Mastermind Operating Manual

Six to eight owners meet on a fixed afternoon each month, in one member's actual place of business, for three and a half hours. Everyone reports against last month's commitment. One person gets an hour on a real problem. Two or three people show working software. Everyone leaves with a written commitment, and the group has dinner.

That is the whole thing. The rest of this manual is the detail that makes it survive past month six — which is where most groups quietly die.

Three design decisions carry most of the weight, and each is doing a specific job.

Most mastermind advice is testimonial. There is, however, one large randomized trial that tested almost exactly this format on almost exactly this population, plus a handful of findings that should change how you run the room.

The findings that shaped specific rules here

Size: six is the floor, eight is the ceiling

Every peer group that runs at twelve to eighteen members pays a professional facilitator whose entire job is rationing airtime. Every peer-facilitated group in the world sits at six to ten. You are not paying a facilitator, so eight is a hard ceiling.

Six rather than five is the less obvious call. At a 60% quorum, five members means quorum is three — two absences puts you at the edge and a third cancels the meeting. At six you can absorb two absences and still have a real room.

The physical test is the honest one: the group should fit around one table and hold one conversation. When it splits into two conversations, it has become two groups sharing a room.

Hard exclusions

Structural, decided before anyone is considered, not negotiable case by case:

• No direct competitors — and more precisely, no two members who could plausibly end up in the same pitch. Industry labels are too coarse. Ask the question directly.
• No economic authority between members — no employer and employee, no investor and founder, nobody who is a dominant client of another.
• No romantic relationships between members.

Stage and domain

The rule the evidence supports is homogeneous on stage and stakes, heterogeneous on domain — same altitude, different mountain. Grouping by similar industry has been tried at scale and underperformed; cross-industry diversity worked markedly better.

Three mitigations, strongest first:

1. Narrow the band. Not identical, but within roughly one order of magnitude on revenue or headcount. A $2M agency and a pre-revenue builder are not peers — they are mentor and mentee, which is a different relationship with different economics.
2. Make airtime structural. Fixed-length slots, a published rotation nobody can skip, and a timekeeper who is not the moderator.
3. Band on stakes rather than revenue. "This is my only income." "I make payroll for other people." "One client is 40% of my book." Shared stakes generate shared problems even at different scales.
4. The most senior person in the room speaks last. Every round, every phase. This is the cheapest intervention available and the evidence behind it is unusually direct.

Twelve dates and twelve venues, in everyone's calendar before the first meeting happens. This is the highest-return single action available to you and it costs one afternoon.

Eleven meetings and a retreat

Eleven monthly sessions plus one overnight retreat a year. Skip a month if you must, but skip it on the published calendar, in advance.

Time of day matters more than day of week

3:00–6:30 PM, rolling into dinner. This takes the back half of a working day rather than the front, so nobody loses a billable morning, and it flows naturally into the meal where most of the actual relationship forms.

Length is not arbitrary either. Below three hours you cannot fit a per-member round and a real deep dive. Beyond about four, attention fails without a meal break.

Travel sets your cadence

Total door-to-door commitment should stay under one working day. At 3.5 hours of meeting plus travel plus dinner you are already at six or seven hours. If any member's round trip exceeds about 90 minutes, monthly is your only realistic in-person cadence — compensate with the written loop in section 08, not by meeting less often.

Two hundred and ten minutes. The clock below assumes six members; at seven or eight, take the extra time out of the topic block, never out of the hot seat.

The round is for the extremes, not the average

The most common way a group of business owners fails is that it becomes an expensive status-update meeting. The instrument that prevents it is a rule about what the round is for.

When the round goes hollow, the repair is specific: have each member recount their last extreme, then separate which parts were facts and which were feelings, and deliberately push toward the feelings. Facts are where owners hide.

The topic block starts in silence

Open discussion is a bad way to generate options, and this is one of the better-tested findings in group research. The fix is a ratio, not a vibe.

Food, phones and drink

• Phones off or silenced for the whole session. Email and voicemail at the scheduled breaks only.
• No eating during the hot seat. Food before or after, never through the serious part.
• Dinner is part of the ritual, not an optional extra — but cap it, and allow two skips a year, so members with childcare or client constraints are not structurally second-class.
• Decide the alcohol question in advance. Half the room disclosing at the 5% level while the other half is three drinks in produces a bad room. The workable default: dry during the session, open at dinner.

