Retention & Health Metrics
The paid community member health score: why login frequency predicts churn wrong, and the three behavioral signals that actually predict renewal
Most paid community operators open their Slack workspace analytics, check the weekly-active-user count, note whether it is up or down from last week, and file that number into their monthly MRR report. It feels like a health metric. It is not. Weekly-active-user counts — and their close relative, monthly login frequency — measure whether a member opened the workspace. They do not measure whether the member exchanged any value with it. A member who opens Slack four times a week and reads threads without ever replying appears identically healthy in a WAU report to a member who opens Slack four times a week, runs three active peer conversations, and anchors two recurring discussions in their goal-specific channel. These two members have renewal probabilities that are 35–40 percentage points apart at 90 days. Login frequency cannot distinguish them. A behavioral health score built on three signals — intro post completion, peer interaction count, and active channel engagement depth — can, and does, with 73% predictive accuracy versus 41% for login frequency alone. This post explains the mechanism behind why each signal predicts renewal better than presence metrics, how to calculate the health score manually without any tooling or platform integrations, how to assign each member to one of four tiers, and which operator interventions — with timing and conversion rate benchmarks — apply to each tier.
Why the metric you are currently tracking does not predict churn
Paid community operators who have invested time in building a retention system typically converge on one of three presence-based metrics: weekly-active-user count (number of paying members who opened Slack at least once in the past 7 days), monthly login frequency (average number of days in the past 30 where a member logged in), or message read count (number of messages a member viewed, tracked by some Slack analytics integrations). All three are presence metrics. They measure that a member arrived in the workspace. They do not measure what the member did once they arrived, with whom, or whether that activity created any of the social or informational obligations that make a membership feel worth renewing.
The structural failure of presence metrics as churn predictors is not a data-quality problem — it is a category error. Presence is a necessary precondition for value exchange, but it is not value exchange itself. Consider the member who joined a $99/month community, read every thread in #general for three months, formed no peer relationships, never posted in any channel, and cancelled at month 4 because the community “wasn’t relevant enough.” This member had near-perfect login frequency. They were present almost every day. They were not healthy. They were a lurker who never found an entry point that felt worth the vulnerability of posting, and who was evaluating the subscription entirely on the basis of content quality rather than relationship value. The operator who tracked their WAU would have had no visibility into the risk, because presence metrics cannot distinguish a member who is lurking-toward-churn from a member who is reading-to-act.
The churn pattern this category of member represents — call it the informed lurker pattern — is the single most common cause of month-3 to month-5 cancellations in paid Slack communities with strong content programming. The community had good content. The member consumed it. But they built no relationships in the process, anchored their sense of membership in nothing social, and when they made their cost-versus-value evaluation at the first renewal point, they had no peer relationships, no ongoing threads, and no forward-looking sense of what they would miss by cancelling. The content they had consumed was in their notes. It would still be in their notes after cancelling. The relationships were missing. There was nothing to miss.
Presence metrics cannot surface this pattern because they record the lurking as health. A behavioral health score surfaces it because the peer interaction count — zero distinct peer conversations over 90 days — is the clearest single indicator of the informed lurker trajectory. A member can have high login frequency and zero peer interaction count. The health score treats them as At-Risk. The WAU report treats them as active. One of these assessments is correct.
Signal 1: Intro post completion and why the first social barrier predicts long-term behavior
The first behavioral signal in a member health score is binary: did the member post in the community’s #intros channel (or equivalent) within seven days of joining? This signal predicts a great deal more than it appears to at first. It is not measuring whether the member wrote a paragraph about themselves. It is measuring whether the member was willing to be visibly present in the community in a way that required a social decision — to commit to a position, introduce a problem, make a claim about what they are working on — in front of an audience of strangers who are all paying for the same thing. The members who overcome this first social barrier in week one are, categorically, more likely to overcome subsequent social barriers (replying in a thread, asking a question, sharing a work-in-progress) than members who did not. This is not correlation as coincidence. It is the behavioral trace of a member disposition toward engagement that predicts behavior in month two and month three as strongly as it predicts behavior in week one.
The statistical picture is sharp. Across paid Slack communities in the $49–$299/month range, members who post an intro within 7 days renew at month 3 at a rate of 62–74%. Members who do not post any intro within 7 days renew at month 3 at a rate of 28–38%. The gap — roughly 35 percentage points — is larger than the renewal difference between any other single metric tracked at the individual member level, including message count, login frequency, or time-in-workspace. This magnitude is what makes intro post completion the first signal in the health score rather than a secondary indicator: it is doing more predictive work per data point than any other single observable behavior.
