Paid community retention

She had 68% bilateral activation and 3.7% monthly churn. Members in months six through twelve were still leaving at 4.9% per month. A peer asked what her longest-tenured members were using the community for between sessions. The answer was not the monthly forum.

Adaeze Nwosu had built the infrastructure. Bilateral activation at 68%, month-three renewal at 78%, monthly churn at 3.7%. Everything from week one through month five was working. The members who had been paying the longest were still leaving at 4.9% per month, and annual billing conversion was 18%. Kweku Mensah ran a comparable community at 2.9% monthly churn and 54% annual billing conversion. He asked her one question at an Edinburgh dinner that she could not answer.

The community and the metrics that did not fit together

Adaeze launched The B2B SaaS Design Leaders Network in October 2024. The community was for senior product designers, design leads, and design directors at B2B SaaS companies — people who had moved beyond execution into the specific set of challenges that leadership in a design function produces: headcount decisions, systems strategy, stakeholder management with engineering and product, the question of what design quality means at scale in a company that is moving too fast to answer it carefully. The membership fee was $89 per month.

By August 2025, month ten: 198 members, $17,622 MRR. She had done the full infrastructure build in the prior six months. Foothold Day-0 DM with a goal-track question (three tracks: research and discovery; systems and components; design leadership and hiring), weekly reply-gap audit, conditional Day-3 routing, Day-10 conditional follow-up, Day-28 value-crystallization check-in, Day-75 renewal conversation. She had followed the programming investment ladder in sequence, fixing bilateral activation before adding programming, adding the renewal conversation before considering a price increase.

The numbers in months one through five confirmed the infrastructure was working. Bilateral activation: 68%. Month-three renewal: 78%, up from 54% before the Day-75 conversation was in place. Monthly churn in months one through five: 3.2% average. She had built three programming elements on top of that foundation: a monthly Design Leadership Forum (a 90-minute live session with a guest VP of Design or Chief Design Officer from a B2B SaaS company in the Series B to Series D range, walking through a real leadership decision — a design system consolidation, a headcount structure redesign after a layoff, a product quality negotiation with a new CPO), a weekly #critique-requests channel (members could post design work, a product decision artifact, or a communication piece for peer feedback), and a quarterly product design audit event (a structured peer review of one member’s product area, facilitated by the operator, with three to four other members as reviewers).

Attendance at the monthly forum by members in their first five months: 61%. The #critique-requests channel had 12 to 18 posts per week. The quarterly audit had a waiting list.

The gap she had not solved: months six through twelve. Monthly churn in that tenure window averaged 4.9%. Annual billing conversion — the percentage of members who, when offered a discounted annual billing option at the end of month six, chose to pay upfront for a year — was 18%. She was losing nearly one in twenty members per month in exactly the cohort that should have had the strongest retention signal: long-tenured, past the activation phase, past the second-month cliff, past the first conscious renewal decision at month three.

The programming fatigue hypothesis and why it did not hold

Her first interpretation was that long-tenured members were experiencing what she called programming fatigue: they had attended the monthly forum several times, had a clear sense of what it offered, and were attending less frequently because the format was no longer novel. The month-six-to-twelve forum attendance data supported this reading. Members in their first five months attended the monthly forum at 61%; members in months six through twelve attended at 29%.

She tested the hypothesis by varying the format. She added shorter 30-minute deep-dives between the full 90-minute forums — a mid-month async thread seeded by a practitioner question rather than a live session. She rotated the guest profile more deliberately, moving from VP-of-Design-at-known-company to CDO-at-less-obvious-company to try to reach members who found the early-stage guest format less relevant to their larger-company context. She experimented with an async version of the quarterly audit event, where the peer review happened in a dedicated channel over two weeks rather than in a single Zoom session.

The months-six-to-twelve churn continued at 4.9%. The format changes had not moved it.

In September 2025, she flew to Edinburgh for a paid-community operators gathering — fourteen operators, two days, no formal agenda. She had come specifically because she had run out of interpretations that fit the data and wanted to describe the problem to people who had seen more versions of it than she had.

What Kweku asked at the Edinburgh dinner

Kweku Mensah ran The B2B Product Operations Network — a paid Slack community for product operations managers, chiefs of staff in product organizations, and senior product analysts at B2B technology companies. The fee was $99 per month. By September 2025: 235 members, $23,265 MRR, 2.9% monthly churn, 54% annual billing conversion. He had launched in March 2024 — seven months before Adaeze.

She described the gap at dinner: the activation infrastructure was working, the month-three retention was working, the programming was good, but months six through twelve were still running at 4.9% and she had no remaining hypothesis that fit the shape of the data.

He listened without interrupting. When she finished, he asked: “Of the members who have been with you for more than six months and are still paying today — the ones with the lowest churn probability in your current cohort — what do they use the community for between sessions?”