Sixty minutes, five phases, and one rule that matters more than the rest: the group shares experience, it does not give advice.

Why the no-advice rule is not politeness

Advice invites the presenter to evaluate the advisor rather than the problem, and it lets the room compete on cleverness. Experience-sharing is also what the evidence supports: mentorship by experienced operators moved profits 20% while abstract business training moved them zero. Specific and lived beats general and correct.

The two failure modes

• The dominant talker. Handle it structurally, not interpersonally. The timekeeper is not the moderator, phases are hard-stopped, and experience-sharing goes round-robin rather than free-for-all. If someone still absorbs the room, that is a private conversation the moderator has once, before it becomes the group's problem.
• The silent member. Round-robin means nobody can pass invisibly — passing is done out loud, with a reason. Repeated silence is a drift signal, and section 11 assigns someone to notice it.

Thirty minutes. Two to five demos. Five minutes each, enforced. This is what makes it a tech mastermind rather than a business forum, and it dies instantly without hard rules.

Running it

• Five minutes, timed. One-minute warning at four. At five the timekeeper stands up. Announce this before the first demo so nobody is surprised.
• Order is set in advance, not volunteered on the day. Strongest demo first and last.
• Two questions maximum per demo if the clock allows. Deeper conversation happens at dinner.
• The presenter states what feedback they want in one sentence before starting. "Tell me whether the onboarding makes sense" gets useful answers; "what do you think?" gets noise.

Craft rules that make five minutes work

• Don't talk and drive. One person narrates, another works the keyboard. Nobody does both well under a clock.
• Show the problem, then the solution. Not a tour of features — that is an inventory, not a demo. Order what you show by purpose, not by where it happens to sit on screen.
• Skip the steps, show the payoff. If adding the thing takes ten clicks, show two and the result.
• Forty-five seconds of live troubleshooting, maximum. Then drop it and move on. With everyone watching, your debugging is worse than usual, and the room is watching you fail at something that isn't the point.
• Have a screencast fallback on the laptop, and a one-click reset if you'll ever show it twice. Assume the network dies thirty seconds in.
• Set your font size before the room, not in it. 14–18pt bold in the terminal.
• A visible mistake, explained well, builds credibility rather than costing it — "this is a common trap, here's what to watch for" is worth more than a clean run.

Four social rules for the demo block

These are borrowed, near-verbatim, from a programming community that has run on them for over a decade. They protect exactly one thing: people's willingness to ask questions and show unfinished work.

• No well-actually's. Correcting something tangential to what the speaker is actually saying. It breaks the flow and shifts attention to the corrector.
• No feigned surprise. Acting surprised that someone doesn't know a thing. It makes people less likely to admit gaps, which is how they stop learning.
• No backseat driving. Lobbing advice from across the room without actually joining the conversation.
• No subtle -isms. The small, probably unintended expressions of bias.

And the part that makes them usable rather than oppressive: you are going to break these, and that is fine. Everyone does. Someone says "that was a well-actually," you say "you're right, sorry," and the conversation continues. Naming a violation has to cost almost nothing socially, or nobody will ever do it.

The meeting is monthly. The accountability is not. This is where follow-through actually lives, and it is four lines of writing every two weeks.

The biweekly written check-in

Posted by every member, same day, same place, four fields:

Rules that keep it alive

• Commit to actions, never to identities or outcomes. "Ship the pricing page by the 14th," not "become the kind of business that charges properly." Publicly announcing an identity goal has been shown to reduce follow-through — and the effect is strongest in the people who care most.
• The commitments list is read aloud at the start of every session. An artifact survives only if a ritual consumes it. Without this the list rots within three cycles, regardless of what tool it lives in.
• Nobody chases anybody. A missing check-in is visible by itself. The drift-check role in section 11 handles it, and it is a peer, not the moderator.

Tooling, kept deliberately small

For six to eight people who see each other in a room, the risk is over-tooling, not under-tooling. A group chat people already have on their phone, plus one durable page per member and one per meeting, is enough. Everything else gets abandoned by cycle three.