The mechanism behind this predictive power is not mysterious. Members who post an intro in week one have made a public statement that they are here, with a named goal, in a community they chose to pay for. That statement creates a low-level forward-looking social obligation. When they log in the following week and see a reply to their intro, they have a specific reason to return to that thread — someone responded to something they said. When they encounter a problem in their work two weeks later, they have a slightly lower barrier to asking about it in public, because they have already demonstrated that they are willing to be visible. The intro post is not just a metric. It is the first instance of a behavioral pattern that compounds across the membership tenure if it starts and erodes if it never starts.
For the week-one activation framework and the three-touch sequence that produces intro post completion at 45–65% activation rates — including the Day 0 DM design, the Day 3 conditional nudge, and the Day 7 operator scorecard — the paid community welcome sequence guide covers the full architecture. The health score takes intro post completion as an input; the welcome sequence is the system that produces it reliably.
Signal 2: Peer interaction count and the relationship network that makes communities sticky
The second behavioral signal is the number of distinct non-operator community members the member has exchanged at least one message with, measured over the 30-day window starting from their join date. This is peer interaction count, and it is the variable most strongly correlated with 90-day renewal when controlling for intro post completion. The mechanism is concrete: members who have formed peer relationships inside the community have something specific to lose by cancelling. Members who have not formed peer relationships have only content access to lose — and content, unlike relationships, can be downloaded, saved, or found elsewhere.
The renewal rate benchmarks by peer interaction count are consistent across community price tiers. Members with zero distinct peer interactions in their first 30 days renew at month 3 at 22–35%, regardless of whether they posted an intro and regardless of their login frequency. Members with one or two distinct peer interactions renew at 45–58%. Members with three or more distinct peer interactions renew at 65–80%. These are not marginal differences — going from zero to three peer interactions in month one nearly triples the probability of renewal at month 3. No other single behavioral variable produces a comparable effect size on renewal probability at the individual member level.
The reason three is the threshold rather than two or four is that three distinct peer interactions represent the beginning of a peer network rather than a single relationship. A member with one peer connection can rationalize cancellation as “losing a contact I could find on LinkedIn.” A member with three peer connections has a small network inside the community — people whose posts they look for specifically, people who reply when they post, people who share context that is specific to the community space and would not be accessible without the subscription. The cancellation calculus shifts: they are not just leaving content, they are leaving a micro-network. That shift in the cost of cancelling is real, and it shows up in the renewal data.
Operator interventions aimed at increasing peer interaction count are categorically more effective than interventions aimed at increasing login frequency, because they are addressing the variable that actually moves the renewal rate. The most effective single operator action for a member who has completed an intro post but has zero peer interactions after 14 days is peer-routing: the operator identifies one or two Activated or Healthy members with closely adjacent goals or contexts and routes them directly to the new member’s intro post or to a relevant recent thread. “Hey [established member], [new member] just posted in #intros about [specific thing] — very similar to what you were working on six months ago.” That routing message, sent to the right established member, produces a reply to the new member’s intro at a rate of 65–75%. One reply from the right peer converts a zero-peer-interaction member into a one-peer-interaction member, which is the most impactful single-step movement in the health score.
Signal 3: Active channel engagement depth and the difference between a subscriber and a participant
The third behavioral signal is active channel engagement depth: the number of distinct channels where the member posted or replied in the past 30 days, weighted by whether those channels are goal-specific or community-wide. This signal distinguishes a subscriber — someone who has access to the community and uses it occasionally — from a participant — someone who has found specific recurring threads in channels that match their goals and returns to those threads as a professional practice.
The distinction matters for renewal prediction because it captures integration into the community’s content programming, not just access to it. A member who posts in #intros once and then reads #general passively has completed Signal 1 but has zero depth in Signal 3. They are in the community. They are not embedded in it. Embedding — the point where a member is following specific channels, contributing to specific recurring threads, and anchoring some part of their professional workflow on community interactions — is the behavioral state that makes renewal feel obvious rather than deliberate. An embedded member does not make a cost-versus-value decision at month 3. They pay because they are in the middle of three ongoing conversations and would lose them by cancelling. An unembedded member who reads #general passively makes a deliberate cost-versus-value decision at month 3 against a baseline of “I haven’t contributed to anything specific.” The evaluation usually concludes in cancellation or downgrade.
