference. An empty community has nobody home. A quiet expert community has plenty of people home, most of them just sitting very still, waiting for a question that is interesting enough to answer.

Picture a dinner party where everyone at the table has spent twenty years in the same field. Nobody’s going to ask what a term means. Nobody’s going to float a half-formed potentially stupid idea. You don’t workshop your thinking out loud in front of people who could correct you in their sleep, you wait until you’ve got something polished enough to survive the table. Multiply that dinner party by a few thousand members and you’ve got most expert-heavy communities I’ve ever worked on.

The more expert a community gets, the higher the invisible bar for speaking rises. And the fewer people feel qualified to clear it… often including your actual experts.

That’s the paradox on its own. And then Dunning-Kruger is what makes it interesting. Quick refresher, since it gets name-dropped a lot more than it gets explained: it’s the idea that the people who know the least about something are often the most confident talking about it, because they don’t yet know enough to see the size of the gap. The people who actually know the subject cold tend to go the other way, they’ve seen enough of the gap to stay a little humble about it. In a community, that effect doesn’t distribute evenly. It splits people into two very different kinds of quiet, and if you’re not looking for both, you’ll only catch one.

On one end, you’ve got people early on the curve who don’t know enough to know what they don’t know. In a beginner-friendly space, this group posts constantly (confidently, sometimes wrongly, and without much fear of being wrong). But drop them into a genuinely expert community and something shifts. The room’s competence is visible enough that even the overconfident start to sense they’re out of their depth. Some still post anyway, loudly, and they’re the barrier newer members point to when they say the space feels intimidating. But a lot of them just go quiet too, for the first time in their posting lives.

On the other end, you’ve got the people who’ve actually earned the expertise, the ones who could answer almost anything in the community without blinking. And this is the group nobody talks about enough: they stop volunteering it. Not because they don’t care. Because when you actually know a subject that well, you know exactly how much nuance gets lost every time you simplify it for a public post, and most experts would rather say nothing than say something they’ll have to caveat five times in the replies. So they lurk too. Just for the opposite reason.

But what does that mean in practice?

Your most expert community can look, from the outside, identical to your least expert one. Low posting volume. Long gaps between threads. A handful of loud, confident voices that don’t necessarily represent the smartest people in the room.

Great. Wonderful. What should we do? What should I do?

If you’re managing one of these spaces, a few things try.

First, stop reading silence as disengagement and start asking which kind of quiet you’re looking at. A beginner-friendly community that’s gone quiet has a different problem than an expert one, and they need different fixes.

Second, build deliberate low-stakes entry points, not for beginners, but for your experts. Give the people who know the most a low-effort way to say a small true thing (a quick take, a “here’s what I’d flag,” a reaction to someone else’s post) without having to produce a fully-defended thesis every time they show up.

Third, be honest that your loudest members might not be your most expert ones, and stop letting volume stand in for authority when you’re deciding who to spotlight, feature, or hand the mic to. (Gives a mean side eye to super user programs…)

And fourth, remember that expertise doesn’t scale confidence. It scales caution. The more someone actually knows, the more they understand what they don’t, and the harder they think before hitting post.

So what do you think? Is your community quiet or empty?

But this isn’t some sort of false modesty or admission of failure. Instead it’s my strategy. Somewhere around my second community build I figured out that the launch decks everyone loves, you know the ones with the fully mapped member journey and the six month content calendar and the engagement tiers with cute on-brand little badge names are mostly theater. Impressive theater, nonetheless. Great for a slide. But still, theater.

The second real members show up, I can guarantee with alarming certainty they’re going to do something with your community you didn’t plan for. Not because you’re bad at your job, but because you are structurally one person (or one small team) guessing at the behavior of a crowd you haven’t met yet in its entirety. If your deck already has an answer for everything, your guess has limited their reality before they’ve even walked in the door.

So, 60/40. Sixty percent locked: the purpose, the guardrails, the non-negotiables, the stuff people can actually lean on. Forty percent left deliriously squishy on purpose, so the community gets a vote in what it becomes.

Speaking from a lot of experience, the 40% never stops feeling like a professional liability. We - the community professionals - are supposed to have the answers; that’s kind of the whole job description. Except the version of me with all the answers never gets asked the question that actually mattered, because nobody in my conference room thought to ask it. Only my members would’ve thought to ask it and there is generally no overlap between those two groups.

The best community strategies are the ones still capable of being surprised.

The 60% is real structure though, not a suggestion. I’ve made the mistake of thinking “we’ll figure it out” was a plan instead of a shrug, and vague is not the same thing as flexible. Your purpose, your moderation floor, your core value exchange - lock those in. People build trust on that part. The squishy part only works because it’s squishy inside something solid, not squishy all the way down to the studs.

Your stakeholders will read your unfinished 40% as a gap every single freaking time, no matter how you frame it. Defending the parts you left undone on purpose, to someone who needs a finished slide for their own boss by Thursday, is its own quiet skill. “We’re leaving room to learn” sounds suspiciously like “we don’t know yet” if you’re not the one who built the thing.

My favorite part though: when a member starts using the community in some way you never planned for… that’s a gift, not scope creep. They’re doing your strategic thinking for you, unpaid, in real time, and all you have to do is notice instead of herding them back to the roadmap.

The ratio moves too, and it should. Month one’s 40% is not month twelve’s 40%. Good chaos eventually teaches you what deserves to become permanent, and the structure grows to match what you’ve actually learned instead of what you guessed at back in the conference room.

The hardest part of all this was never the math though. Not really. It’s sitting across from someone who wants certainty and telling them, straight faced, that a big ol’ chunk of the plan doesn’t exist yet on purpose. And that doesn’t feel like confidence unfortunately. It feels like showing up slightly underdressed to your own launch.

The communities I’ve watched fail the hardest weren’t the ones that left room for surprise. They were the ones so fully mapped on day one that no member ever got to leave a fingerprint on the place or members bucked so hard against the rigidity they broke the foundation. You can build something beautiful and airtight and still end up with a building nobody wanted to move into.

Leave the 40% empty. Somebody’s going to want to fill it in for you.

They shouldn’t be.

A community that answers its own questions well enough that people stop asking them isn’t failing. It’s finishing the job it was designed for. Hurrah!

We built our metrics around visible activity because visible activity is easy to count. Questions asked, threads started, replies posted. All of it assumes that a healthy community is a busy one, forever generating fresh confusion for people to resolve together. But that’s not maturity. That’s a community that hasn’t figured most things out yet.

A community that stops generating the same questions over and over isn’t dying. It’s the thing you were trying to build in the first place.

Think about what actually happens when a knowledge base does its job. Someone hits a snag, searches, finds the answer that three other people already worked out and documented, and moves on with their day. No thread. No reply count. No ping in your notifications to reassure you that people still need you. Just a problem that quietly stopped being a problem. You built infrastructure specifically so people wouldn’t have to ask, and then you built a dashboard that punishes you for succeeding.