She said: the monthly forum, the #critique-requests channel, reading threads when something relevant came up.

He said: “That’s what they do in the community. I asked what they use it for. Is there anything they couldn’t do without the community? Something they return to that isn’t a session or a channel?”

She thought about it. She described two or three members who she knew had formed genuine peer relationships inside the community — who had told her explicitly that a connection they had made through the community had been professionally significant. She named one: a design director who had connected with another member while discussing a reorg in the #critique-requests channel, and who had subsequently hired that member six months later. Another: a member who had been navigating a difficult design system consolidation and had found a peer who had done the same thing at a comparable company, had a series of private conversations with that person, and had later cited the community as the reason she had renewed.

Kweku said: “Those are the members who will renew at month twelve. The common thread is not which sessions they attended. It’s that they found a specific person and had a one-to-one conversation that solved a specific problem. The community was how they found that person. The question is whether every member in months six through twelve has a way to do that, or only the ones who happened to be in the same channel thread at the right moment.”

The member directory effect

He had launched a searchable peer directory in his community seven months earlier. He described what it looked like, what had changed, and what he now called the member directory effect.

The mechanism, as he described it, was a function of the member lifecycle that he had not understood until he had data from two full cohort cycles.

In months one through five, a new member’s primary value experience is discovery: finding out what the community contains. Every session is potentially the first of its kind; every channel is unfamiliar; every member is someone who might be relevant. This is an active orientation that produces reliable engagement. Members open Slack frequently, attend sessions, post questions and work. The discovery phase is the most high-engagement period of the member lifecycle, and it is not driven by the quality of the programming — it is driven by the novelty of an unexplored environment that promises relevant things.

The discovery phase has an endpoint. By month six, most members have a complete mental model of the community: they know which channels are worth following, which sessions they find valuable, which members they are most likely to learn from. The unexplored environment has been explored. What remains as the ongoing value proposition is: the programming they attend when it is scheduled, and the chance that a useful thread appears in a channel they already know at the moment they happen to be looking.

This is a fundamentally weaker value proposition than it appeared at month one. Not because the programming has gotten worse or the community has gotten quieter. Because the member’s relationship to the community has changed: they are no longer discovering what is here. They are waiting for something relevant to appear. The passive-wait experience of a six-month member is structurally different from the active-discovery experience of a two-month member, even if they are doing approximately the same things in the community.

“The question I could not answer,” Kweku said, “was what I was actually selling to a member in month eight. Not what I was offering. What was I selling to someone who already knew every channel, had attended seven monthly sessions, and was now deciding whether $99 a month was worth paying for the next twelve months. The programming was the same programming. They already knew what it was. What else was there?”

A peer directory changes the answer. It converts the community’s value proposition from an event subscription — programming that members attend when it is scheduled — to a professional network infrastructure: a searchable, always-available resource for finding a specific person with specific experience who is open to a direct conversation. The value of that infrastructure does not diminish as the discovery phase ends. It increases, because as the member accumulates context about their specific challenges and the specific kinds of experience that would help them solve those challenges, they become better at using a directory to find exactly the peer they need.

A member who is six months into a senior design leadership role and is navigating a specific problem — how to build a design system that survives a major product acquisition, how to manage a peer relationship with a new VP of Engineering who has different assumptions about where design quality lives, how to structure a first-time design hire when the role requirements are genuinely ambiguous — can search a directory for a peer who has solved that specific problem at a comparable company stage and reach out directly. That is not a session they attend when scheduled. It is infrastructure they use when they need it. The community is the directory. The directory makes the community’s value permanent.

What Kweku built and what he observed

His directory was a Notion table. He built it in an afternoon. The fields: member name (linked to their Slack profile), role title, company type (B2B SaaS, B2B fintech, B2B infrastructure), company stage (seed, Series A–B, Series C+, post-IPO), current focus (three choices from a dropdown: systems and tooling; stakeholder management and headcount; pricing and packaging ops), open to intro (yes / on request / not currently), timezone.

He sent an invitation to the full member base: a brief DM from the operator noting that he was launching a member directory so that members could find peers with specific backgrounds relevant to their challenges. Filling out the profile would take about three minutes. Members who wanted to review introduction requests before he connected them could set their intro preference to “on request.”

Directory completion rate after two weeks: 71% (167 of 235 members). He sent one follow-up to non-completing members seven days after the original invitation; 14 additional members completed after the follow-up. The remaining 29% did not complete — either inactive members in a low-engagement period or members who preferred not to be searchable.

He did not structure any programming around the directory. There was no monthly “find a peer” prompt, no formal introduction event. The directory existed as a passive infrastructure element — accessible from the community’s Slack sidebar as a pinned link. Members who wanted to find a specific peer could search it. Members who wanted to be introduced to a peer they found could DM the operator.