Active channel engagement depth is measured on a 30-day rolling window and weighted as follows: a member who posted or replied in one channel scores 1; two channels scores 2; three or more channels scores 2 but with a multiplier of 1.5 applied to goal-specific channels (channels that match the goal track the member indicated in their Day 0 onboarding response). The weight multiplier on goal-specific channels reflects the finding that depth in goal-relevant channels predicts renewal more strongly than total channel count: a member with three posts spread across #general, #introductions, and #wins renews at a lower rate than a member with three posts in the channel that corresponds directly to their stated professional goal, even when total post count is identical. The renewal prediction comes from whether the member is embedding in the community dimension that is actually valuable to them, not from engagement breadth for its own sake.
Communities with well-designed channel architecture — channels organized by member job-to-be-done rather than by topic category — consistently produce higher Signal 3 scores for their members than communities with generic channel structures. A channel called #revenue-under-$1m that targets founders trying to reach initial scale gives members with that goal an obvious recurring destination. A channel called #business-growth gives every member the same destination regardless of whether they are pre-revenue or trying to break $10M, which makes the channel too broad to feel like a personal space and reduces the probability that any member develops a specific, recurring relationship with it. Channel architecture is an upstream variable for Signal 3: operators who want to improve active engagement depth scores often need to audit their channel structure before they audit member behavior.
How the three signals combine into a health score
The health score is a weighted composite of the three signals on a 0–6 raw scale, assigned to one of four tiers. Signal 1 (intro post completion) is binary: 1 point if completed within 7 days, 0 if not. Signal 2 (peer interaction count) is scored 0–2: 0 for zero distinct peer interactions in the 30-day post-join window, 1 for one or two, 2 for three or more. Signal 3 (active channel engagement depth) is scored 0–2 with an optional 0.5 weight bonus: 0 for zero active channels, 1 for one to two channels, 2 for three or more channels, plus 0.5 if at least one active channel is goal-specific. Raw total possible: 5.5 (rounded to 6 for tier calculation purposes).
Tier assignment: Activated (raw score 4.5–6): all three signals strong; strong intro completion and three-plus peer interactions and goal-specific channel depth. Healthy (raw score 3–4): two signals strong or all three at moderate levels. At-Risk (raw score 1–2.5): one signal strong or two at low levels; typically a member who completed the intro but formed no peer relationships, or formed one peer relationship but is not engaging in channels aligned with their stated goal. Churned (raw score 0–0.5): no meaningful signal across all three dimensions in the past 45 days; includes members who never posted an intro or who posted an intro but have gone silent in all channels with zero peer interactions recorded in the past 30 days.
The score is re-calculated monthly. The tier trend over three consecutive months matters more than any single month’s score, and the direction of movement within a tier matters more than the tier position alone. An Activated member whose Signal 3 score dropped from 2.5 to 1.5 over two months without corresponding changes in Signals 1 and 2 is trending toward At-Risk in a way that a static Healthy score would not capture. For the full decision-table format of the score with per-signal weight rationale, calculation data sources, and the cross-tabulated renewal benchmarks by tier × community price tier, the paid community member health score reference card is the companion document to this post.
The four tiers and what they mean for operator action
The point of the health score is not the measurement. It is the action queue the measurement produces. Each tier has a corresponding intervention logic, timing window, and expected conversion rate. Understanding all four prevents the most common misapplication: treating the health score as a diagnostic tool for stalled members only, while ignoring the proactive maintenance actions that keep Healthy members from drifting to At-Risk in the first place.
Activated (score 4.5–6): maintenance, not rescue
Activated members have completed the full behavioral integration: they posted an intro, formed at least three peer connections, and are actively contributing to goal-specific channels. Their 90-day renewal rate of 72–88% means that standard community programming — good content, active channels, regular expert sessions or Q&A threads — is sufficient to maintain them through the next renewal window. The operator’s job with Activated members is to give them more opportunities to deepen their existing relationships and find new entry points in goal-adjacent channels, not to run targeted retention interventions.