But alas… this is where the maturity curve gets mistaken for decline. Early on, a community is loud by necessity: nothing’s been answered yet, so everything gets asked. That’s when the metrics look great and the leadership meetings feel easy. But the curve should bend. Questions should get more novel, less repetitive, further out on the edge of what your content already covers. If your question volume looks the same in year two as it did in month two, that’s not stability at all. More like a sign your knowledge base never actually absorbed anything. Oof.

So what do you actually do with this?

First, stop reporting question volume as a headline number without context. A raw count tells you nothing about where a community sits on its own curve. Pair it with something like question novelty (are people asking things you’ve genuinely never addressed) or repeat-question rate (how often is someone re-asking what’s already answered, which tells you your content isn’t surfacing, not that your community is thriving).

Second, get specific about what “declining engagement” could actually mean before you let anyone panic in a meeting about it. A drop in basic questions paired with a rise in novel, complex ones is a community leveling up. A drop across the board, novel and basic alike, might mean people left, or gave up, or found somewhere else to ask. Same number, opposite story. Your dashboard can’t tell the difference. You have to.

Third, this changes what you build toward. If the goal was always fewer repeat questions, you should be celebrating your best FAQ articles and your clearest documentation as much as you celebrate your busiest threads. Right now most of us report on what happened in the community and quietly ignore what the community made unnecessary. That’s backwards. The unnecessary is the win.

Fourth and potentially uncomfortable: this means part of your job, maybe the best part of your job, produces no evidence of itself. The four support tickets that never got filed because the doc was clear. The onboarding question that never got asked because the welcome flow already answered it. You will never get credit in a headcount meeting for a fire that didn’t start, and you have to make peace with that or you’ll spend your whole career optimizing for the wrong signal because it’s the only one that shows up on a chart.

Fifth, and this is the one I’m living right now: if the absence of a question is the win, then seeding a brand new community with those answers before anyone asks is not busywork, it’s the whole strategy. I’m building a community from zero right now, and the temptation is to wait, let real questions surface organically, then respond and look responsive. But every question you preempt is a person who got value in their first five minutes instead of their first five weeks. A new member who searches on day one and finds the answer already sitting there doesn’t experience that as “nobody’s talking in this community yet.” They experience it as “oh, this place already gets it.” Seeding isn’t faking activity. It’s compressing the distance between showing up and getting something out of showing up, and that distance is the whole battle in the early days.

I’ve got ways of tracking this now, question novelty scoring, repeat-rate baselines, the works. I’m not going to hand you the exact framework, because the version that works for my community is tuned to my community, and a framework built for one group’s maturity curve is a great way to make a completely different group’s leadership panic for entirely new reasons. Build your own. Watch your own curve. The point isn’t the specific formula, it’s that you need one, and “trust me” without any formula at all doesn’t survive budget season.

But the reframe matters even without the perfect metric. The next time someone tells you engagement is down because the questions dried up, ask them what dried up first, the questions or the confusion. Those are very different problems, and only one of them should be keeping you up at night.

But anyways.

Not only have I built a lot of community launch decks, but I’ve sat through probably 50x as many community decks from platform vendors, other communities, and more. And more than a small handful have a similar slide: a stunning landing page, a big ol’ friendly button that invites users to join, and someone excitedly explaining onboarding functionality like it’s the only way anyone will ever arrive. One door. One hallway. One version of hello.

That was never fully true, but it’s now catastrophically untrue. Someone finds your community through a Google search that dumps them straight into a three year old thread about a bug that was fixed two releases ago. Someone else asks an AI assistant a question and gets a paraphrased answer with a link buried at the bottom, if they get a link at all. A third person hears about you at a conference bar because a colleague name dropped your forum while complaining about something unrelated. Three people, three completely different rooms, and you designed a reception for… well… none of them.

I gave a talk last year about how architectural principles shape human experience across physical and digital environments, and I keep coming back to it because it maps onto this almost too cleanly.

A front door isn’t just an opening in a wall. It’s a invitation. You walk in, there’s a reception, maybe some much needed information, and a beat where your eyes adjust and your nervous system registers I am allowed to be here. Then you’re guided toward wherever you’re actually going via hallways, corridors, walls, signage, and more. And that’s truly just grand if you need that experience.

Now let’s say you already know this place and you know exactly where you want to go. You know there’s a side door that takes you right to where you want to be. Also truly grand.

But both of those are very intentional behaviors. You entered the space with a desired outcome; the first with exploration and the second with direction.

So now let’s say you didn’t see the front door, grabbed the handle of the side door. And the side door just skipped all of that ‘truly grand’ stuff from above. It drops you straight into the kitchen mid dinner party; no reception, no adjustment beat, just you and a room full of context you don’t have.

Uh… am I allowed to be here?

Search and AI summaries are side doors. They don’t route people through your front door and to reception. They drop them into a specific room, mid conversation, and expect that room to make sense on its own. So does it?

Most communities have spent a decade perfecting the front door and completely ignoring what happens at the side entrances.

That’s the thesis, and it’s worth sitting with, because it explains a specific kind of failure you’ve probably felt but not named: the sense that your community’s best content, the stuff that actually converts curiosity into belonging, gets buried three clicks deep, while a half baked thread with a so-so title is quietly doing the work of your entire acquisition funnel.

So what does it actually mean to design for three doors instead of one.

Your reception content (the landing page, the welcome thread, the about page) still matters, but it’s no longer your primary entrance. Its job shifts from persuasion to orientation: helping someone who’s already inside find their bearings, not convincing a stranger to walk through the wall.

Your side door content, the stuff that gets surfaced by search or scraped into an AI answer, has to work as a complete room. No hallway, no reception, no assumed context. A piece of knowledge that can’t stand alone and still be true, still be useful to someone who has never heard of your community, is not going to survive being extracted. Write for the person who arrives with zero reception.

Word of mouth is its own kind of door, and it’s the one most people forget is architectural at all. Someone vouched for you before the visitor ever showed up, which means the visitor is walking in with trust already extended on your behalf. Over explaining insults that trust. Under delivering breaks it. The room needs to be exactly as good as the person who recommended it implied it would be: no more theatrical, no less substantial.

The part that’s genuinely uncomfortable if you’ve spent years optimizing a single funnel: you don’t get to choose which door most people use anymore. The AI summary door is only going to get busier. The side entrances outnumber the front one now, probably permanently, which means the room by room experience of your community matters more than the building’s facade ever did.

I still love a good reception. First impressions matter and that the welcome experience deserves real design attention. But if every door except the front one drops a stranger into an unfurnished room with no context and no signage, the reception as entry point never mattered. It was just the door you happened to be standing at when you built the place.

Go check your side doors. I promise you’ll find at least one room with the lights off and nobody there to make sense of the chaos.

The ones who are quietly dismantling their communities because AI is going to make them obsolete anyway.

The ones who are carrying on like it’s 2019 and nothing fundamental has shifted.