Eight weeks after launch: 23 member-initiated peer introductions had been facilitated through the directory. Kweku reviewed the tenure data for the members who had initiated introductions. Of the 23: 19 were members in months six through twelve of their tenure. Four were members in months two through five.

The pattern was not a surprise to him in retrospect, but it had been a surprise when he first saw it. He had expected newer members to be the primary users — people who were still in the discovery phase and were using the directory to accelerate their exploration of who was in the community. Instead, the directory was being used almost entirely by members who had passed the discovery phase. They were not using it to find out who was in the community. They knew. They were using it to find a specific person for a specific purpose at a specific moment, in a way that the general community activity was not reliably producing for them.

What Adaeze built and how she launched it

After the Edinburgh dinner, she built a Notion database for The B2B SaaS Design Leaders Network. The fields she chose: name (linked to Slack profile), role title, company type, company stage (seed / Series A–B / Series C+ / post-IPO), design challenge focus (three options: research and discovery; systems and components; design leadership and hiring), open to intro (yes / on request / not currently), timezone.

She chose a different launch approach than Kweku had used: a concierge launch. Rather than sending a general invitation and asking members to fill out their profiles from scratch, she reviewed the full member list and pre-populated directory entries for the 47 members she knew well enough to fill in the basic information — role, company type, company stage, approximate challenge focus based on what they had posted in channels. She sent those 47 members a personal DM: “I’ve pre-populated your directory profile based on what you’ve shared in the community. Here’s the entry. Can you check it, add or correct anything, and set your intro preference? Takes about a minute.”

Forty-three of the 47 responded within five days. All 43 confirmed or corrected their entry and set their intro preference. Four did not respond; she left their entries in the directory with a note that the entry was operator-generated and not yet self-confirmed.

For the remaining 151 members, she sent the standard invitation: “I’m launching a member directory so you can find peers with specific backgrounds relevant to your current challenges. The directory already has 43 entries from long-tenured members. Three minutes to fill out your profile — or you can set your intro preference to ‘on request’ if you want to review connections before I make them.”

The framing mattered. Showing that 43 entries already existed — that the directory was not empty, that the concierge cohort had already populated it with real data — produced a higher completion rate than the cold invitation would have. Directory completion at end of week two: 78% (154 of 198 members). The remaining 22% either did not respond or set their preference to “not currently.”

The directory in use

She facilitated all introductions personally. Her process: when a member DMed her asking to be connected to a specific directory entry, she reviewed both members’ profiles and sent a brief message to the target member before making the connection: “I have a member who would like to connect with you — they’re working on [specific challenge from their profile], and I thought your experience with [specific background] would be directly relevant. Would you be open to a brief conversation?”

She processed 31 introduction requests in the first ten weeks. Thirty-one introductions meant reviewing 62 profiles, drafting 31 pre-introduction messages to the target members, and facilitating the actual connections. It took about twenty minutes per introduction request, managed in batches of four or five on Thursday mornings. Around week eight, she noticed that the introduction requests were running faster than she could process them in a single weekly batch and started a brief waitlist for lower-urgency requests.

The tenure distribution of the requesting members confirmed Kweku’s pattern: of the 31 introduction requests in the first ten weeks, 26 came from members in months six through twelve of their tenure. The remaining five came from members in months three through five.

She tracked the outcomes of the introductions at 30 days and 90 days: did the members actually have a conversation, and did either member report that it had been useful? At 30 days, 27 of the 31 introductions had resulted in at least one direct message exchange. At 90 days, 19 of the 31 reported — in a brief follow-up DM from Adaeze — that the conversation had been directly useful for something they were working on. Six said the conversation had been useful but not immediately relevant to their current challenges. Six did not respond.

The results: months-six-to-twelve retention and annual billing conversion

The data tracked over six months from September 2025 through February 2026:

She also tracked the behavioral correlation between directory usage and retention. Of the members who converted to annual billing between September 2025 and February 2026: 81% had used the directory at least once (either initiated an introduction request or been the target of one). Of the members who churned between months six and twelve in the same period: 89% had not used the directory at all.

She was careful about interpreting the correlation. Members who used the directory were probably already more engaged with the community than members who did not — the direction of causation was not clear. A more engaged member is both more likely to use a directory and more likely to stay past month twelve. The directory may have been a symptom of high engagement rather than a cause of retention. She could not rule this out.

But the alternative interpretation — that the directory had no causal effect and was only measuring existing engagement — did not fit the distribution of introduction requests by tenure. If directory usage merely tracked engagement, she would expect newer members (more actively engaged in the discovery phase) to use it at least as much as tenured members. The opposite was true: 84% of introduction requests came from members in months six through twelve. That was the tenure cohort where engagement was declining, not increasing. They were not using the directory because they were the most engaged members of the community. They were using it because it was the only available mechanism for the specific form of value — a targeted peer conversation about a specific current challenge — that asynchronous channel activity and monthly sessions could not reliably produce for them.