Chat logs are effectively unsearchable at twelve months on every platform. What makes a year-old discussion retrievable is that someone wrote twenty structured lines with stable fields — not search quality.

This section exists because your group is unusual. A room of client-serving owners can send each other real work — the largest single upside available to you, and the source of nearly all the risk.

Keep referrals a gift. Pay only for work performed.

This is the most important line in the section, and the practitioner consensus and the economics agree on it.

Stopping the network curdling into obligation

• Make asking structured and universal. Everyone posts an ask every cycle, in the same field, at the same time. When everyone asks, asking stops being a favour-debt and becomes a turn. Roughly 90% of workplace helping happens in response to a direct request — so the ask is the upstream variable, not generosity.
• Do not keep a ledger of who owes whom. Generalized reciprocity outperforms bilateral scorekeeping: "pay it forward" motives were found to be stronger and more durable than reputation-tracking, which decayed. Log contributions to the group, never debts between pairs.
• Cap the cost of saying no. Every intro is double opt-in — ask both sides separately, connect only if both say yes, then drop off the thread. "I'll mention you, no promises, no hard feelings if either side passes" is said out loud as the default.
• Let recognition carry the incentive. A visible count of referrals given converts giving into status, which is the one incentive that does not crowd out the norm.

When a member-to-member engagement goes badly

Worth saying plainly: the public record here is empty. No peer organisation publishes a policy for what happens when one member's paid work for another fails. What follows is structural reasoning, not established practice — but the failure mode is real, and it is how client-services groups die.

• Separate the two relationships in writing before the first engagement. The group is a forum; the engagement is a contract. Scope, price, timeline and termination live in the contract, not in the room.
• Recuse. While two members are in an active commercial dispute, neither processes the other's issues, and the dispute is not group business.
• The group never arbitrates. Peers cannot be both a confidential support forum and a tribunal. Name a neutral outside path — a mediation clause in the engagement contract — and keep the room out of it.
• Pre-agree the exit. Decide now whether a failed engagement is grounds for one party leaving, and who decides. Deciding that while angry is how groups end.
• Never discuss remediation pricing as a group. See the next section — that conversation has a legal name.

Guardrail one: antitrust

This is the largest real legal exposure your group has, it is criminal at the top end, and almost no mastermind material mentions it.

Five to eight owners serving clients in overlapping markets, in a room together, are competitors — for clients, and for the same contractors and staff. That is precisely the configuration the antitrust agencies target.

Wage-fixing and no-poach agreements between competitors are per se illegal, and the agencies have stated they may bring criminal felony charges against both individuals and companies. In the best-known civil case, seven technology companies paid $435 million over agreements not to cold-call each other's employees.

Guardrail two: confidentiality

Confidentiality is not ceremony. It is the input to the only mechanism the group runs on: psychological safety drives learning behaviour, which is what mediates the path to any actual result.

The clause that works, and its exclusions:

Everything discussed in a meeting, and any information about what took place in a meeting, is confidential. It is not shared outside the room with anyone at any time, including spouses and partners. Confidentiality is forever — it survives your leaving the group. Confidential information is not discussed with anyone, in or outside the group, unless the person it concerns is present or has approved it.

Confidential information does not include publicly known information, the names and occupations of members, or the location of a meeting.

The exclusions are what make it workable rather than absurd — without them the rule is unenforceable and people quietly ignore all of it. The "present or approves" clause is the one that stops two members debriefing about a third, which is how a backchannel forms.

Guardrail three: intellectual property

• Get a lightweight signed mutual NDA in place before anyone demos unreleased work. Free, reusable, sign-once standard forms exist. Its real value is not litigation — it is establishing that you took reasonable measures to keep things secret, which is what trade-secret protection turns on.
• If any member might ever file a patent outside the US, this matters more than it sounds. The US gives a one-year grace period for the inventor's own disclosures. Europe gives none. A genuinely confidential disclosure is generally not "available to the public" — but only if the confidentiality obligation is real.
• Co-building needs a written IP assignment before the work starts. This is the trap almost everyone misses: absent an agreement, a US patent co-owner can license or assign without the other's consent and without accounting for profits. Casual brainstorming rarely reaches inventorship. Hands-on co-building of code or designs crosses that line easily.