And then there’s a smaller, quieter group. We’re a mix of deeply understanding what’s happening and adapting accordingly… or those just getting on with it because why the heck not?

This article is for that third group. But the other two are welcome to stay. Maybe we’ll convert you to the dark side.

But first, we’re going to need you to accept a really harsh reality: AI didn’t change what good community knowledge looks like. It just made it impossible to pretend that bad knowledge architecture was fine.

Do you know the move I’m talking about? Someone posts a question your community has definitely answered before, and instead of pointing to a crisp, documented resource, someone types: “Oh, great question, you should ask Jillian, she handled this exact thing last quarter.” And Jillian answers, because Jillian is generous and competent and has the institutional memory of a small government agency. Crisis averted. Thread resolved. Nobody notices that the actual answer lives entirely inside Jillian’s head.

That worked. Until it didn’t.

The communities that are panicking about AI right now were always one Jillian-on-PTO away from collapse.

Think about what “ask Jillian” actually means. It means your knowledge was never really in your community. It was in a person, and that person happened to be reachable. The member who joined six months ago and doesn’t know Jillian yet? They either figured it out themselves, gave up, or got lucky. That’s not community knowledge. That’s a single point of failure with good vibes.

And yes yes - we know the retort: “But but but… AI!”

AI didn’t create this problem. It removed the social scaffolding that let us ignore it.

Here’s where I’d gently push back on a story we tell ourselves about human cognition: we like to think we navigate knowledge relationally, that we’re fundamentally different from machines because we connect ideas through people and context and lived experience. And sure, there’s something to that. But mostly? We know where things are because we have context. We remember that the onboarding doc lives in the folder that so-and-so reorganized last spring. We know the workaround for that integration bug because we were in the Slack thread when it happened. That’s context. Not magic. Not some ineffable (how great is that word?) human quality that machines can never touch.

The gap between how humans find knowledge and how AI finds knowledge is narrower than we’d like to admit. (I know, I know. Uncomfortable. Stay with me.)

Which means the communities that will hold up, the ones that serve members well regardless of whether those members arrive having already asked ChatGPT something, run it through their company’s AI tool, or just typed a search query like it’s 2011, are the ones that were already doing knowledge architecture right. Clear, contextual, findable, not dependent on any single person’s availability or generosity.

This is not about making your community “AI-ready,” a phrase I’d actually be gloriously happy to retire. Communities serve humans. Full stop. But humans now move through information differently. They come in with partial answers. They paste things into tools and come back with follow-up questions. They expect to find what they’re looking for without having to know who to ask or which part of your site to go to. That’s not a new expectation, actually. It’s just newly consequential when it goes unmet.

So what does this mean practically? It means documentation that stands alone, with enough surrounding context that a reader (human or otherwise) doesn’t need backstory to make use of it. It means tagging and titling that reflects how people ask questions, not how the internal team thinks about categories (I’m glaring at you personas). It means periodically asking: if Jillian left tomorrow, what would actually be findable? What would disappear with her? (Not much actually; Jillian knows to document her brain on the regular but Jillian is weird.)

It also means letting go of the fantasy that relational knowledge transfer scales. It doesn’t. It never did. Jillian has been subsidizing your knowledge architecture for years, and she deserves a vacation. Italy or France will do.

The communities worth building, right now, are the ones that work when nobody knows anybody. They’re also the ones that become richer when people do connect, when Jillian’s expertise gets documented and built upon rather than just tapped and forgotten.

And y’all… that’s not an AI story. That’s just a community worth having.

Go give Jillian some time off.

A former boss of mine (Rachel Happe: LinkedIn | Engaged Organizations | Control is for Amateurs) used to gleefully and mischievously say “Control is for amateurs”; so much so that she made us t-shirts. She’d just drop it into conversation, totally unprompted, like it was obvious. It took me an embarrassingly long time to understand what she meant.

I kept waiting for the second half of the sentence. The “so instead, do THIS” part. The framework. The playbook. The three steps to structured freedom or whatever. It never came. Because that was the point. The people who feel compelled to reach for control, to define the outcomes, to spell out exactly what participation looks like and what it’s for… those are the amateurs. The masters are doing something sneakier (and sneaky she was - still IS!). And more interesting.

Here’s the thing we tell ourselves in community work: clarity drives participation. Make the ask specific. Show people exactly where their contribution lands. “Your answer helped 42 people this month.” “Your post was the top resource in Q3.” Explicit impact, explicit prompt, explicit result. Tight loop, measurable, reportable to stakeholders. Clean.

And it’s not wrong… exactly. It works. People respond to it. But it’s also a ceiling.

Here’s what social physics (yes, that’s a real field, and yes, it will make you feel like a puppet and a puppeteer simultaneously) keeps showing us: human behavior is not nearly as spontaneous as humans believe it to be. We are wildly, almost insultingly predictable. The conditions you set determine the behavior you get, far more than the individuals making choices inside those conditions. Which means the actual job of a community designer is not to control behavior. It’s to architect the conditions that make the behavior you want feel like the member’s own idea.

The best community design is invisible. Not because it’s subtle. Because it’s a theme park.

Think about how a theme park works. Every path, every sightline, every queue is engineered. The “spontaneous” detour through the gift shop? Planned. The feeling that you discovered that hidden churro cart? Designed. The mechanical box you never noticed because it was painted “Go Away Green” (no, really - it’s a thing)? You are being guided at every moment, and you are having the time of your life, and you will go home and tell people you just wandered around and it was magical. Disney didn’t hand you a map with the route highlighted and a timer. They built an environment where you couldn’t help but do exactly what they needed you to do, while feeling entirely free.

That is what mastery looks like in community design.

So what does this actually mean for how we work? A few things worth munching on.

First: Stop over-narrating impact. When you tell someone exactly what their contribution accomplished, you close the loop for them. You also close the door on their imagination, which is almost always more generous than your metrics. A member who wonders “I hope that helped someone” will carry that question into their next visit. A member who knows “that helped 12 people” has completed a transaction. Leave some darn room, willya?

Second: Watch what happens when you remove the explicit prompt. Most community programs are prompt-heavy: “Share your story,” “Answer this week’s question,” “React if you agree.” These work. They also train members to wait to be asked. The communities with the most durable contribution cultures tend to have softer structures, ones that create occasion without dictating behavior. There’s a difference between a room with assigned seats and a room with good acoustics. Both fill up. Only one feels alive.

Third: Trust the container more than the content. Your job is to build the walls, the lighting, the ambient conditions. What happens inside is, genuinely, up to the people in the room. The instinct to over-program, to fill every silence, to make sure every interaction has a clear purpose and a tracked outcome… that instinct is not serving your members. It’s serving your anxiety. (I say this as someone who has definitely built a spreadsheet tracking “spontaneous” member interactions. We contain multitudes.)

Fourth, and maybe the hardest: Get comfortable not knowing exactly what worked. Ambiguity in impact is not a measurement failure. It’s a sign that people are doing things you didn’t design for, which is the whole goal. The moment your community is perfectly legible is the moment it has stopped surprising you. And a community that can’t surprise you is one that your members have already outgrown.

