Foothold

Metrics & retention

My retention rate still said 63% six months after I fixed onboarding. The number wasn’t wrong. It just hadn’t caught up yet.

Tomás Reyes improved week-one activation from 29% to 68% in May 2025. Six months later, his trailing 12-month rolling retention rate still read 63% — squarely in the “marginal compounding” tier. He thought the sequence had been less impactful than he expected. A friend walked him through a cohort breakdown and showed him why the rolling number was the wrong instrument for measuring a recent improvement — and when the rolling rate would finally catch up.

Published 22 October 2026 · 12 min read · Reference: Slack community retention rate — how to calculate it and what a good number looks like →

The community: The E-Commerce Operators Guild

Tomás Reyes launched The E-Commerce Operators Guild in November 2024. The positioning: a paid Slack community for DTC founders, heads of growth, and brand operators at consumer companies doing $1 million to $30 million in annual revenue, $119 per month. Not e-commerce in general — specifically the operational layer: customer acquisition cost as unit economics, email and SMS retention strategy, contribution margin by channel, and the operational decisions that scale brands between their first $1M and $30M.

He had spent six years at DTC apparel companies, most recently as head of growth at a brand that scaled from $4M to $22M in three years before a strategic acquisition. The peer conversations he had valued most in that period — about Meta attribution windows, COGS structure under inflationary pressure, how to evaluate a 3PL change — had happened in Slack DMs with two or three people he had met at a trade show. He built the community to institutionalise those conversations.

One year in, in November 2025: 310 members, $36,890 MRR. He had grown at a steady pace since launch, adding 20 to 30 new members per month from a combination of his personal following on LinkedIn, word of mouth from founding members, and occasional mentions in e-commerce operator newsletters.

The annual retention calculation he had been avoiding

Tomás had been tracking month-over-month membership numbers and MRR since launch, but he had never calculated a formal annual retention rate. By November 2025 — the community’s one-year anniversary — he had enough data to run it.

The formula he used, from the Slack community retention rate guide: trailing 12-month rolling retention equals one minus the quotient of members-at-period-start plus new-members-joined minus members-at-period-end, divided by members-at-period-start plus new-members-joined.

His numbers:

He worked it out step by step from his Stripe export. Over the full 12 months, 329 unique subscribers had been active at some point (47 founding + 282 who joined later). At the end of the period, 310 were still active. That meant 19 had churned — but this ignored the members who joined and churned within the same 12-month window. From his Stripe data: 207 members of the 329 were still active 12 months after their respective join dates (or, for those who joined mid-period, were still active at November 2025). The rest had cancelled.

The trailing rolling retention came out to approximately 63%.

He looked up the benchmark table. Sixty-three percent: marginal compounding. Growing slowly, but churn still the dominant story. One acquisition slowdown and net growth reverses.

He had expected to be in the strong-compounding tier — above 66%, ideally approaching 70%. Six months earlier, he had made what he thought was a significant operational change to his onboarding. The retention number did not reflect it.

The May 2025 onboarding overhaul

Tomás had run the activation audit in April 2025 after reading about the concept in a community operations newsletter. The audit procedure: export the full Slack member list, sort by messages-sent ascending, count zero-message members.

His result: of 185 paying members at that point, 112 had never sent a single message in the workspace. That was 60.5% of total membership.

He segmented the 112 by tenure:

He ran a 30-DM manual test on members from the 0–45-day group. He had been collecting checkout data — company revenue range, primary growth challenge (two to three sentences required), and biggest operational decision on their plate right now — since launch. He wrote personalised DMs using that data: opened by naming the specific challenge they had described at checkout, routed each member to the current week’s most active thread in the closest channel match, and asked one question they could answer in a single sentence.

Results: 21 of 30 posted within 72 hours. That was a 70% activation rate, against a 29% baseline on his existing welcome-channel post approach.