This section is research, not legal advice. Given that members will subcontract to each other, thirty minutes with a lawyer once, at founding, is proportionate.

The organizer trap is the most common way a group this size dies. The defence is that jobs rotate while people stay.

The drift check is the highest-value role here

In a randomized study of 14,766 newcomers to an online community, those who received a personal invitation were significantly more likely to still be active weeks and months later — and the effect held even for people who never used the thing they were invited to. The invitation itself was the intervention.

The same body of work found the inverse: automated, impersonal enforcement opened a conversation about 7% of the time, while the same decision delivered by a person did so about 60% of the time. Keep the attendance rule mechanical so nobody has to be the enforcer — but make sure a human makes the call.

Buy one day of professional facilitation a year

Hiring a facilitator for a monthly six-person meeting is not economic. Hiring one for the annual retreat is — and the retreat is where the highest-stakes work happens: clearing the air, reviewing the constitution, handing over the moderator role. Brief them explicitly to teach the group to run it themselves afterwards.

What dues do buy is roster retention and perceived value. What they do not buy is people in the room. And the comparison in section 02 is the one to hold onto: where both have been tested against each other, meeting rhythm moved behaviour severalfold while financial stakes moved it essentially not at all. If you are choosing where to spend effort, spend it on the calendar, not the penalty schedule.

Two refinements follow:

• Bill quarterly, not annually. The pull of a payment decays with time since paying. Annual dues buy attendance in month one, not month ten. Quarterly billing re-triggers commitment four times a year.
• Keep dues at cost. Venue, food, one shared tool. No margin, and no facilitator fee unless you actually hire one.

Fines, if you use them

Small, immediate, mechanical, and paid into the group's own pot — never to an individual. Group-contingent stakes substantially outperform identical individual stakes, and the advantage outlives the money: in a controlled trial a group-split incentive beat an identical per-person incentive, and the difference persisted twelve weeks after payments stopped.

The point of a fine is not punishment. It converts lateness from a social event into a transaction, which is less corrosive. What makes small fines credible is one hard backstop: more than fifteen minutes late and you do not come in, and it counts as an absence.

Peer groups almost universally do not instrument themselves. There is no published benchmark for most of this — which means the targets below are drawn from adjacent evidence, and that you will be creating your own baseline rather than matching one.

On satisfaction scores: use them to start conversations, not to make decisions. The claim that net promoter scores are superior predictors of growth failed to replicate when independently tested across 21 firms and more than 15,500 interviews. Renewal and attendance are your real metrics.

The admission ladder

Admission is the most consequential act the group performs — good peers and bad peers cancel each other out, so who you let in matters more than anything you do once they are in.

Attendance and removal

Exit

Growth

Cap at eight and keep a waiting list. When the list reaches five or six qualified people, branch rather than split: two members leave to seed the new group while the original persists, with a standing joint dinner or annual joint retreat between them. Groups resist splitting something they love — sending two people out is far better tolerated than cutting one group in half.

Run in six-month seasons with an explicit re-up decision at each boundary. This gives everyone a legitimate, non-humiliating exit ramp, and gives the group a scheduled moment to fix what is not working.

Two tests, run at every season boundary

The annual maintenance rhythm

Without this the constitution becomes a dead document within a year:

• Constitution re-read by every member annually, amended only by unanimous consent of those present.
• Clearing the air, quarterly. Announce it one meeting ahead — the announcement alone often prompts people to resolve things privately first, which is the actual mechanism. Schedule it at the start of a meeting with slack, so it cannot run out of time.
• Moderator scorecard at month four of the term. Anonymous, collected at the end of a session, read alone afterwards.
• Moderator handover at the retreat.
• A high-energy, interactive session after any long break. Holidays are the dangerous window — activity dips become spirals.

In order. Everything here happens before the first real session.