My former boss was right. Control is for amateurs. The professionals build the theme park and then buy a ticket if for nothing else but the people watching.

And trust me, y’all are predictably entertaining.

But I’m going to suggest a uncomfortable swap, though. Well, uncomfortable for you. I got comfy with this a long time ago.

What if the lurker isn’t the edge case you’re tolerating? What if they’re the actual user, and the person posting is just... infrastructure?

Marinate in that for a second, because it changes everything downstream. Most community programs are built, measured, and pitched to leadership around the people who show up and talk. The 1% writing the posts, asking the questions, starting the threads, they’re treated as the community itself, the thing you’re trying to grow. Everyone else is a rounding error, or worse, a problem to be “activated.”

But think about how you actually use most communities you’re not active in. You search a Slack workspace for how someone else solved the bug you’re stuck on. You read three pages of a forum thread before deciding not to post. You scroll a subreddit for twenty minutes during lunch, contribute nothing, ever, and leave knowing something you didn’t before. None of that shows up in a dashboard. All of it is the thing working.

The 1% aren’t the community. They’re the maintenance crew for a building full of people they’ll never meet.

I don’t love writing that sentence as I am, after all, part of the 1% in most of the communities I belong to, and “maintenance crew” is not exactly the origin story I tell myself (I prefer “architect of chaos”). But if the lurker is the primary user, the people posting are providing a service to someone else, not building a thing for themselves. Bit of an inversion, eh?

So if we take that inversion seriously, a few things stop making sense the way they used to.

The whole vocabulary of “community” starts to wobble. Trust, reciprocity, belonging, a sense that other people are there with you, those words assume two-way contact. If your primary user never posts, never replies, never even creates an account, do they have any of that? Or do they get a version of it anyway: trusting the same FAQ a thousand other people trusted, absorbing the same norms by reading instead of participating, feeling like “people like me are here” because the archive proves it, just distributed across time instead of synced across people?

I genuinely don’t know. Some days it feels like community, evolved, quieter, asynchronous, but still very very real. Other days it feels like a ghost of community from the days gone by: a town that’s all storefronts and no residents, kept running by the handful of people who still show up to stock the shelves, while everyone else just... walks through, takes what they need, and leaves. Same buildings. Different town.

And maybe that’s the part worth not resolving. Because here’s what doesn’t change, regardless of which version you believe: if your strategy, your roadmap, and your metrics are all built around the 1%, you’re optimizing for the maintenance crew and ignoring the building’s actual occupancy.

That means the read experience needs as much design attention as the write experience, because for most of your users, reading is the entire interaction. It means “active member percentage” stops being a meaningful headline number (a concept I’ve thoroughly explore in other articles here), because it was never measuring your primary user in the first place, it was measuring your staff. It means onboarding has to account for someone who’s already had four useful visits before they ever consider creating an account, because by the time they show up in your numbers, they’ve already gotten most of what they came for. And it means the loud, visible, engaged minority deserves a different kind of recognition, not because they’re “the community,” but because they’re doing labor on behalf of people who will never know their names.

Whether that’s still community or something quieter wearing community’s name, I’m not sure it matters as much as we think it does. What matters is whether you’re building for the people who show up in your reports, or the people who show up, get what they need, and never appear in one at all.

Go ahead and thank the second group. They won’t see it. They’re not logged in.

What doesn’t make it into the onboarding deck is this really finicky little thing: the moment your first members walk in, you start making promises. Not in your guidelines. Not in your welcome post. In the design itself and your behavior. In who you let in first, how you responded to the first difficult thread, whether leadership showed up in the forums or stayed behind glass. In all of it.

By Year 3, those promises are big burly walls.

We (me included) talk a lot in this profession about community design like it’s a series of discrete decisions: category structure, membership tiers, moderation policy, gamification. And it is. But it’s also something harder to name. Every design choice in Year 1 sends a signal about what this space values, who it’s for, and how status works inside it. Members absorb those signals. They build their participation habits around them. And then, quietly, they build their identity around them too.

So therefore: your founding ethos isn’t what you wrote in the mission statement. It’s what your earliest members learned to expect.

This is the thing that makes Year 3 redesigns so much more treacherous than they look on paper. You’re not updating a category taxonomy or refreshing a recognition program. You’re renegotiating a social contract that your community didn’t know it had signed, with members who didn’t know they were signing it, but who will absolutely notice when you try to change the terms.

Take something as seemingly low-stakes as expert-to-novice knowledge flow. If your community launched with strong subject matter experts front and center, and your early moderation rewarded definitive answers over exploratory questions, you’ve done more than establish a posting norm. You’ve told your members what intelligence looks like here. You’ve created a hierarchy, and some people have climbed it. When you decide in Year 3 that you want more peer-to-peer conversation, more vulnerability, more “I don’t know, what do you think?” energy, you’re not just asking for different posts. You’re asking the experts who built their reputation on certainty to step down a rung. And you’re asking the community to redefine what it means to contribute well.

And while it feels like a content problem, it’s not. It’s an identity problem.

The same dynamic shows up in access and responsiveness. If your team was highly visible in Year 1, answering questions fast, showing up in threads, being genuinely present, members calibrated to that. It became the baseline. The expectation. When your team grows and bandwidth stretches thin, or when the strategy shifts toward member-led support, the community doesn’t read that as a natural evolution. They read it as withdrawal. Because from their perspective, something was offered and then taken away.

And yes, you can explain the change. You can announce it thoughtfully. But you’re still arguing with the founding ethos, and the founding ethos has three years of precedent on its side.

None of this means Year 1 decisions are irreversible. Communities do change, and some of them change well. But the ones that manage it tend to share a quality: they treat the redesign as a renegotiation, not a correction. They name the original ethos explicitly, honor what it built, and bring members into the conversation about what needs to shift and why. That’s slower and messier than a clean product update. But it also actually works.

The harder lesson, the one I’d have wanted someone to hand me early in my career, is that intentionality in Year 1 isn’t just about building something good. It’s about building something you can actually change later. Because you will need to change it. But will you be renegotiating or apologizing?

Choose your founding ethos on purpose. Your Year 3 self will be living inside it either way.

And so once again, we’re going to visit a metric I despise: monthly active members. Some community dashboards will give you daily actives. A few will break it down by post type, session length, time of day. You can slice your membership into cohorts and watch them drift in and out of your engagement funnel like tides.

What none of them will show you is the conversation, conversion, or behavioral shift that happened because of your community, somewhere else entirely.

Your member who lurked for six months before pulling a thread on customer onboarding and sharing it in her company’s internal Slack. The practitioner who cited a three-year-old discussion in a presentation to his leadership team. The person who never posted once but quietly absorbed enough that they stopped making a particular mistake. That value left your platform. It’s living somewhere you’ll never see. And it’s been happening since the first community went online. And it’s probably the best thing your community will ever do and keep on doing.