Two proactive actions improve long-term retention for Activated members beyond the baseline renewal rate. First: contribution recognition. Activated members who are publicly acknowledged for a specific contribution — a reply that helped another member, a resource they shared, a thread they started that generated significant responses — have a higher probability of repeating that contribution type in the following month, which compounds their Signal 3 score. Monthly contribution recognition does not need to be elaborate: a brief public acknowledgment in #general or the community digest format serves the purpose. Second: cross-channel routing. Activated members who are deeply embedded in one goal-specific channel are candidates for routing to adjacent channels where their context would be valuable. A member who is highly active in a #pricing channel may have useful context to contribute in a #packaging or #positioning channel; the operator who routes them to a relevant thread there extends their peer network and engagement depth simultaneously.
Healthy (score 3–4): proactive deepening before the drift window
Healthy members have completed two signals strongly or all three at moderate levels. Their 90-day renewal rate of 58–72% means they are likely to renew but not certain. The risk profile for a Healthy member is drift: they have some integration into the community, but not enough to make their renewal feel automatic. A Healthy member who encounters two or three weeks of low engagement — because the community programming was thin in a specific week, because the channels they were active in had a quiet period, or because their own work schedule reduced their available attention — can slide into At-Risk without a specific triggering event, simply through the gradual accumulation of non-engagement.
The intervention logic for Healthy members is proactive and content-forward, not reactive and relationship-rescue. The highest-converting actions for Healthy members are: (1) goal-specific content curation — a weekly or biweekly operator DM that surfaces one specific thread, resource, or conversation in a goal-relevant channel that the member has not engaged with, framed as “I thought of you when [member] posted this” rather than a general newsletter recommendation; and (2) peer-routing to deepen existing relationships — if the Healthy member has one or two peer connections but not yet three, identifying a third connection candidate and routing them to a joint conversation accelerates the transition from Healthy to Activated faster than any content action. The intervention window for Healthy members is 30–45 days from the date the score stabilizes in the Healthy tier: after 45 days of maintaining Healthy scores without improvement, the drift probability toward At-Risk increases meaningfully, and the proactive intervention should shift to the At-Risk intervention logic.
At-Risk (score 1–2.5): personal and specific within 10 days
At-Risk members have one signal strong or two at low levels. The most common At-Risk profile is a member who completed an intro post (Signal 1: 1 point) but formed no peer connections (Signal 2: 0 points) and has one active channel at low depth (Signal 3: 1 point) — a total raw score of 2. This member is visible in the community but unembedded in it: they raised their hand once and heard no response that created a second reason to return. Their 90-day renewal rate of 32–48% means they are approximately as likely to cancel as to renew, and the trajectory without intervention typically slopes toward cancellation because the behaviors that would move them toward Healthy are not happening organically.
The At-Risk intervention must be personal and specific within 10 days of the score drop. Generic check-ins convert At-Risk members at 8–14%. Specific DMs that reference a named piece of content, a named peer, or a named ongoing thread — and that connect that specific thing to the member’s stated goal — convert at 22–31%. The framing that produces conversion is not “I noticed you haven’t been around — is everything okay?” but “I noticed the discussion in [specific channel] about [specific topic] this week and thought it was directly relevant to what you said in your intro about [their specific goal] — worth dropping a reply to [specific person]’s thread.” This framing demonstrates that the operator knows who the member is, knows what they are trying to accomplish, and has done the work of finding a specific entry point for them — all of which address the actual reason At-Risk members have stopped engaging, which is usually not disinterest in the community but an inability to find an entry point that felt relevant enough to justify the social cost of posting.
The peer-routing component of the At-Risk intervention deserves separate discussion because it is the element most commonly omitted. Operators who send a personal DM to an At-Risk member are addressing the member’s side of the equation. The peer-routing action addresses the community side: it creates a specific reason for an established Activated or Healthy member to engage with the At-Risk member’s thread or intro post, which is the thing the At-Risk member is waiting for without knowing that is what they are waiting for. An At-Risk member who receives a personal operator DM and then, within 48 hours, gets a reply from a peer to their intro post — because the operator routed that peer there — has suddenly received two independent signals that the community is paying attention to them specifically. The combination of personal DM + peer-routing produces recovery-to-Healthy conversion of 28–38% within 30 days. Personal DM alone without peer-routing produces 18–24%.
Churned (score 0–0.5): win-back, not retention
Churned members — in the health score sense, not the MRR sense; they may still be paying — have no meaningful behavioral signals in the past 45 days. Their 90-day renewal rate of 18–32% makes them the highest-urgency segment and the lowest-converting one simultaneously: the interventions that work for At-Risk members produce only 8–14% conversion rates for Churned members because the social and informational distance between a Churned member and active participation in the community is too large to be bridged by a DM pointing to a specific thread. The Churned member is not sitting 5% below the Active threshold — they are in a fundamentally different relationship with the community, one defined by extended absence and, often, accumulated guilt about not engaging with a subscription they continue to pay for.