He built the full sequence in May 2025: a four-question checkout intake form (company revenue range, primary growth challenge, current biggest operational decision, and what they hoped to get from the community specifically), five channel routes (#acquisition-and-paid (#ad-buying, attribution, CAC), #email-and-sms (retention, flows, list building), #operations-and-3pl (fulfillment, COGS, supplier), #brand-and-product (positioning, expansion, SKU decisions), #unit-economics (contribution margin, LTV, cohort modeling)), and a Day-0/3/7 sequence.

The sequence:

Week-one activation from May 2025 onward: consistently 65 to 70%. Month-three renewal for cohorts who went through the sequence: 73 to 79%, against 55 to 58% for the pre-sequence cohorts he could compare. The change had worked by every measure he could track within 90 days of a member joining.

And then he ran the annual retention calculation in November 2025 and got 63%.

The question he brought to a peer

Tomás had a call in late November with Rebeca Suárez, a friend who ran a paid community for consumer brand founders and had been operating for three years. He described the situation: activation improved dramatically, month-three renewal improved dramatically, and yet the annual retention rate looked unchanged from what he would have expected before the sequence.

Rebeca asked him one question: “What percentage of your current 12-month trailing window is cohorts that actually went through the new sequence?”

He had not thought about it that way.

She walked him through the math.

The retention rate refresh lag

The trailing 12-month rolling retention rate as of November 2025 covered members who had joined in the community at any point between November 2024 and October 2025. That was 12 cohort-months of membership history.

Tomás had deployed the sequence in May 2025. That was cohort-month 7.

Of the 12 cohort-months in the rolling window:

The rolling rate was blending both. The November 2024 founding cohort — the only cohort with a full 12-month window completed — had retained at 49%. The January 2025 cohort had a 10-month window and was tracking toward 51 to 53%. These pre-sequence cohorts were dragging the rolling average down, correctly reflecting their actual retention performance under the old approach.

The post-sequence cohorts (May through October 2025) had 1 to 6 months of data. Their 6-month cohort retention was 79 to 83% for the older post-sequence cohorts. The newest ones had barely started their windows. They were positive contributors to the rolling calculation, but they had not yet had time to demonstrate their 12-month retention.

Rebeca made the key point: “The rolling rate is not wrong. It is correctly summarising the experience of every member who was in your community across the past 12 months. Half of that experience was under the old system. The number you’re reading is accurate. It just measures the past 12 months, not the next 12 months.”

The retention rate refresh lag works like this: when an operator deploys an onboarding improvement in month M, the trailing 12-month rolling rate will not fully reflect that improvement until month M+12 — when every cohort in the rolling window has been onboarded through the new sequence. Before M+12, the rolling rate is blending pre-improvement and post-improvement cohorts in proportions that shift by one month each month.

In Tomás’s case:

He had been measuring his improvement with the right instrument for investor reports and annual reviews. He had been using the wrong instrument for confirming whether a recent operational change had worked.

The cohort breakdown that showed the actual improvement

Rebeca helped Tomás build a per-cohort analysis from his Stripe export. The procedure: group subscribers by join-month, then for each cohort calculate what percentage of members were still active 6 months after their join date.

The 6-month cohort retention data:

Pre-sequence average 6-month cohort retention: 57%. Post-sequence average: 80%. A 23-percentage-point improvement in 6-month cohort retention, beginning with the first cohort to go through the sequence.

The improvement was not gradual or ambiguous. It was a step change that appeared in the May 2025 cohort and held through every subsequent cohort. The sequence had worked. The rolling annual rate was not seeing it yet because it was averaging across a window that was half pre-sequence history.

Rebeca ran the forward projection. The pre-sequence cohorts (November 2024 through April 2025) had 6-month cohort retention of 54 to 61%. Based on the first complete cohort (November 2024, which had a full 12-month window), 12-month cohort retention for pre-sequence cohorts was approximately 49 to 54%. The post-sequence cohorts had 6-month cohort retention of 78 to 82%. Extrapolating to 12-month — using the same month-over-month decline pattern visible in the pre-sequence cohorts — the post-sequence 12-month cohort retention would likely land at 72 to 76%.