This is not new behavior. Communities have always been leaky by design, in the best possible way. We learn things and we take them with us. That’s the whole point. A community that produced knowledge people hoarded inside it would be a pretty useless community.

But alas… The problem isn’t that value escapes the platform. The problem is that we built dashboards that can only see inside it and expectations only aligned to what we can see.

Think about a library. A library doesn’t know that you recommended a book to your sister, or that you spent three hours with a resource and it changed how you approached a problem, or that the idea you found in the stacks eventually ended up in a report that influenced a decision at your organization. It knows you checked out the book. Maybe it knows how many people checked out that same book. The rest is invisible, which doesn’t make it any less real.

Community platforms work the same way. We instrument what happens inside the walls because that’s what we can count. Logins, posts, replies, reactions. Useful signals, all of them. But they’re not the whole picture, and for a long time we let the measurable crowd out the acknowledgment that something larger was happening beyond it.

We designed for re-entry when we should have been designing for portability.

That means asking different questions about what we build. Is your best content accessible to someone who hasn’t logged in recently, or is it locked behind a gate that requires fresh authentication? Are your templates, frameworks, and resources structured so members can actually take them somewhere? Is your community producing the kind of distilled, usable knowledge that travels well, or is it mostly in-the-moment conversation that loses context the moment it leaves? These aren’t arguments against a logged-in experience, but they are arguments for being intentional (and realistic) about what you’re making and where it’s going to end up.

But why is this feeling really uncomfortable right now? Why are you squirming reading this going ‘oh, I feel seen and I don’t like it’?

A lot of us are confronting this belatedly because our new AI overlords have made the invisible visible. Language models trained on forums, help communities, and practitioner discussions surfaced, at scale, the knowledge our communities had been producing for years. The value was indexed… outside our communities. It was findable… outside our communities. It turned out our members’ collective intelligence was genuinely useful and widely usable… outside our communities.

However, we didn’t measure that. We didn’t build for it. We were watching active member percentages while something else entirely was happening with the content we’d spent years cultivating.

There’s something both validating and humbling about that moment. Validating because it confirmed what community builders have always believed: this work produces real, lasting value that’s so good that WE are the ones feeding AI the valuable stuff. Humbling because the entity that made that legible wasn’t our analytics platform. It was someone else’s model.

So here we are. The behavior was always there. The knowledge was always escaping the walls and making its way into the work. We just finally have a mirror big enough to see it.

A little too late, but better late than never.

So am I alone in this? Or is this a thing?

The reality is that just one bad review can outweigh five good ones. Not because people are irrational, but because recency and specificity do something to our brains that aggregate ratings can’t undo. A glowing 4.8 with 200 reviews means nothing if the most recent entry says “got food poisoning, still not over it.” You close the app. You pick somewhere else. (Reminds me to go leave a review on the place that gave me food poisoning last weekend…)

I think about that a lot when I’m looking at community spaces.

We spend a lot of time in this profession talking about what our best interactions look like: the member who got a thoughtful response, the feedback thread that turned into a product feature, the support question that became a knowledge base article. We curate those wins. We screenshot them for our quarterly reviews. And none of that is wrong, exactly. But it might be obscuring the more important question which is: what did the worst interaction in your community teach every lurker who watched it?

Because someone is always watching.

The psychological cost of participating publicly in a community, whether to ask for help or critique a product, isn’t fixed. It fluctuates based on what people have already seen happen to others. Before anyone posts, they do a quick, mostly unconscious threat assessment. They scroll back. They read a few threads. They’re not looking for the average experience. They’re looking for evidence of the worst case.

This is why support and feedback feel like the same problem even though they’re not quite the same risk. When someone is considering posting a support request, the vulnerability on the line is competence. What if I’m the only one who doesn’t understand this? What if someone tells me to read the docs in a way that makes me feel stupid? When someone is considering posting product feedback, the vulnerability is credibility. What if I suggest something that’s already been rejected a hundred times? What if the response makes me look uninformed? Different fears, same mechanism. They’re both scanning for evidence of how this goes when it goes badly.

The floor of your community sets the ceiling for participation.

This is the thing we underinvest in. We optimize for making great interactions great, and we let the bad ones quietly do their damage. A cold, dismissive response to a frustrated user doesn’t just affect that user. It’s a public record. It sits there, searchable, permanent, training every future reader on what this space does with vulnerability.

So what do you do with that?

First: your worst public moments deserve the same strategic attention as your best ones. Most community teams have a plan for amplifying wins. Very few have a plan for actively rehabilitating bad threads. That doesn’t mean deleting them (almost never the right call). It means going back, adding context, following up, making sure the public record reflects that the community course-corrected. A thread that starts badly and ends well tells a better story than a thread that starts well and gets abandoned.

Second: if you’re designing for participation, you’re designing for the lurker. The person considering whether to post is your real audience. They outnumber active posters by a lot (honestly, by a lot… most community benchmarks suggest somewhere between 80 and 95 percent of community members never post at all). They’re not going to tell you what scared them off. They’re just going to stay quiet, or leave, or pursue a 1:1 avenue instead. So when you’re thinking about how to respond to a tricky feedback post or a frustrated support thread, think past the person in front of you. Think about who’s reading.

Third: response quality to difficult posts is a brand decision, not a support task. I know that framing makes some people uncomfortable, because it sounds like it’s adding pressure to already-pressured community managers. But consider the alternative framing: every response to a hard post is an opportunity to publicly demonstrate what this community is for. The member who posts “I’ve been asking for this feature for two years and nothing has happened” is handing you a moment. What you do with it is up to you.

The floor matters more than the ceiling. Your community’s best thread is a nice thing to have. Your community’s worst thread is what people make decisions based on.

Next time you’re reading back through your space, try to find the one that makes you wince a little (or a lot). The one where the response was curt, or the feedback got dismissed, or the frustrated member never got a real reply. It’s still there, doing its quiet work. That’s the one worth fixing first.

Right?

Not really. The problem is that by the time your content looks stale, the decay happened a long time ago. You’re working from a lagging record, not a live one. Knowledge decay isn’t just a content problem. It’s a participation problem that eventually shows up in your content. The content is just where it becomes visible.

Outside of all things nerd related these days, I’m working on trying to get the last few details figured out for a new home build. And of all freaking things - water tables and aquifers are very much on my mind.

So we’re gonna run with that analogy.

Think about an aquifer. The water table underneath a region can drop for months, sometimes years, before anyone notices the wells running low. On the surface, everything looks fine. The tap still works. You don’t know the recharge stopped until you’re already in trouble (arguably well into trouble). Community knowledge works the same way. The knowledge base can look intact, even active, while the conditions that keep it alive are quietly disappearing underneath.

And therefore… the leading indicators of knowledge decay aren’t in your content. They’re in your participation patterns.