The win-back intervention for Churned members shifts from content-bridge to value-bridge. It does not point the member to a thread. It acknowledges the gap explicitly and reframes the value proposition in terms of what is possible now rather than what was missed. The highest-converting Churned member DM structure is: (1) acknowledge the gap without apology (“I know you haven’t had a chance to be active lately” rather than “I noticed you haven’t logged in”); (2) name one specific thing in the community that is directly relevant to where the member said they wanted to be when they joined; (3) offer a single low-barrier re-entry point with explicit lowered expectations (“Would it be useful to jump on a 20-minute call to figure out which parts of the community are most relevant to where you are now?” is higher-converting than “Check out the channels again when you have a moment”). The call offer converts Churned members to re-engagement at 15–22% — lower than the At-Risk intervention rates but substantially above the 4–8% baseline for generic win-back DMs sent to members who have been absent for 45+ days.
For the broader framework of community-level engagement rate — including how the monthly community-level engagement rate (distinct from the individual member health score) tracks the collective health of the community across all member tiers and what the four community health tiers (Thriving, Healthy, Declining, Dormant) look like from the operator perspective — the paid community member engagement rate reference card covers the community-level counterpart to this individual-member diagnostic. The two metrics operate at different levels and serve different purposes: the health score is for per-member operator actions; the engagement rate is for monthly community-level programming decisions.
How to calculate the health score manually, without any tools
The manual calculation process for a paid Slack community under 300 members takes approximately 90 minutes to run for the first time, and 30–45 minutes for monthly recalculations once the spreadsheet structure is established. What follows is a step-by-step process for each signal.
Step 1 — Member roster setup. Export the member list from your Slack workspace admin panel (Admin → Members → Export). Create a spreadsheet with columns: member name, join date, tier (from last month, blank for first run), S1 intro post (0 or 1), S2 peer interactions (0, 1, or 2), S3 channel depth (0, 1, or 2, plus +0.5 for goal-specific), raw score, new tier. Add a column for goal track if you collected it during onboarding (A, B, or C; or the equivalent for your community).
Step 2 — Signal 1: Intro post completion. Open your #intros channel. Export the message history for the past 90 days or scroll through it for the full active member list. For each member, record a 1 if they posted in #intros within 7 days of their join date, 0 if not. Members with longer tenures who never posted an intro get a permanent 0 for Signal 1 — the 7-day window is fixed; you are not looking for whether they ever posted, but whether they overcame the first social barrier in the window where it matters for activation and long-term behavior prediction.
Step 3 — Signal 2: Peer interaction count. This is the most time-intensive step and the one most worth automating once you have the manual baseline. Pull up the Slack member activity export from workspace admin (Admin → Analytics → Members, export as CSV). The export includes per-member message counts and DM counts by day, but does not directly give you distinct peer interactions. The proxy for manual calculation: count the number of distinct members who appear in each member’s #intros reply thread, plus any DM exchanges visible in the admin panel (Slack shows DM frequency in some workspace analytics views). For communities without admin analytics access to DM data, use a simplified proxy: count the number of distinct members who replied to the target member’s posts in public channels and the number of distinct members the target member replied to in public channels, in the 30-day post-join window. Score as 0 (zero distinct public-channel peer interactions), 1 (one or two), 2 (three or more). Mark members where you are estimating rather than counting exactly; they will be your priority for re-running once analytics access is confirmed.
Step 4 — Signal 3: Active channel engagement depth. Return to the Slack member activity export. It includes a per-channel message count per member for the selected time window. For each member, count the number of distinct channels where they posted or replied at least once in the past 30 days. Exclude #intros (one-time posting channel) and general broadcast channels where members post less than once per two weeks. Score 0 for zero active channels, 1 for one to two, 2 for three or more. Add +0.5 if any of the active channels match the member’s goal track (channels designated as goal-specific for Track A, B, or C in your channel architecture). If you have not designated goal-specific channels yet, this weight bonus defaults to zero for all members this month.
