Member Retention
How to do a paid community member health audit: a structured process for diagnosing and acting on member health states
When a paid community goes quiet — fewer posts, lower attendance, a creeping sense that the energy has shifted — the most common operator response is a broadcast message to the full member list asking what people want to see more of. This produces responses from the 20% of members who were engaged anyway and near-silence from the 80% who are quiet, which is exactly backwards. The already-engaged members are not the diagnostic target. The quiet members are — and a broadcast message is structurally incapable of reaching them, because quiet members do not respond to broadcast messages. What reaches quiet members is a targeted, personalised approach that treats their silence as a specific problem with a specific cause rather than as undifferentiated “disengagement” that a town-hall question can diagnose. A member health audit is the framework that makes that possible: it segments the member base into four health states, runs a different diagnostic on each one, and produces four different action plans rather than one intervention applied indiscriminately to the full list.
The four member health states and how to define them
Every paying member of your community is in one of four states at any given point in time. The definitions are behavioral, not attitudinal — they are based on what the member has done, not on how satisfied they report feeling. Attitudinal measures (NPS surveys, satisfaction ratings) are lag indicators that capture sentiment after a retention outcome has largely been decided. Behavioral states are leading indicators: a member in the at-risk state is not yet cancelled, but their behavioral trajectory makes cancellation in the next 30–60 days substantially more probable than renewal, and the gap between the behavioral signal and the billing event is the operator’s intervention window.
Activated: A member who has completed a first meaningful value-exchange action within their first week. Activation is not a login event, a channel browse, or a reaction to a post. It is the first action in which the member contributed something to the community rather than consuming it passively: an introduction post in the designated channel, a first substantive reply to another member’s question, a resource shared in a relevant channel, or a first question posed to the group. The specific behavioral definition of activation varies slightly by community type — in a career-advancement community, activation might be the first job-search update posted; in a revenue-focused community, it might be the first tactical question or revenue number shared — but the common feature is that the member has crossed the threshold from observer to participant. Members who activate in the first week are 2–3× more likely to be active at month three than members who do not, which is why first-week activation rate is the most important leading indicator in community health.
Engaged: A member who activated in the first week and has remained behaviorally active in the last 30 days. Active means at least two qualifying events in the last 30 days — posts, replies, reactions, or attendance at live programming. One event per month is a marginal engagement state that is statistically difficult to distinguish from at-risk; two or more events is the threshold at which the member is demonstrably integrating the community into their regular workflow rather than checking in occasionally out of obligation. Engaged members are your most valuable asset and your most reliable referral source. They should not be the target of re-engagement campaigns; they should be the target of upgrade conversations and testimonial requests.
At-risk: A member who was previously activated or engaged but has had no qualifying activity in the last 30–90 days, and is still paying. At-risk is the most actionable health state: the member is still in the community, still being billed, and has demonstrated in the past that they can find value in it. The intervention window is open. The diagnostic question for an at-risk member is not “are they about to cancel?” (the answer is probably yes if no action is taken) but “why did they go quiet, and what would have to be true for them to re-engage?” The answer is almost always one of three things: a life event that temporarily reduced capacity and is now over, a programming or content gap that removed the reason they were posting, or an unresolved friction point (onboarding failure, interpersonal conflict, or channel structure problem) that made continued participation feel effortful without visible payoff.
Churned: A member who has cancelled their subscription, or who has had no qualifying activity in more than 90 days and has not renewed. For communities on monthly billing, churned members are the source of the most actionable data in the audit — not because they can be easily won back (most cannot), but because the pattern across churned members from a given join cohort reveals systematic problems in the onboarding sequence that are currently causing the same failure in newer cohorts. A member who churned is a signal about something that happened 30–90 days before they churned, not about the moment of cancellation.
How to build the segment inventory: the data you need and where to find it
The segment inventory is a spreadsheet with one row per paying or recently-paying member and four columns: name, join date, last qualifying activity date, and current billing status. Building it requires pulling data from two sources: your Slack workspace admin and your billing tool (Stripe, MemberSpace, Memberstack, or whatever you use to collect subscription payments).