The first thing to watch is expert withdrawal. Every community has people who actually know things, not just people who post a lot. When those people start going quiet, the knowledge base has lost its primary source of new deposits. The existing content doesn’t change overnight. But the system that generates and validates knowledge has already started shutting down. If your subject matter experts are still showing up in your metrics but only reacting, not responding, that’s worth noticing.

The second signal is question deflection. Pay attention to where hard questions go. When community members stop expecting good answers from the community itself and start saying “DM so-and-so directly” or “ask in the Slack channel,” that’s a participation pattern telling you something. The community is no longer the authoritative source. People still show up, but they route around it for anything that actually matters.

Third, watch where correct answers come from. This one takes some active monitoring, but it’s worth it. Are your best answers coming from community members drawing on lived experience and collective memory? Or are they increasingly citations to external docs, support tickets, and search results? A community that answers its own questions is a recharging aquifer. A community that outsources its answers is running on reserves.

The last one is the hardest to catch because it looks like humility. Watch for hedging in responses. When declarative answers (”here’s how you do this”) start giving way to qualified ones (”I think this used to work, but I’m not sure if it’s changed”), you’re watching knowledge confidence erode in real time. Nobody decided to stop knowing things. The community just stopped being the place where knowing things gets reinforced and updated.

What do you do with this? The honest answer is that recharging a depleted aquifer is much harder than maintaining one that’s healthy. But you can’t fix what you can’t see, and most community managers are watching the wrong layer. Start tracking expert participation rates separately from general participation. Map where complex questions get answered, not just whether they get answered. Notice when your answer quality shifts in register.

You don’t fill an aquifer by updating your FAQ. You fill it by creating the conditions that make expert participation worth sustained investment, and catching early enough when those conditions are starting to fail.

The tap still works. For now.

I’d like to yeet that into the sun.

It treated like a pulse check. If it’s going up, the community is healthy. If it’s going down, something is wrong and you need to fix it.

But hear me out: That number is going to go down no matter what you do, and that this is mostly fine.

Not “fine” in the way we say things are fine when they’re not. Fine in the way that radioactive decay is fine - predictable, measurable, and not actually a crisis once you understand the physics.

When someone joins a consumer community - a fandom, a hobby forum, a subreddit for people who love their air fryers - they’re there because the thing is inherently interesting to them. The participation is the point. There’s no asymptote. They’ll never effectively reach 100%.

But B2B communities (which is my bread and butter) work differently. Your members joined because they needed something: help getting started, answers to configuration questions, workarounds for edge cases, someone to tell them they’re not the only one confused by the onboarding flow. The community is instrumental. It’s a means to competence (which is 100%), not an end in itself.

The moment a member achieves competence, their relationship to the community fundamentally and wildly changes. They stop asking questions because they know the answers. They stop searching the knowledge base because they’ve internalized it. They can troubleshoot solo. They don’t need the community anymore, and in the absence of need, participation drops. Not because you failed them. Because you succeeded. Pat yourself on the back. Buy yourself a cake. YOU DID IT!

(Uh, great. What did I do? Do you have a fork for this cake?)

The half-life of a B2B community member is determined by the length of their learning curve, not the quality of your content calendar.

Every member enters at peak engagement potential - lots of questions, lots of unknowns, lots of reasons to show up. Over time, that potential converts into competence and their active participation drops. Some become lurkers. Most eventually go inactive. And in a growing community, new members are continuously entering at the top of their curve while older members slide toward the asymptote. The denominator keeps growing. The numerator can’t keep pace (unless your company is experiencing wild growth).

If you’re reporting % active members to leadership as a health metric, you’re setting yourself up to spend a lot of time explaining a phenomenon that isn’t a problem.

What’s more interesting to track is engagement density during the learning window. Are new members getting value while the community is most useful to them? Are questions getting answered? Is the knowledge base covering the actual gaps people hit? Is time-to-competence improving? That’s the community doing its job. Raw active percentage is the community aging.

Extending the half-life for members who’ve cleared the learning curve is possible, but it requires genuinely different work. Change logs that surface the right information to the right users. Feature feedback loops that give mature members a reason to show up; not to ask, but to influence. Early access programs. Beta cohorts. The value proposition has to shift from “get help” to “have a voice,” and that shift doesn’t happen through a monthly newsletter blast to your entire member base.

The community doesn’t fail when members graduate. It fails when members arrive and can’t find what they need, when questions go unanswered, when the same confusion cycles through every six months because no one captured the answer the first time.

So the next time someone asks why your active member percentage is declining: the community is working. Members are learning. The curve is doing what curves do.

Then maybe quietly suggest you all agree on a better metric.

You already know what was in it: posts per week, active members, monthly content volume, some variation of “how much stuff is happening in here.” Every single line was about quantity. Not one of them touched quality or consumption.

And y’all know how I feel about those metrics. (See The Metrics We Measure Because We Can)

I’ve been working to change that since the moment my eyeballs landed on it, because this is the thing I keep coming back to: volume tells you that activity happened. It tells you almost nothing about whether any of it mattered.

We’ve built a whole professional reflex around visible participation. Partly because visible participation is what our platforms surface easily. Partly because a graph that goes up is easier to explain than a nuanced story about member satisfaction. And in the earliest days of a community, volume genuinely can be a useful proxy for health. You need to know that anyone is showing up at all.

But communities don’t stay early-stage forever (in fact, early stage is a far shorter phase than you might think). Yet, our metrics often stay stuck in that early-stage phase… for a looooooong time.

Here’s what community maturity actually looks like: members stop asking beginner questions because they already know the answers. They get value from reading, not posting. They show up when something specific prompts them. Conversations get fewer but deeper. Response times slow down because the people responding actually think before they type. (Wild, eh?)

To a volume-first dashboard, this looks like decline. A failure. A proper catasrophe. To someone who understands community dynamics, it can look like exactly the opposite.

A healthy community can be a quiet community, and confusing the two is one of the most expensive mistakes we make.

Think about the neighborhoods you’ve actually wanted to live in. Not the ones where everyone is outside all the time, performing neighborliness, hosting block events every weekend. The ones where you know your neighbors well enough to borrow something without asking twice. Where the group chat is mostly quiet but lights up reliably when someone needs help. That’s not an inactive neighborhood. That’s a neighborhood with trust baked into the infrastructure.

Online communities work the same way. The noisiest ones aren’t always the most valuable. Sometimes they’re the most anxious, the most algorithmically gamed, the most dependent on manufactured prompts to keep the feed moving. What gets mistaken for vibrancy is sometimes just churn.

So what do you do with this practically?

Start with the diagnosis. Before you panic about a dip in posting volume, look at what kind of posting dropped. Fewer people asking for help might mean the knowledge base got better, or members are finding answers without posting at all (which is something I love). Fewer introductions might mean the community skews toward longer-tenured members rather than constant churn. Context changes the story completely, and volume gives you none of it.

