Every company
you build joins
the ones already
selling.

Cold at one percent. Warm at ten.

Every company Cupel builds acquires customers in a defined buyer segment. Those relationships do not disappear when a company reaches its reading. They accumulate. After three years of cohorts concentrated on repeat segments, the portfolio holds hundreds of customer relationships inside markets Cupel already understands.

A new cohort company can either start cold against that market, or launch into it warm, introduced by companies its buyer already pays. Cold outbound converts somewhere near one to two percent. A warm introduction from a vendor the buyer already trusts converts at six to fifteen. That gap is the whole argument.

Cohort one starts from nothing. Cohort twelve launches into a base of existing customers, a dataset on what converts in that segment, and a set of founders who have already sold into it. The GO rate stops being a fixed assumption and becomes a function of portfolio age.

At ten active companies a new build clears the $100K threshold from the network alone. At twenty it clears it twice over before a single cold email goes out.

Load-bearing condition. This only holds if cohorts concentrate on repeat buyer segments. Rotating markets every cohort produces a scattered portfolio with no shared customer base and destroys the entire effect. Concentration is not an operational preference. It is the thing the compounding depends on.

• Customer introductions. The main event. A member introduces its customers to another member's product where it genuinely fits.
• Co-selling. Two or three members serving the same buyer sell as a package.
• Shared services. Compliance, tooling, insurance, and processing negotiated at portfolio rates.
• Benchmarks. Anonymised operating metrics across the portfolio, segmented by buyer.
• Talent. Operators, contractors, and founders who have worked inside portfolio companies.

Who belongs.

Any company Cupel has built, at any reading, that opts in. Membership is offered at formation and the company may decline with no penalty and no effect on anything else in its agreement.

Membership survives Cupel's equity exit.

This is the clause that determines whether this is a network or a leash. A founder who repurchases Cupel's stake stays in, on identical terms, because the network is worth belonging to independently of who holds equity. If membership depended on the cap table it would break the moment anyone could afford to leave, and every member would know it.

Membership terminates on change of control.

An acquirer has no reason to inherit these obligations and every reason to resent them. Commissions already earned on closed deals survive. Forward obligations end at closing.

Twenty percent of first-year contract value.

Introductions need to be paid for. Goodwill produces a flurry in month one and nothing by month four. At a $500 monthly contract that is $1,200 per closed referral. High enough that a founder will spend a Tuesday afternoon writing introductions, low enough that the receiving company is still far ahead on a customer it would not otherwise have.

On the attribution window. Earlier drafts used ninety days. That is too short for the contract sizes in this portfolio. Considered B2B purchases at $500+ monthly routinely run four to six months from first conversation to signature, and a window that expires before the deal closes teaches members that referring is pointless. A hundred and eighty days matches the actual sales cycle.

First year only, deliberately. Recurring commissions create perpetual obligations that follow a company into diligence and complicate every future transaction. A one-year term is standard, clean, and does not surface as a liability in an acquisition.

What counts as an introduction.

A warm introduction to a named individual, made by someone at the referring company, that receives a response. Logged at the time it is made. Mentioning a company at a conference, posting about it, or forwarding a link does not qualify. The bar is deliberately specific because attribution disputes are the fastest way to poison a network.

Attribution conflicts.

First logged introduction wins. If two members introduce the same buyer, the earlier ledger entry takes the commission. Retroactive logging is not permitted. This rule is blunt on purpose. An adjudication process would cost more in relationships than the commissions are worth.

If the customer was already in the receiving company's pipeline, the receiving company may dispute within fourteen days of the introduction being logged, with evidence of prior contact. After fourteen days the introduction stands.

Cupel administers the ledger and never touches the money.

That distinction matters. A shared record of who introduced whom is administration. Collecting and disbursing funds between companies would make Cupel a payment intermediary, with the licensing questions that carries. Members settle bilaterally against a statement Cupel publishes.

• Members log introductions as they make them, in the shared ledger.
• Members log closed deals against introductions, with contract value.
• At quarter end Cupel issues a statement to each member showing gross owed, gross due, and the net position against every counterparty.
• Members settle directly, within thirty days of the statement.
• Unsettled balances are flagged on the following statement and visible to the counterparty.

Netting is what makes this workable. A member with eight introductions out and five in does not process thirteen invoices. It receives one statement with a net figure per counterparty, and in most quarters settles with two or three companies rather than a dozen.

Non-payment.

A member more than sixty days late on a settled balance is suspended from receiving introductions until it clears. That is the only enforcement mechanism and it is sufficient. The cost of exclusion exceeds any commission worth withholding.

Where two or more members sell into the same buyer, they may package. Revenue splits by list price contribution unless the parties agree otherwise in writing before the deal is presented to the customer. No commission applies inside a co-sell. The split replaces it.

One company owns the relationship and the contract. Packaged selling with divided account ownership produces confusion at renewal and finger-pointing at churn. The owning company is named before the first joint call.

Keep the allocation simple. Elaborate formulas kill deals that would otherwise close, and the amounts at stake in early cohorts do not justify complexity.

Individually no leverage. Collectively a portfolio.

Compliance is the clearest case. SOC 2 is table stakes for every B2B software company selling upmarket. A first audit runs $25K to $40K standalone. A portfolio relationship with one auditor cuts materially into that. Across ten companies the saving is six figures.