Each month from November 2025 through May 2026, one pre-sequence cohort would roll off the trailing 12-month window and be replaced by a post-sequence cohort. By May 2026, every cohort in the rolling window would be post-sequence. The trailing rolling rate would cross 75% sometime between March and May 2026 — assuming acquisition pace held and the sequence continued performing at its current level.

“Your retention rate is going to say 75% in about six months,” Rebeca said. “The work is already done. The clock is just running.”

What the month-three renewal rate had been telling him all along

After the call, Tomás looked back at the monthly data he had been collecting. He had been tracking month-three renewal rate — the percentage of members who joined 90 days ago who were still active today — alongside MRR and total members since launch. He had not paid it much attention because he was focused on the annual number.

The month-three renewal rate had been telling the story he could not see in the annual rate:

The month-three renewal rate had jumped 20 percentage points the month the first post-sequence cohort reached 90 days, and had held at that level for every subsequent month. He had been watching this in his spreadsheet but had been dismissing it as insufficient evidence because the annual rate “wasn’t moving.”

The month-three renewal rate is a leading indicator for annual retention in exactly the way the annual rate is a lagging one. Members who renew at month three are retaining at the highest-risk churn point — the first billing-cycle renewal after the initial “this community is probably worth keeping” evaluation. A 20-point improvement in month-three renewal for every post-sequence cohort was the most direct possible confirmation that the sequence had solved the problem it was designed to solve: first-week activation, which determines whether a member reaches month three with a formed opinion about the community’s value or with a vague sense that they never quite engaged.

He had the confirmation signal the whole time. He had been looking past it at the annual rate.

The diagnostic sequence he now runs quarterly

After the conversation with Rebeca, Tomás formalised a quarterly retention review that looks at both instruments in the correct order:

Step 1: per-cohort 6-month retention by join-month. The leading indicator for whether any onboarding or programming change worked. Run this for every cohort in the last 12 months. Look for step changes — a cohort where the number jumped materially, which indicates a change that month had an impact. This is the instrument for confirming improvements in near-real-time, not waiting for the annual rate to catch up. See the cohort retention calculation guide for the exact formula and how to run it from a Stripe export.

Step 2: month-three renewal rate, trended over the past 6 months. The confirmation signal for activation-driven improvements. If week-one activation improved, month-three renewal should show it within 90 days. If the month-three rate improved but the annual rate didn’t, you are in the refresh lag window — the annual rate will catch up in 6 to 12 months. If both month-three renewal and 6-month cohort retention have improved but the annual rate hasn’t, the lag is structural, not a sign the changes are failing. For the relationship between the activation cliff and month-three churn, see the cohort-month breakdown guide.

Step 3: trailing 12-month rolling retention. The investor-report metric and long-term benchmark. Not useful for detecting recent changes; useful for annual reviews, kill-or-continue decisions, and comparing the community to external benchmarks. If this number is below 65% and per-cohort data shows recent cohorts retaining at 75%+, you are in the refresh lag. If per-cohort data also shows low retention, the problem is current, not historical. See the benchmark table for the four-tier framework (treadmill / marginal / strong / exceptional).

The key insight from the sequencing: operators who look at the trailing annual rate first and see no improvement will often conclude that a recent intervention “didn’t work” and either undo it or replace it with a new intervention. This cycles through changes without ever holding one in place long enough to appear in the annual rate. The correct order is cohort-level data first (confirms within 6 months), month-three renewal second (confirms within 3 months), annual rolling rate last (confirms over 12 months, used for reporting). Each instrument has its own time horizon. Using the annual rate as the primary feedback loop for month-level interventions is like measuring whether a medicine is working by looking for long-term outcomes when the drug’s effect should be visible in a blood test in 72 hours.

Results as of November 2025

The monthly churn rate improvement — 5.8% to 2.3% — was the most direct signal. Monthly churn rate reflects the most recent month’s experience rather than the trailing 12-month average, so it responded to the onboarding improvement within 60 to 90 days rather than 12 months. A 5.8% monthly churn rate annualised to approximately 50% annual retention (a treadmill community requiring constant acquisition to hold flat). A 2.3% monthly churn rate annualised to approximately 76% annual retention — strong compounding. The community that existed in November 2025 was a different community from the one that had existed in April 2025. The trailing annual rate was still measuring the average of both.