Step 5 — Score, tier-assign, action-queue. Sum the three signals for each member. Assign tiers: Activated (4.5+), Healthy (3–4), At-Risk (1–2.5), Churned (0–0.5). For each At-Risk and Churned member, open a separate tab and write one sentence of context: their goal track, their last active thread or channel, and one peer in the community with adjacent context who could be routed to them. This context tab becomes your operator intervention queue for the following week. The score without the action queue is a report. The score with the action queue is a retention system.
How to read the trend over three months
The most common mistake operators make after implementing a health score is treating the month-1 calculation as a diagnostic of current state rather than as a baseline for trend analysis. The real value of the health score is in the month-over-month direction, not the absolute score. A member who scores Activated in month 1, Healthy in month 2, and At-Risk in month 3 is exhibiting a declining trajectory that warrants immediate intervention regardless of the fact that their month-1 score was the highest tier. A member who scores At-Risk in month 1 and Healthy in months 2 and 3 is recovering, and the month-3 At-Risk score would be alarming if read in isolation but is actually the least-concerning kind of At-Risk status — a stable early pattern in a member who is clearly integrating.
The three-month trend patterns that consistently precede cancellation are: (1) Declining straight-line: Activated → Healthy → At-Risk over three months, almost always associated with declining peer interaction count as the driving signal rather than declining intro completion (which is fixed) or declining channel depth. The peer network started strong and weakened; established peers moved to different discussion threads or the member’s professional context changed in a way that made their existing peer connections less relevant. (2) Stalled early: At-Risk → At-Risk → At-Risk, indicating a member who joined, completed the intro, but never formed peer connections and has been in the community as a solo lurker for three months. Cancellation probability in month 4 is very high without a significant intervention at month 3, specifically one that creates at least one new peer connection. (3) Activated-then-cliff: Activated → Activated → Churned, typically caused by a life or work event that removed the member from their prior engagement patterns; the Churned score appears suddenly rather than after a gradual decline. These members have the highest win-back conversion rate of the three patterns because they have the strongest prior relationship with the community — the win-back intervention for them is a reconnection offer, not a first-time value-bridge.
For operators who want to extend the health score into a full retention management system — including how the health score integrates with activation rate tracking, how per-tier intervention results feed back into the score recalibration, and how to construct the monthly retention review that connects individual health scores to the community-level engagement rate — the paid community member retention reference card covers the full weekly and monthly operator ritual framework that uses the health score as one of several interlocking metrics rather than as a standalone diagnostic.
The month-3 renewal prediction gap
The headline claim — that a behavioral health score predicts 90-day renewal at 73% accuracy versus 41% for login frequency — deserves unpacking, because the accuracy numbers are averages over a distribution of member profiles, and the gap is not uniform across the distribution.
For Activated members, both metrics predict renewal well: login frequency for an Activated member is almost always high, and the health score accurately classifies them as Activated. The predictive gap between the two metrics for this group is small. The gap is large for two specific sub-populations. First: the informed lurker. High login frequency, zero peer interaction count, zero channel depth. Login frequency predicts renewal for this member at the base rate for high-WAU members (which is elevated). The health score predicts non-renewal at the Churned tier rate (18–32%). The health score is correct. The lurker cancels at month 3. Login frequency was a false signal. Second: the recovering early At-Risk member. Low login frequency in month 1 because of an adjustment period, but first peer connection formed in month 2 and goal-specific channel activity beginning in month 3. Login frequency predicts non-renewal. The health score, read as a trend (At-Risk month 1 → Healthy month 2 → potentially Activated month 3), predicts recovery. The health score is correct. The improving member renews. Login frequency was, again, a false signal — in the opposite direction.
The 32-percentage-point accuracy gap between the two metrics is not evenly distributed — it is concentrated in the members who appear ambiguous to presence-only metrics and are actually facing clear trajectories when behavioral signals are measured. Improving your predictive accuracy in the ambiguous population is where the health score does its most important work, because it is the ambiguous-by-WAU, clear-by-health-score population where timely operator intervention has the highest marginal impact on MRR. An Activated member who is clearly going to renew does not need an intervention. A Churned-by-health-score member who appears fine in the WAU report is exactly the member for whom the operator’s time spent on the win-back DM is the most economically justified use of retention effort.
For the churn-by-tenure analysis that contextualizes health score patterns against the timing of when members in your community are most likely to cancel — including the month-3 evaluation cliff, the 6-month value-stagnation pattern, and the 12-month “what have I actually done here” review — the member churn by tenure reference card covers the tenure-specific cancellation patterns and the per-tenure intervention logic that pairs with the health score tiers. The health score tells you where a member stands. The churn-by-tenure framework tells you how much time you have before the cancellation decision is likely to be made, given where they stand.