From Slack workspace admin, export the member list from the Members section. The export CSV includes name, email, join date, and last active date. “Last active date” in the Slack export is a login event, not a qualifying activity — a member who opened the app to check a notification is counted as active even if they did nothing. You will need to manually check last-post or last-reply date for members whose last login was more than 30 days ago but whose Slack export shows “last active” within the last two weeks. This is tedious but necessary. In most communities under 500 members it takes 45–60 minutes. For the at-risk diagnostic specifically, you need to know what the member’s last substantive action was, not just that they opened the app. The difference matters: a member who opened the app three days ago but hasn’t posted in 60 days is in a very different state than a member whose last post was two weeks ago.
From your billing tool, export the subscription list with join date, billing interval, status (active, cancelled, paused, past-due), and most recent payment date. Match this list to the Slack export on email address. Members who appear in billing but not in Slack have typically been removed from the workspace after cancellation; their billing record is still valid for the churned-member diagnostic. Members who appear in Slack but not in billing are either in a free trial period or are grandfathered guests — exclude them from the health audit unless your community has a significant free-tier population that you are trying to convert.
Once you have the merged spreadsheet, classifying each member into a health state takes 15–20 minutes for a community of 200 members, scaling roughly linearly. The classification rules are: activated-but-not-yet-30-days goes in Activated; active within 30 days and 30+ days since join goes in Engaged; active more than 30 days ago and still paying goes in At-Risk; cancelled or 90+ days inactive goes in Churned. Express each segment as a count and as a percentage of total paying members. This distribution is your baseline. For community health metrics benchmarks by community size and price tier, see the full reference guide.
Running the diagnostic per segment: what questions to ask and what to look for
The segment distribution tells you the current state. The diagnostic tells you why each segment is in that state. The diagnostic is different for each segment, and conflating them — trying to understand at-risk members and churned members with the same analysis — produces conclusions that are true for neither group.
Activated and Engaged diagnostic: The goal here is not to identify a problem but to identify a pattern that you can replicate in newer cohorts. What do your activated members have in common that your non-activated members from the same cohort do not? Common findings: activated members joined in a month when you ran a specific onboarding challenge or cohort-kick-off event; activated members came from a referral channel (they knew another member before joining) rather than a paid or organic search channel; activated members were contacted within 24 hours of joining via a personal DM from the operator or a designated ambassador. Each of these findings has a direct implication for your onboarding sequence — and the fact that they are drawn from your own member base rather than from industry benchmarks makes them more actionable. The benchmark says 55–70% of members should activate in the first week; your historical data tells you which specific onboarding actions produced that rate in your community and which ones did not.
At-risk diagnostic: Pull the at-risk members from your spreadsheet. For each member, note: (1) the month they joined, (2) their last qualifying activity, (3) what that last activity was, and (4) how many months they paid before going quiet. Look for three patterns across the group. First, cohort clustering: if a disproportionate share of your at-risk members joined in the same 1–2 month window, the problem is likely in the programming or content cadence that was running at that time — a period of low programming density, a major community topic shift, or an event that didn’t land well. Second, tenure clustering: if at-risk members cluster at specific tenure marks (3 months, 6 months), this indicates a recurring inflection point in the member lifecycle where the community stops producing visible value. The 3-month mark is the most common: members who activate, spend the first three months in orientation mode, and then go quiet because the orientation content has been consumed and the community has not transitioned them into an execution or accountability mode. Third, activity type: if at-risk members’ last activity was attending a live event (rather than a post or reply), the community may be running too much passive-consumption programming relative to participatory programming. Members who last engaged by watching something are more likely to go quiet than members who last engaged by contributing something.
Churned diagnostic: The churned diagnostic is a cohort analysis, not a member-by-member review. Group churned members by their join-month cohort and calculate the activation rate and first-60-day retention rate for each cohort. Look for the cohort months with the worst activation rates — these are the months where something in the onboarding sequence failed systematically. Also look for the months with the best first-60-day retention rates and compare them to what was different about the community programming, onboarding, or acquisition channel during that period. The goal is to find at least one “what was different about the good months” insight and at least one “what was different about the bad months” insight. These two findings anchor the action plan for the next 90 days: replicate the good-month condition for incoming cohorts, and remove or replace the bad-month condition. See the paid Slack community churn rate guide for the full cohort-analysis framework.