The refresh lag as a trap for motivated operators

Tomás’s situation — making a real improvement and then doubting it because the annual rate didn’t reflect it — is common enough to have a specific failure mode. Operators who run a sequence for three months, see 63% trailing annual retention (unchanged), and conclude the sequence “didn’t work” will often switch tactics: try a different onboarding format, add AMAs, focus on engagement programming, or simply deprioritise onboarding in favour of acquisition. Each of these pivots abandons a working improvement before it had time to appear in the measurement instrument the operator was watching.

The most common version: an operator deploys a sequence in month 6, checks annual retention in month 9 (three months after deployment), and concludes it moved the rate by only 2 points. That is arithmetically inevitable — the rolling window has replaced 3 of 12 months with post-sequence cohorts. Three months of 80% 6-month retention averaging in against nine months of 57% 6-month retention produces a blended improvement of roughly (3 × 80% + 9 × 57%) / 12 = 63% versus the prior 57% — a 6-point improvement. The underlying per-cohort performance has improved 23 points. The annual rate reports a 6-point improvement because it has only seen 3 months of the new regime.

The operator interprets “6-point improvement in annual retention” as partial success at best. The correct interpretation is “3 months of data from a 23-point improvement, viewed through a 12-month averaging window.” The annual rate is doing exactly what it is designed to do — reporting the average experience of all members over the past year. It is being asked a question it is not designed to answer: whether a change from three months ago is working.

The two-cliff model is relevant here in a specific way. The model describes two distinct reasons members churn: Cliff 1 is the week-one activation cliff (members who never post cancel in months 1–3 before ever engaging); Cliff 2 is the month-three engagement cliff (members who activated in week one but whose engagement faded as content or programming went quiet). An onboarding sequence addresses Cliff 1. Engagement programming addresses Cliff 2. A 12-month rolling rate blends both cliffs together. If you deploy a sequence that solves Cliff 1 and see the annual rate improve only partially, you may have solved Cliff 1 and still have a significant Cliff 2 problem — or you may be in the refresh lag and Cliff 2 is fine. The cohort data and month-three renewal rate differentiate between the two situations. The annual rate alone does not.

The forward projection he monitors monthly

Tomás now tracks a single projection alongside his rolling annual retention rate: the expected rolling rate for each future month, based on the per-cohort 6-month retention data he already has and the projected 12-month cohort retention rate for post-sequence cohorts.

The projection works as follows: each month, one pre-sequence cohort exits the rolling window (having completed its 12-month tenure) and one new cohort enters. The pre-sequence cohorts that are still in the window are tracking toward 49 to 54% 12-month retention. The new cohorts entering each month have 6-month retention of 78 to 82% and are projected toward 72 to 76% at 12 months. As each pre-sequence cohort exits and each post-sequence cohort reaches its 12-month mark, the blended rolling rate improves.

His projection for the rolling annual rate:

The 74 to 76% projected rate in May 2026 represents a community that has moved from the marginal-compounding tier to the strong-compounding tier — the tier where the membership base is stable, activated members refer, and acquisition cost per net-new member decreases as retention compounds. The difference between those two tiers in operational terms: at 63% annual retention and 310 members, Tomás needs approximately 115 new members per year just to hold flat. At 75% annual retention, he needs approximately 78 new members per year to hold flat — 32% less acquisition just to maintain the same size, with more MRR available to reinvest. The compounding math changes qualitatively, not just quantitatively.

He checks the projection monthly against actual cohort data as each post-sequence cohort reaches its 6-month mark, updating the 12-month projection if cohort performance is tracking above or below the baseline. The projection is not a guarantee; it is a running estimate that tells him whether the improvement is holding at the cohort level while the annual rate catches up.

Frequently asked questions

Why hasn’t my Slack community retention rate improved after fixing onboarding?

