How do I know who is about to cancel on Skool?

Let us be straight about the framing first, because it changes what you should expect. You cannot know who is about to cancel. Nobody can. There is no reliable model that reads a Skool community and outputs a list of people who will leave next month, and any tool claiming otherwise is dressing up a heuristic in confident language.

What you can do is see drift — a member whose behaviour has changed relative to their own past — and drift shows up weeks before the cancellation. That is a smaller claim and a genuinely useful one, because the entire value is in the timing. A cancellation you see coming three weeks out is a conversation. A cancellation you see on the billing report is a receipt.

What is the actual signal that a member is about to leave?

The signal that a member is about to leave is change, not silence: their activity has fallen off relative to their own past behaviour. The most common mistake is to treat quiet members as at-risk members. They are not the same population and confusing them wastes all your effort.

Plenty of members are quiet forever. They read every post, apply the ideas, never comment, and renew for two years. If you message all of them asking if everything is okay, you have done nothing except make silent people self-conscious.

The signal is a member whose activity has fallen off relative to their own past. Someone who commented three times a week for four months and has posted nothing for three weeks has changed. Someone who has never posted has not changed at all.

Compare each member against themselves, never against the community average.

This also means you cannot do this from a leaderboard. A leaderboard ranks people against each other. It will show you your top contributors and hide exactly the person you needed to see — the member who slid from fifth place to nowhere.

Which behaviours actually mean a Skool member is drifting?

Six behaviours reliably indicate that a Skool member is drifting, and all six are measured against that member's own history rather than the community average. Ranked roughly by usefulness:

1. Drop in posting or commenting relative to their own baseline. The core one. Look for a member whose recent activity is a fraction of their normal rate, over a window long enough to not be a holiday.
2. Stopped opening the community. If someone was a daily presence and has not been seen in ten days, that is a change even if they never posted much.
3. Went quiet after a reply that never landed. A member who asked something and got no answer, then stopped. This one is a wound you inflicted, and it is the most fixable.
4. Stopped appearing in threads they always appeared in. The person who always answered beginner questions, absent from the beginner questions.
5. Reduced depth, not just volume. Long thoughtful comments becoming a thumbs-up emoji is a real signal that most counting misses.
6. Active elsewhere on Skool, silent with you. Not quite the same signal — this is often a member who never built the habit here rather than one who is leaving — but it is worth separating out, because the fix is different. See activating lurkers.

Which churn signals are misleading?

Five commonly cited churn signals are misleading in a Skool community, because each one is normal behaviour for members who are perfectly happy and staying:

• Raw silence. Half your healthy members are silent, permanently, and always have been.
• A single quiet week. People take holidays, get sick, ship projects. Two to three weeks is a reasonable floor before you call it drift in a weekly-rhythm community.
• Not opening emails. Email behaviour tracks email fatigue more than community fit.
• Low leaderboard position. Position is relative to everyone else's activity, so it moves when other people get busier. Useless as a personal signal.
• Not attending calls. Most members of most communities never attend the calls and stay for years.

Which drifting members should I deal with first?

Deal first with drifting members who pay you, who joined recently enough that onboarding is still the explanation, or who used to be heavy contributors. Two members with identical drift patterns are not equally urgent.

• Paying members outrank free members. Obvious, and worth being explicit about because generic risk scoring tends to bury them. In Skoolgrades, buyer status deliberately outranks churn risk when member lifecycle is resolved, so a paying member who is drifting is surfaced as a paying member first rather than being lost in a pile of quiet free accounts.
• Tenure changes the meaning. A newly joined member going quiet after four days is an onboarding failure. A three-year member going quiet is a different, slower story. Lifecycle in Skoolgrades resolves across six lanes and three tenure bands — newly joined at 14 days or less, mid at 15 to 90 days, long-tenured beyond 90 days — for exactly this reason.
• Formerly high contributors outrank everyone. The member who used to carry threads and has gone quiet costs you twice: their renewal and their contribution to everyone else's experience.

How do I check for drifting members each week by hand?

Checking for drifting members by hand takes about twenty minutes a week and works well up to roughly a hundred members. If you are under that size, genuinely do it by hand:

1. Open your members list and sort by recent activity.
2. Scan for names you recognise as previously active who are now near the bottom.
3. Write down three.
4. Message those three, individually, with something specific.
5. Note in a spreadsheet who you messaged and what happened.

That routine outperforms most tooling, because the bottleneck at that size is attention, not data.

Past a few hundred members it stops working, because you can no longer remember who used to be active. The thing you are missing is not analysis — it is memory. That is where a tool earns its place.

How does Skoolgrades spot members who are about to leave?

Skoolgrades reads your community through a Chrome extension running on your own logged-in Skool session, then places every member into a lifecycle state through the Cohort Blender: star, rising, steady, fading, drifted, ghosted, untapped, or going dark. Those states are the memory you do not have. Fading and drifted are the drift signal. Going dark separates the member who is leaving from the one who was never here.

It also resolves member lifecycle across six lanes and three tenure bands, so a drifting newly-joined member is never mixed in with a drifting three-year veteran, and it keeps a churned-member archive that freezes the member's full state at the moment they churned — which is what lets you look back and see what the last ten departures had in common.

Once you have the names, it also writes to them. Build the list, and Skoolgrades drafts each message using what it knows about that member — verbatim to everyone or personalised per member — you approve every draft, and you send from the extension panel one click at a time. Nothing goes out unattended, which is a design decision rather than a gap: a human reading each message is what keeps your account safe and your members' inbox worth opening.

Two honest limits, stated plainly:

• There is no predictive model and no backtest. Snapshot history is deliberately shallow, and long-range trend lines are withheld until there is enough data to justify them. What is surfaced is behavioural: activity falling off relative to that member's own past. Not a probability of cancelling.
• Large communities are sampled. Over about 3,000 active members, collection uses a fixed sample rather than a full census, so figures at that scale are estimates.

What should I say to a member who has gone quiet?

Say something specific from that member's own history and ask nothing else — never open with a reference to their absence. "Haven't seen you in a while!" makes the member feel caught and usually gets a polite lie.

A message that opens on something concrete they told you looks like this:

Hey Dana — you mentioned you were rebuilding the onboarding flow a few weeks back. Did that land?

Ask nothing, sell nothing, and let the conversation go where it goes. Sometimes you get a reply that reveals exactly why they went quiet, which is worth more than the save. Sometimes you get nothing, and that is fine too — the message cost you ninety seconds.

Then log it. Ten of these, with outcomes, and you will start seeing the pattern behind your own churn far more clearly than any dashboard would have shown you. See reducing churn for what to change once the pattern is obvious, and building a churn baseline for how to tell whether it worked.