Taking targeted action: re-engagement templates, win-back messages, and exit surveys
The audit is only valuable if it produces targeted action. Broadcast messages to the full member list are not targeted action. A broadcast message announcing a new feature or a re-engagement challenge reaches the engaged members (who don’t need it), the at-risk members (who may open it but rarely act on it), and the recently churned members (who are no longer in the community and don’t receive it). It does nothing for the segment that matters most in an audit: the at-risk members who are still paying but are one more quiet month away from cancellation.
For at-risk members, the action is a personal DM from the operator within two weeks of running the audit. The message should not look like an automated re-engagement campaign. It should be written in the operator’s voice, reference something specific about the member’s prior activity, and ask one diagnostic question rather than making a pitch. A template that works consistently:
Hi [Name] — I noticed you were active in [channel/thread] back in [month], and I haven’t seen you around lately. Is [the goal they stated when they joined, or the type of problem they were working on] still something you’re focused on? I want to make sure the community is actually useful for where you are right now, not just where you were three months ago.
This message does several things correctly. It names something the member actually did (not a generic “we miss you”). It asks a yes/no diagnostic question rather than an open-ended request for feedback that creates decision friction. It frames the operator’s goal as making the community useful rather than preventing a cancellation. The response rate on this message for at-risk members who have been inactive 30–60 days is 35–55% when the specificity is real. The response rate drops sharply if the specificity is fabricated or vague — “you were active a while ago” produces far lower response than “you were active in #career-transitions in March.”
For churned members — members who have cancelled within the last 4–6 months — a win-back message is appropriate in two cases: if your cohort diagnostic identified a specific onboarding failure that has since been fixed (and you can name the fix concretely), or if you have added a specific new programming element that maps directly to the outcome the churned member originally signed up for. The win-back message should acknowledge the cancellation directly, name the specific change, and offer a low-friction re-entry path (a free 30-day trial or a single-event invite, not a discount on the subscription). Members who cancelled because of a recognised failure and who are offered a specific, relevant fix are responsive to win-back at a meaningful rate (15–25% trial re-entry for a well-targeted list). Members who cancelled for reasons unrelated to a fixable failure are not worth a win-back attempt — the economics do not support the time cost.
For very recently churned members — members who cancelled in the last 30 days — an exit survey is more valuable than a win-back attempt. A three-question exit survey sent within 72 hours of cancellation produces the most honest feedback you will get from any member. The three questions: (1) “What was the main reason you cancelled?” with five options (too expensive / not using it enough / the content wasn’t right for me / a life change / other). (2) “Was there one thing we could have done that would have changed your decision?” (open text, optional). (3) “Is there any chance you’d rejoin in the future?” (yes / maybe / no). This survey, sent by email from the operator personally (not from an automated billing tool), gets 40–60% response rates from members who cancelled within the last week and 20–35% from members who cancelled 2–4 weeks ago. The responses to question 2 are the most actionable data in the audit: they tell you the specific intervention that was closest to being on the table before the member made the cancellation decision.
When to run the audit and how often
The first audit should be run as soon as you have at least 60 days of member history — enough for the first cohort to have completed the activation window and for at least some members to have moved through the at-risk state. Earlier than this, the sample sizes in each segment are too small to support cohort-pattern analysis and the action plans you produce will be speculative rather than evidence-based.
For communities under 300 members, a quarterly cadence is appropriate: three months of new cohort data is enough to identify emerging patterns without running the audit so frequently that the action plans from one quarter are not yet implemented when the next audit begins. The rule of thumb is that an audit should not be run until the actions from the prior audit have been executed and have had at least 30 days of data to evaluate. If the at-risk DMs from the last audit are still being sent, wait.