Almost certainly the retention rate refresh lag. The trailing 12-month rolling retention rate includes 12 months of cohorts. If you deployed an onboarding improvement in month M, the rolling rate as of month M+1 still includes 11 months of pre-improvement cohorts — cohorts who churned at the old rate. Each month, one pre-improvement cohort exits the rolling window and one post-improvement cohort replaces it. The rolling rate will not fully reflect your improvement until month M+12. The correct instrument for confirming that your improvement is working before the rolling rate catches up is per-cohort 6-month retention: calculate the percentage of members from each join-month who are still active 6 months later. If cohorts who joined after your sequence deployment are retaining 15 to 20 points higher than pre-deployment cohorts at the same 6-month mark, your improvement is working. Full calculation guide at Slack community retention rate — the two methods.

What is trailing 12-month rolling retention and how do I calculate it?

Trailing 12-month rolling retention measures what percentage of all members who were active in your community in the past 12 months are still active today. Formula: 1 − ((members at period start + new members who joined in the 12-month window − members still active at period end) ÷ (members at period start + new members who joined)). Pull the inputs from your billing system (Stripe, Memberstack, or equivalent). This is the right metric for investor updates, annual reviews, and benchmarking against external data. It is not the right metric for detecting whether a recent onboarding change worked — it averages that change’s impact across the 12-month window it occupies, which dilutes the signal until the window fully refreshes. See the full two-method guide for when to use rolling retention versus per-cohort retention.

What are the retention rate benchmarks for a paid Slack community?

For paid communities in the 200–2,000-member range: 50% or below is a treadmill (losing half the community per year, acquisition required just to hold flat); 51–65% is marginal compounding (positive net growth, but an acquisition slowdown reverses it, and word-of-mouth is limited by high early churn); 66–80% is strong compounding (stable base, activated members refer, acquisition cost per net-new member decreases over time); 81%+ is exceptional (typically a very tight ICP or near-irreplaceable value for a specific professional community). Most communities in the 200–500-member range land between 55% and 70% by year two. Moving from 65% to 80%+ is where the compounding math changes qualitatively — it is the difference between a community that needs constant acquisition and one that grows with reduced acquisition effort. The single lever with the highest impact on that move is week-one activation rate.

How do I calculate per-cohort retention for my Slack community?

From your billing export, group subscribers by join-month. For each cohort, count how many members are still active 6 months after their join date. Divide by the total who joined that month. This gives you 6-month cohort retention per join-month — the instrument for detecting onboarding improvements within 6 months rather than 12. If you have a full year of data, you can also calculate 12-month cohort retention: of members who joined in month N, how many are still active in month N+12. Compare across cohort-months to identify step changes that correspond to onboarding or pricing changes you made. The pre-post comparison should show a clear jump at the month of the change if it had an impact. If the jump is present in 6-month cohort retention but the rolling annual rate is not yet reflecting it, you are in the refresh lag — the annual rate will catch up over the next 6 to 12 months as the pre-change cohorts exit the window.

What is a good month-three renewal rate for a paid Slack community?

Month-three renewal rate — the percentage of members who joined 90 days ago who are still active paying members — is a fast-moving leading indicator for annual retention. For communities with structured onboarding (a Day-0/3/7 sequence), month-three renewal of 73–80% is achievable. Without structured onboarding, 55–65% is typical. The month-three rate matters specifically because it captures the first major churn cliff: members who have not activated in their first 90 days are confronted with their first or second billing renewal without having formed a clear opinion about the community’s value. A 15 to 20-point improvement in month-three renewal is the fastest confirmation signal that a new onboarding sequence is working — it shows up within 90 days of the sequence deployment, while the annual rate takes 12 months to fully reflect the same improvement.

Related: Slack community retention rate — two calculation methods, benchmarks, and the four levers · The cohort-month breakdown: finding whether your churn is Cliff 1 or Cliff 2 · The two-cliff model: why six months of engagement programming didn’t move the retention rate · The denominator trap: why your poster count is rising and your engagement rate is falling