1. Name the purpose in one sentence. Not a list. Purpose dilution is the most-cited reason groups die — try to do many things and you never get great at the one that matters.
2. Draw up the candidate list. You want twelve to fifteen plausible names to land six to eight members. If you cannot get that many genuinely interested, do not convene yet.
3. Run the conflict check across the list. Who could be in the same pitch? Who employs whom? Who is whose dominant client?
4. Run the admission ladder on each candidate, including the site visit. Slowly. This is the highest-leverage work you will do.
5. Book the founding half-day. Longer than a normal session.
6. At the founding session, write and sign the constitution together. Not circulated for approval — written in the room. Section 17 is the clause list.
7. Say the antitrust rule out loud at that session, and agree the redirect for the pricing question before it is ever asked.
8. Sign the mutual NDA before anyone shows unreleased work.
9. Run one practice round and one practice hot seat. This matters more than the document. Most groups fail at norms because nobody has ever seen the norm executed.
10. Lock twelve dates and twelve venues into everyone's calendar before you leave the room.
11. Publish the rotation — who hosts, who is in the hot seat, who demos — for the whole season.
12. Set up the two tools and nothing else. A phone-native group chat, and one shared doc.
13. Post the first written check-in two weeks before session one, so the loop is already running when you first sit down.

Write these together at the founding session. The signing is not ceremony — it is what makes the automatic rules feel legitimate rather than imposed when they eventually fire on someone.

1. Purpose — one sentence, and the test you will apply to future decisions: does this take us toward it or away from it?
2. Confidentiality — the full clause from section 10, including the exclusions and the "present or approves" line.
3. Antitrust — the never-discussed list, and the named redirect for pricing questions.
4. Attendance — two misses or the retreat means automatic removal; unanimous vote to return; fiscal year defined explicitly.
5. Lateness — half an absence plus a fixed amount; late means one second past the start time by a named clock; past fifteen minutes you do not come in; explicit carve-out for travel delays and genuine emergencies.
6. Preparation — the hot seat frame circulates a week ahead, or presenter and coach both pay in.
7. Devices — off or silenced throughout; email and voicemail at breaks only.
8. Food, drink and dinner — nothing eaten during the hot seat; the alcohol decision; dinner expected, with two skips a year.
9. Business between members — any transaction above a named threshold is disclosed to the group. Pick the number now.
10. Referral norms — referrals are gifts, subcontracted work is invoiced at commercial rates, no ongoing commissions inside the group.
11. Disputes — one-to-one first, then the moderator, then a vote. The group never arbitrates a commercial dispute; that goes to an outside mediator.
12. Removal — two thirds of unaffected members for values misalignment.
13. Exit — one month's notice, exit presentation required, unspent dues forfeited otherwise.
14. Emergency meetings — available to any member in need; attendance not mandatory.
15. Review — re-read annually, amended only by unanimous consent of members present.

This manual was built from research across five areas: peer-advisory protocols, technical-builder meeting formats, small-group operations and retention, value creation and measurement, and session-design craft. Where a claim rests on a specific study, the study is named in the text.

What is solid

• The anchor randomized trial on monthly owner peer groups — large, well-identified, published in a top-five economics journal, with effects persisting after the intervention ended.
• Progress monitoring, goal-setting and implementation-intention effects — large meta-analyses, consistent direction, honest effect sizes.
• The motivation-crowding literature on paying for things that were previously gifts.
• The antitrust guidance and case law, quoted from primary agency documents.

What was deliberately excluded

• The "95% accountability partner" statistic. Untraceable to any real research — see section 02.
• The "1953 Yale goal-setting study." Debunked in 1996; it never happened.
• Vendor-published referral and community statistics where the organisation's own regional pages contradicted each other.
• Widely-quoted figures that do not appear in the papers they are attributed to. Several popular productivity numbers fall into this category.

Where the public record is genuinely empty

Three things in this manual are structural reasoning rather than established practice, and are marked as such in the text:

• What happens when a member-to-member commercial engagement fails. No peer organisation publishes a policy on this. Section 09's guidance is reasoned from adjacent evidence.
• Response rates for community ask-channels. No aggregate data exists anywhere public. Your scorecard will be creating this benchmark, not matching one.
• Any survey ranking what mastermind members actually value most. None exists. The ranking used here is inferred from causal studies instead — stronger evidence than a satisfaction survey would have been, but still an inference.

On the legal sections: this is research, not legal advice. The antitrust and IP material is quoted from primary sources, but your jurisdiction, your industries, and your members' regulatory regimes all matter. One conversation with a lawyer at founding is proportionate to what a room of client-serving owners is actually exposed to.