The two phases of paid community value

Kweku’s formulation, from the Edinburgh dinner, was the one Adaeze used when she described the member directory effect to other operators in the months that followed: “There are two different value propositions in a paid community, and they serve two different member states. The first is the discovery proposition: here is an environment full of people and activity that is relevant to you. Explore it. The second is the infrastructure proposition: here is a resource that lets you find a specific peer for a specific problem at any time. Use it. Most communities only ever build the first. They optimize programming, improve sessions, refine channels. And then they wonder why engagement drops at month six — because they have exhausted the first proposition without ever building the second.”

The programming investment ladder describes the infrastructure that makes the discovery proposition work: the bilateral activation in week one that creates the felt sense of belonging, the reply-gap audit that ensures no member is left without a response to their first post, the conditional Day-3 routing that keeps newly activated members engaged through the second-month cliff. All of that infrastructure is designed to convert a new member into an engaged participant in the discovery phase. It is the foundation of every other retention investment the community can make.

The member directory effect describes what comes after the discovery phase is over. The discovery infrastructure ensures that members arrive in month six as genuinely engaged participants who have formed peer relationships and know the community well. The directory ensures that those members, once they have arrived, have a mechanism that remains valuable regardless of how well they already know the community — one that gets more valuable, not less, as their professional challenges become more specific and their ability to articulate what they need from a peer conversation becomes more precise.

Together, the two infrastructures address the full member lifecycle from week one through year two: bilateral activation creates the felt community member in week one; live programming creates the engaged community participant in months one through five; the peer directory creates the invested community infrastructure user in months six through twelve and beyond.

The practical implication: annual billing conversion is the metric most directly affected by the directory because it requires the member to make a deliberate forward commitment to the community. A member who has used the directory to find a specific peer and had a conversation that directly helped them with a specific challenge has done something that makes the forward commitment feel concrete rather than speculative: they know what they would be paying for. “This community is where I would find the next person like that if I need one” is a fundamentally different value proposition than “this community has monthly sessions that are sometimes useful.” Annual billing is the member’s answer to the question of which value proposition they have experienced.

What the directory needs to work and what it does not need

The directory does not require a sophisticated technical implementation. Kweku’s first directory was a Notion table. Adaeze’s started as a Google Sheet for the first three weeks before she moved it to Notion for better filtering. The format matters less than three things: searchability (can a member find a peer by role and current challenge?), opt-in quality (does the member feel comfortable with how they are represented and who can reach them?), and density (is there enough of the member base represented that a search for a specific profile type will return at least two or three results?).

The two directory fields that most predict successful peer introductions, in Kweku’s data: current challenge focus and company stage. Role title — the field operators most often include and members most often fill out first — is the least predictive. A member searching for a peer to help with a stakeholder management problem at a Series B company does not primarily need someone with the same role title. They need someone who has navigated a similar stakeholder situation at a comparable company stage. Challenge focus and stage are the fields that make the match; role title is the field that provides context once the match is made.

The concierge launch approach — pre-populating profiles for known members before sending the general invitation — produces meaningfully higher completion rates than a cold general invitation, because it reduces the blank-page friction for members who would fill out a profile if they had a starting point but will not start from scratch when the invitation is one of many things competing for their attention. The ten to fifteen percentage points of additional completion that the concierge approach produces represent a meaningful difference in directory density: a 78% completion rate on a 198-member community gives a directory that reflects the full range of roles and company stages; a 63% completion rate gives a directory that may have meaningful gaps.

The directory also does not require a formal usage structure. No monthly “find a peer” prompt is necessary, and no mandatory introduction events are needed. The directory works as passive infrastructure: members who need it will find it; members who do not currently need it will know it is there. The operator’s job is to make it findable (a pinned link in the community sidebar or channel description), to keep it updated (a quarterly review of profiles to update company stage and challenge focus for members whose situations have changed), and to facilitate introductions when members request them. Those three tasks represent about 90 minutes per month for a community of 200 members with an active introduction volume.

If you want the bilateral activation tracking and conditional routing infrastructure that makes the discovery phase work — the foundation that the directory builds on — Foothold handles layers one through four of the programming investment ladder in a single Slack app install. The peer directory, the reply multiplier, and the programming decisions at layers five through seven remain the operator’s work. But the infrastructure question for months six through twelve is not about whether the monthly sessions are good enough. It is about whether every member, in the specific month when they have a specific professional challenge and a specific question they cannot answer from what they already know, has a mechanism to find the peer who can answer it. The directory is that mechanism.