Then there’s the leadership conversation. Volume metrics feel safe to report because they’re legible. But you can build the case for quieter indicators without abandoning the dashboard entirely. Member retention, time-to-first-value for new members, the ratio of questions that get answered versus questions that go dark: these tell a more honest story, and most of them can still be graphed. You’re not asking leadership to trust vibes. You’re asking them to trust a better set of numbers.

The harder shift is structural. A community program built around publishing cadences and weekly prompts designed to keep the feed alive is optimizing for volume in ways that actively work against depth. Every low-effort prompt posted to “spark engagement” is also training your members that this is what participation looks like here. That’s a ceiling you set, and it’s genuinely hard to raise later.

And then: get comfortable not knowing and living in the perpetual squishiness. Some of the value your community generates is invisible by design. Members who read a thread and make a better decision at work. Members who avoid a costly mistake because someone posted about it six months ago. Members who feel less alone in a hard job because they know this place exists, even on weeks they don’t log in. You will never fully measure that. Your job is to argue for its existence anyway.

The next time someone pulls up that volume chart looking worried, you don’t have to pretend the number is fine. You can say something more interesting: let’s talk about what the quiet is actually telling us.

That’s not a dodge. That’s the work.

It’s a content strategy we all know super well. So what’s the problem?

The problem is that everyone figured it out at roughly the same time. And now we’re all swimming in correct, (mostly) useful, thoroughly optimized content that says approximately the same things in approximately the same order. Your search results. Your LinkedIn feed. Your newsletter inbox. A sea of answers, and almost no one willing to have an original thought about them.

But you want to know what’s actually scarce? Perspective. Not information. Not even insight, really. Genuine, specific, shaped-by-someone’s-actual-experience point of view. The kind that makes you think “I’ve never heard it put that way” or, more usefully, “I completely disagree with that.”

Which is a problem, because perspective is also kind of terrifying to produce.

Think about the last time you had a hot take at work. Not in your head, not vented to a trusted colleague over Slack, but actually said out loud in a meeting or written down somewhere with your name on it. There’s a particular kind of vulnerability in that moment. You’re not just sharing information. You’re making a claim about how you see the world, and inviting other people to tell you you’re wrong. Most of us have learned, through one too many bruising experiences, to hedge. To qualify. To present “considerations” rather than conclusions.

When the cost of having a public opinion feels high, people stop having them publicly. And what’s left is a lot of very careful content that doesn’t actually say a darn thing.

This is where community comes in, and not in the soft, feel-good way people usually mean when they say “community is important.” I mean structurally. Mechanically. Communities lower the personal cost of having a take.

When you post a question or a half-formed opinion in a community of peers, something specific happens. You get responses. Some of them agree and extend your thinking. Some of them push back and sharpen it. Some of them are wrong in ways that make your original point clearer. The opinion doesn’t just sit there in the void, waiting to be judged. It enters into a conversation. It becomes a contribution rather than a declaration.

That changes the mathematics entirely. It’s the difference between standing up at a podium alone and saying something in a room full of people who are actually present also trying to figure it out. The social context doesn’t just make it more comfortable. It makes it more useful. Your perspective gets tested, refined, and returned to you in better shape than it left.

This is the thing communities offer that no amount of content can replicate. A blog post, a newsletter, an AI-generated summary: they can deliver information efficiently. What they can’t do is witness your specific perspective and respond to it. They can’t tell you where you’re right and where you’re missing something. They can’t turn your half-formed take into a real argument.

For community builders, this reframes the job in ways worth sitting with. If one of the core functions of your community is to give members a low-stakes place to think out loud, then the design question isn’t just “how do we drive more content?” It’s “how do we make it safe enough to say the uncomfortable thing?” That’s a very different problem. It requires different moderation norms, different conversation starters, different tolerance for disagreement. It means occasionally letting a spicy thread run instead of smoothing it over. (I know. That one’s hard. But gosh… I adore spicy threads.)

It also means the communities that will hold value as content gets cheaper to produce aren’t necessarily the ones with the most resources or the most members. They’re the ones where people are actually willing to say what they think. Where the culture has built enough trust that someone will post the thing they’d never put on LinkedIn. Where dissent is treated as a contribution rather than a disruption.

Infinite content raises the floor on information. It makes “here’s what the research says” essentially free. What it can’t touch is “here’s what I actually think about what the research says, and here’s where I think it’s wrong.”

That’s yours. And the communities where people feel safe enough to share it? Those are going to matter more, not less.

Go build the room where someone can say the uncomfortable thing.

The optimistic answer is that it would thrive (said with jazz hands and sparkles). That members would self-organize, hold each other accountable, and the community would demonstrate exactly the kind of organic health that makes leadership teams feel good about the budget line. The pessimistic answer is that it would devolve into chaos within a week (followed by an evil villain laugh). Both of these answers miss the actual answer.

It would still be moderated. Just not by anyone you appointed.

We talk a lot in this industry about “self-governing communities” as if that’s a thing that spontaneously happens when you get the right people in a room and step back gracefully. It doesn’t. Not really. What actually happens when formal moderation disappears is that informal moderation creeps and slithers in to fill the space. Someone sets the tone. Someone decides which questions are worth engaging and which ones get the cold shoulder. Someone’s interpretation of the rules becomes the interpretation, not because they were given that authority, but because they showed up consistently and had the time and inclination to assert it.

Power doesn’t leave when moderators do. It just stops being legible in the way you knew it.

When moderation goes underground, it doesn’t become democratic. It becomes a function of availability and appetite.

Think about what that actually looks like in practice. The member who’s online at 11pm every night and has Opinions about how things should go. The long-tenured member whose contributions are genuinely valuable, but who has also quietly accumulated enormous social influence over what’s acceptable and what isn’t based on their perspective. The person who doesn’t technically break any rules but whose presence shapes what other people feel comfortable saying. These folks exist in almost every community, moderated or not. The difference is whether you’ve thought about them.

Formal moderation structures, for all their messiness and overhead (and yes, there is a lot of both), do one thing that informal power cannot: they make accountability visible. You know who to go to when something goes wrong. You know whose job it is to make a call. You know, at least in theory, what the decision-making process looks like. Remove that structure and you don’t get a flat community. You get a community with a hidden org chart and a concept of guidelines.

This matters for how we think about distributed governance models, member-led spaces, and the general trend toward “lighter touch” community management. None of those things are bad ideas. But they require more intentional design, not less. If you’re going to distribute moderation authority across a member base, you have to think hard about who ends up holding it and why. You have to make the informal org chart legible. You have to build in ways for power to be contested, transferred, and held accountable, because if you don’t, you’re not creating a self-governing community. You’re just outsourcing the governance to whoever wants it most.

Which, to be clear, is sometimes fine. Organic community leaders emerge for good reasons. But “see, it worked out” is not a governance model.

The moderator-free thought experiment is useful precisely because it surfaces what we usually don’t want to look at: the informal power structures that already exist in our communities, right now, running alongside the formal ones. Who in your community is already doing de facto moderation without the title? Whose approval does a new member implicitly need before they feel like they belong? Whose disapproval quietly ends threads?