For communities above 300 members, a monthly cadence for the at-risk segment (the most time-sensitive segment) and a quarterly cadence for the full cohort analysis is appropriate. The at-risk segment changes every month: members who were at-risk last month may have re-engaged (and should be reclassified as engaged) or may have cancelled (and should move to churned). Monthly at-risk reviews prevent the 30-day intervention window from closing while the operator is waiting for the next quarterly audit.
The best operational home for the audit is a recurring calendar block on the last day of the month, blocked as a 2-hour working session. The output of the session is: (1) updated health distribution with month-over-month delta, (2) at-risk DMs sent or queued for the next 5 business days, (3) one cohort finding noted for the quarterly review, and (4) one update to the onboarding sequence if the at-risk or churned diagnostic produced a clear fixable signal. Communities that run a monthly at-risk review consistently for 6+ months typically see their at-risk proportion fall from the 25–40% range (typical for communities that have never run a structured audit) to the 10–20% range, with a corresponding 8–15 percentage-point improvement in month-3 renewal rate.
Frequently asked questions
How do you audit a paid community?
A paid community audit has four steps. First, segment your current member base into four health states — activated, engaged, at-risk, and churned — using behavioral data from Slack workspace admin and your billing tool. The classification rules: activated means a member completed a first meaningful value-exchange action (introduction post, first answer, first reply) within the first week; engaged means a member has been active in the last 30 days after activation; at-risk means a member was previously activated or engaged but has had no activity in 30–90 days and is still paying; churned means a member has cancelled or had no activity in more than 90 days. Second, calculate the size of each segment and express it as a percentage of total paying members to get your health distribution. Third, run a segment-specific diagnostic on each group: for at-risk members, look for cohort and tenure clustering; for churned members, run a cohort-activation-rate analysis; for engaged members, identify what they have in common that non-activated members do not. Fourth, take targeted action per segment: personal DMs for at-risk members, win-back messages for recently churned members with a relevant new offer, exit surveys for members who cancelled in the last 30 days.
What is a community health audit?
A community health audit is a structured diagnostic process that segments a paid community’s member base by behavioral engagement state and runs targeted analysis on each segment to identify root causes of disengagement. It is distinct from a routine analytics review in three ways: it is segment-specific rather than aggregate, it is root-cause-oriented rather than descriptive, and it produces targeted actions for each segment rather than a single intervention applied to the full member base. A community health audit is typically run quarterly by community operators who want to move from reactive management (responding to individual cancellation requests) to proactive management (identifying and acting on the precursors to cancellation before the cancellation request arrives).
How do you re-engage inactive paid community members?
Re-engaging inactive paid community members requires different approaches depending on how long the member has been inactive and what they did before going quiet. For at-risk members — inactive 30–90 days but still paying — the highest-success approach is a personal DM from the operator that references something specific the member did before going quiet and asks one diagnostic question about whether their original goal is still active. This message has a 35–55% response rate when genuinely personalised. For recently churned members, a win-back message is appropriate only if you have a concrete, relevant change to offer — a fixed onboarding failure or a new programming element that maps to their original goal. Do not send win-back messages as general re-engagement campaigns; send them only when you have something specific and relevant to say. For very recently cancelled members, an exit survey sent within 72 hours produces 40–60% response rates and provides the most honest, actionable feedback you will get from any member.
What does a healthy paid community look like?
A healthy paid community has four measurable characteristics. A high new-member activation rate: 55–70% of members in the most recent join cohort should have completed a first meaningful value-exchange action within their first seven days. A low at-risk proportion: no more than 20–25% of currently paying members should be classified as at-risk (30–90 days inactive). A stable month-3 renewal rate: healthy paid communities in the $49–$199/mo tier have month-3 renewal rates of 70–80%; below 65% indicates a systematic onboarding or ongoing-product problem. A manageable monthly churn rate: 3–5% for communities under 500 members, 2–4% for communities above 500. Communities that run a quarterly health audit consistently for 2+ years typically operate in the upper range of each benchmark — not because auditing is magic, but because it creates the systematic feedback loop that surfaces problems when they are still fixable rather than after they have compounded into an elevated churn rate across multiple cohorts.