Frequently asked questions
What is a paid community member health score?
A paid community member health score is a composite behavioral metric that predicts an individual member’s probability of renewing their subscription at the next billing interval. It is built from three behavioral signals: intro post completion (did the member overcome the first social barrier in week one?), peer interaction count (how many distinct peers did the member exchange at least one message with in their first 30 days?), and active channel engagement depth (in how many distinct channels did the member post or reply in the past 30 days, weighted for goal-specific channels?). The three signals combined score 0–6 raw and assign the member to one of four tiers: Activated (90-day renewal 72–88%), Healthy (58–72%), At-Risk (32–48%), Churned (18–32%). For the full scoring decision tables, calculation data sources, and renewal benchmarks cross-tabulated by tier and community price tier, see the paid community member health score reference card.
Why does login frequency fail to predict paid community member churn?
Login frequency measures whether a member opened the Slack workspace. It does not measure whether they exchanged value with it. A member who logs in every weekday and reads threads without ever replying appears identically healthy to a member who logs in every weekday and runs five active peer conversations. These two members have renewal probabilities 35–40 percentage points apart at 90 days. Login frequency cannot distinguish them because it is a presence metric, not a value-exchange metric. The specific sub-population where login frequency generates the most false signals is the informed lurker: high login frequency, zero peer interactions, zero active channel engagement. Login frequency predicts renewal for this member at the high-WAU baseline rate. The health score classifies them as Churned at 18–32% renewal. The health score is correct; the lurker typically cancels at month 3 or month 4.
What are the four paid community member health score tiers?
Activated (score 4.5–6): all three signals strong; 90-day renewal 72–88%. No rescue intervention needed; operator maintains through contribution recognition and cross-channel routing. Healthy (score 3–4): two signals strong or all three at moderate levels; 90-day renewal 58–72%. Proactive deepening within a 30–45 day window through goal-specific content curation and peer-routing to add the third peer connection. At-Risk (score 1–2.5): one signal strong or two at low levels; 90-day renewal 32–48%. Personal and specific operator DM within 10 days, referencing a named thread and a specific peer, paired with a peer-routing action; combined intervention produces 28–38% recovery-to-Healthy in 30 days. Churned (score 0–0.5): no meaningful signals in 45 days; 90-day renewal 18–32%. Win-back DM acknowledging the gap and offering a direct reconnection path; generic check-ins convert at 4–8%, structured win-back at 15–22%.
How do you calculate a paid community member health score manually without any tools?
The manual process requires three data points per member, extracted from Slack workspace admin exports: Signal 1 — scroll #intros for the past 90 days and record a 1 for each member who posted within 7 days of join, 0 for those who did not. Signal 2 — use the Slack member activity CSV export to count distinct peer interactions per member in the 30-day post-join window; score 0 for zero peers, 1 for one to two, 2 for three or more. Signal 3 — use the same CSV export to count distinct active channels (posts or replies, not reads) per member in the past 30 days; score 0 for zero channels, 1 for one to two, 2 for three-plus, plus +0.5 if any active channel is goal-specific. Sum the three signal scores; assign tier based on total: 4.5+ is Activated, 3–4 is Healthy, 1–2.5 is At-Risk, 0–0.5 is Churned. First run takes 60–90 minutes for a community under 300 members; monthly recalculations take 30–45 minutes once the spreadsheet structure is established.
What should a paid community operator do when a member’s health score drops to At-Risk?
The At-Risk intervention has two components and must happen within 10 days of the score drop. Component one: a personal operator DM that references a specific thread, resource, or peer by name and connects it to the member’s stated goal. Generic check-ins produce 8–14% conversion; specific, named-reference DMs produce 22–31%. The framing that converts is not “I noticed you haven’t been active” but “I noticed the discussion in [specific channel] about [specific topic] this week and thought it matched what you said in your intro about [their goal] — worth dropping a reply to [specific person]’s thread.” Component two: peer-routing. Identify one established Activated or Healthy member with context adjacent to the At-Risk member’s stated goal and route them to the At-Risk member’s intro post or last active thread. Personal DM alone produces 18–24% recovery-to-Healthy in 30 days. Personal DM combined with peer-routing produces 28–38%.