If you can’t answer those questions, you don’t actually have a picture of how your community governs itself. You have a picture of the part you built intentionally, which is probably not the whole picture.

So needless to say, the goal was never fewer moderators. It was accountability that doesn’t depend on any one person showing up. That’s a much harder design problem, and it doesn’t get easier by pretending the power isn’t there.

It’s just quieter.

It’ll turn into a disaster. What if people say mean things about us or to us? And: We don’t have the bandwidth to manage it.

Both fears are understandable. Both are also, in most cases, solving for the wrong problem.

Let’s start with hostility, because it tends to be the scarier one. This is the:

“Your product sucks and I’m going to let everyone know.”

“Do you people even think before releasing this junk? I’m demanding a refund.”

And so on.

There’s a persistent assumption in organizations that online communities are inherently volatile, that if you give people a space to talk, the worst voices will fill it. And yes, that happens. But it doesn’t happen randomly. It happens by design, even when nobody intended to design it that way.

Hostile communities are almost always communities where hostility works. Where the person who posts the most inflammatory take gets the most replies. Where complaining loudly is the fastest path to getting your problem solved. Where drama generates visibility and visibility generates status. If your community rewards bad behavior with the thing bad actors want most (attention, validation, answers, reputation), you will get more bad behavior. That’s not a community problem. That’s a feedback loop problem, and it starts with the environment you built.

And if this is sounding familiar already - you either have this in your community or you read When Your Community Turns Into a Stage.

So tl;dr… Hostility isn’t a community’s default state. It’s a community’s response to its incentive structure.

This is actually good news, because incentive structures are something you can design. It means thinking carefully about what behaviors your platform surfaces and amplifies. It means not rushing to answer the loudest complaint in the room while quieter, constructive voices go unacknowledged. It means having a communication plan for the moments you know are coming: the product changes, the pricing updates, the industry news that’s going to hit your members’ feeds before it hits your forum. Proactive communication doesn’t eliminate negative sentiment, but it takes away the conditions in which panic and speculation become the loudest voices in the room.

The second fear, the resource question, is where it gets interesting.

Leaders imagine moderation as a full-time job requiring a dedicated team, a ticketing system, and probably a therapist on retainer. (Fair, honestly.) And for some communities… at some moments… that’s true.

But here’s what tends to happen in communities with generally healthy cultures and satisfied members: the members don’t want the ruckus either.

This is the neighborhood watch model, and it’s more common than we talk about. In communities where people have co-built something they value, they show up to protect it. Not because a community manager asked them to. Not because they have a “trusted member” badge. Because they live there, and they care what it feels like. They reply to the off-tone newcomer before the mod team even sees the post. They downvote the bad-faith question into obscurity. They tell someone directly, and often more effectively than any official warning, that this isn’t how things work here.

And honestly: this is such a fun thing to watch (both because it’s behavioral modeling in action… and also watching users say what you can’t say).

But alas, I want to be careful here: I’m not arguing for turning moderation entirely over to the community. That’s a different kind of chaos. Moderation accountability has to live somewhere official, with people who have both the authority and the responsibility to make hard calls or call in backup (security, legal, etc.). But in a healthy community, that team isn’t carrying the weight alone. They’re supported by members who have internalized the culture deeply enough to act as an informal first line of response.

The organizations that figure this out spend less time playing defense. They also tend to have communities that are more durable, because the culture isn’t dependent on any single moderator or policy update. It’s distributed.

Here’s the thing the two fears have in common: both are solved by the same underlying work. Building a community where members are satisfied, where good behavior gets rewarded and bad behavior doesn’t find footing, where the culture is legible enough that people understand what they’re protecting. You don’t get the self-sustaining neighborhood watch without first building a neighborhood worth watching.

That’s the part that requires real investment: not an army of moderators… not a panic response every time someone posts something spicy. It’s the upfront work of being intentional about culture, incentives, and communication.

So next time a leader in your organization raises the hostility fear or the resource fear, here’s your reframe: those aren’t arguments against building a community. They’re arguments for building one carefully. The neighborhood takes care of itself, once you’ve built something the neighbors actually want to protect.

They mean it kindly. They are beaming at you. And somewhere behind your professional smile, a small part of you is screaming. For multiple reasons.

One because we’re terrible at taking compliments.

Two… because you know what runs itself: a rock. A parking lot. A PDF. Communities do not run themselves. They run because someone is running them, and that someone has spent years getting good enough at the job that it no longer looks like a job. Congratulations. You have achieved professional invisibility.

This is not a metrics problem, exactly. We’ve been told for years that the solution is better measurement: track response times, post volume, engagement rates, NPS scores. And look, metrics matter. But dashboards haven’t solved the core issue, because the core issue was never about data. It’s alwways been translation.

The invisibility of community labor isn’t a measurement problem. It’s a legibility problem. The solution isn’t more dashboards; it’s learning to speak your organization’s language instead of asking your organization to learn yours.

Think about the building superintendent. In a well-run building, you never think about the super. The heat works. The pipes don’t rattle. The lobby is clean. The super is invisible by design, because the super is doing their job. But the moment the boiler breaks, suddenly everyone knows exactly who they are and what they do and why it matters.

Community managers are the super. Except we don’t usually have to deal with heating, cooling, or rattling pipes literally and unless things really go sideways leadership doesn’t always fully register that someone is down there running it. The better you perform, the more you disappear. It is a genuinely perverse incentive structure and I think about it more than is probably healthy.

So what do you do with that?

First, stop documenting only what happened and start additionally documenting what didn’t. A post count tells your leadership how many conversations occurred. It says nothing about the conflict that got quietly defused before it became a thread, the new member who almost churned after an awkward first post but didn’t, the question that got answered in four minutes instead of four days. Near-misses are data. Start treating them that way.

Second: stop asking leadership to learn community vocabulary. They won’t. Not because they’re incapable, but because they already have a vocabulary that works fine for them. If your organization is obsessed with churn, your community work is a retention play. If they care about pipeline, your users are warm leads. If they track support ticket deflection, your peer-to-peer answer threads are saving someone real money. The translation is your job. I know that’s annoying. But alas… it’s also true.

Third, make your labor visible before you need it to be. The worst time to explain what community management actually involves is during a budget review or, worse, a reduction in force. By then you’re playing defense. The case for your work - and YOU - should be ambient, woven into regular reporting, mentioned casually in team meetings, present in the room before anyone thinks to question whether you should be.

Here’s the part that’s hardest to say: a lot of us were trained, explicitly or culturally, to make hard work look effortless. To handle the crisis gracefully, clean up quietly, and never let them see you sweat. That norm made us better at the job and worse at keeping it. The community that “runs itself” is also the community that leadership looks at and thinks, hm, do we really need a full-time person or team for this?

The answer is yes. Obviously yes. But invisible labor doesn’t argue for itself. Don’t hide in the boiler room. Go introduce yourself. Tell someone what’s down there.