The words, and what each one changes.

The share of replies written by members rather than by your team.

Below half and the room is a support queue. It is the single most honest number in community work and almost nobody reports it, because it can fall.

Of people who post once, the share who ever post again.

Tests welcome, onboarding and whether the first experience was good, all at once. Usually terrible the first time anybody measures it.

How long a new question waits before any human answers.

Every unanswered question teaches the audience not to ask. The damage lands on the watchers, not the asker.

Proportion of members who posted or replied in a period.

Only meaningful next to churn. Reported alone it flatters growing communities and hides dying ones.

A member who reads and does not post.

The normal majority in every community that has ever existed. Designing to shame them into posting mostly drives them out entirely.

Everybody who joined in the same period, tracked together over time.

The only way to see whether onboarding improved. Blended averages hide a better product behind worse acquisition, and vice versa.

Members who go quiet or leave in a period.

A third of a cohort going quiet within months is ordinary. Communities that never report it cannot explain why a room of ten thousand feels empty.

Something that happens on the same schedule, which members come to expect.

Removes the decision about when to show up. It is what lets a room survive its founders getting busy, and it costs nothing to start.

What happens to somebody in their first hour and first week.

Automating the welcome saves time you do not yet need to save, and teaches new members that nobody is really there.

The written rules, and who applies them.

Every room has unwritten rules. New members break them, feel stupid, and leave. Writing them down is the cheapest retention work available.

Deliberately creating the behaviour you want in the first weeks.

Whatever happens in the first fortnight becomes the norm. Founders posting announcements produces an announcement channel, permanently.

Using the community itself as an acquisition and retention channel.

Works when members have a reason to talk to each other. Slow in a way most budgets do not tolerate, and attribution will never be clean.

A small group given status, access and a recurring task.

People join for status, not merchandise. Twelve engaged people beat two hundred on a list, every quarter.

Staff posting about their work in their own voice.

Resharing company posts does not work: same content, smaller account, overlapping audience. What works cannot be scheduled.

Working with an individual who already has the audience you want.

You are renting trust, not reach. The moment the content sounds like your marketing team wrote it, the trust does not transfer.

How many people see a post without paid distribution.

Follower count stopped meaning distribution years ago. Compare reach against followers on your last ten posts to find your real audience size.

Deliberate actions a reader takes beyond scrolling past.

They survive ranking changes because they require a person to decide something. Impressions do not, which is why impressions are reported.

How often you publish on a channel.

The number people ask about first and the one that matters least. Consistency beats volume, and volume without saves is just noise you paid for.

Sharing that happens where you cannot measure it: DMs, private groups, forwarded links.

Most real recommendation happens here and none of it appears in analytics. Traffic labelled direct is frequently this.

The small number of members producing most of the answers.

Usually under five percent of a room. Losing two of them quietly halves your community, and nobody notices for a month.