• Insurance. E&O and cyber cover at portfolio rates, which small companies cannot negotiate alone.
• Payment processing. Volume tiers that no individual member reaches.
• Tooling. Support, analytics, and CRM licences at aggregated pricing.
• Legal templates. MSAs, DPAs, order forms, and security questionnaire responses, drafted once and reused.
• Contractor bench. Vetted designers, engineers, and SDRs who have already delivered inside the portfolio.

Participation is optional and priced at or near cost. This is not a services revenue line. Its purpose is to make the portfolio structurally more profitable than the sum of its parts, which raises the value of every position Cupel holds.

Every member reports a short monthly set: ARR, new logos, churn, average contract value, sales cycle length, and stage conversion. Cupel aggregates and publishes anonymised medians and quartiles by buyer segment. Each member sees its own position against the distribution.

This costs almost nothing and is worth a great deal. A founder who cannot tell whether eleven percent annual churn is good or catastrophic in their segment is guessing, which is the thing this entire operation exists to stop. It also feeds the wedge bank directly. Cohort results become the evidence that scores the next set of wedges.

No company-level figures are ever shared between members. Aggregates and distributions only, with a minimum of five companies in any segment before it is reported.

The most under-appreciated channel.

A founder whose company received a KILL reading has spent ninety days building and selling under real pressure and now has no company. That person is frequently the strongest available candidate for an Assemble seat in the next cohort, or for an operating role inside the studio.

The network maintains a bench: founders between companies, contractors who delivered inside the portfolio, sales people who have sold into the relevant segments. Members hire from it first.

The founder network outlives the companies. Membership in the talent side is personal, not corporate, and survives a KILL.

• The referring company owns the relationship. Introductions come from that company, to its customer, for a product it genuinely believes helps. Cupel does not run outbound into another member's customer base and members do not receive lists.
• No customer data moves. Ever. Not a list, not an export, not a CRM sync. An introduction is a person deciding to connect two other people. Anything else creates privacy exposure and destroys the trust the whole thing runs on.
• Reciprocal by construction. Every member is simultaneously a channel and a beneficiary. There is no tier that receives without contributing.
• Free-riding self-corrects. A member that takes introductions and makes none simply pays commissions and receives none. No quotas, no contribution requirements, no committee deciding who is pulling their weight.
• Quality over volume. A bad introduction costs the referring company its customer's trust, which is worth far more than $1,200. The commission is calibrated to make good introductions worth making, not to make many introductions worth making.
• Cupel takes nothing. No override, no administration fee, no percentage. The network's value to Cupel is the compounding GO rate across every future cohort, which is worth an order of magnitude more than any skim. Taxing it would slow the adoption that produces the value.

Networks like this normally exist on a slide. Four mechanisms prevent that.

• The directory. Every member with its buyer, product, contract value, customer count, and the segments it sells into. Visible to all members, updated monthly.
• The channel. A shared Slack or Discord where introduction requests happen in public. A member posts what it is looking for. Anyone who can help responds. Public requests get answered. Private ones get ignored.
• The quarterly round. Ninety minutes, every member present. Each nominates three to five introductions it can make in the coming quarter and commits to them in writing. This single mechanism produces more activity than any amount of encouragement, because a commitment made in front of peers gets kept.
• A named owner. Network activation is somebody's job, not everybody's. Jack through the first cohorts, the Head of Go-to-Market from year two. Unowned, it decays within two quarters.

Terms are set by Cupel and published. Rate, window, settlement mechanics, and the rules above.

Changes require sixty days' notice and apply prospectively only. Introductions already logged settle under the terms in force when they were made.

Once the network exceeds fifteen active members, material changes go to a member vote. Simple majority of active members, with Cupel holding no vote. At that scale the members have more at stake in the mechanics than Cupel does.

Removal.

A member may be removed for repeated non-payment, for sharing customer data, or for introductions made in bad faith. Removal requires written notice and a thirty-day cure period except in the case of data sharing, which is immediate.

Exit is always available.

Any member may leave on thirty days' notice. Obligations on closed deals survive.

The sequencing mistake to avoid is building the full apparatus for Cohort 01. Three companies with forty customers between them do not need a settlement process. They need each other's phone numbers and a norm that introductions get paid for.

The Network Participation Agreement is a standalone document, separate from the founder agreement, executed between the member company and Cupel as administrator, incorporating the terms above by reference.

Standalone because it must survive Cupel's exit from the member's cap table. If the network obligations lived inside the founder agreement they would terminate with it, and a founder who bought out Cupel's equity would fall out of the network on a technicality.

The founder agreement contains only the opt-in clause, which references the Network Participation Agreement and confirms the right to decline at formation without consequence.

• The commission is a contractual obligation between member companies. Cupel is administrator, not a party to the payment obligation, and never holds funds.
• No exclusivity. A member may refer to non-members freely and owes nothing on those.
• No representation or warranty passes between members on introduced products. An introduction is not an endorsement and the referring company carries no liability for a member's performance.
• Confidentiality covering the directory, the benchmarks, and the ledger, surviving termination.
• Data protection: the agreement should state expressly that no personal data of customers is transferred between members, since an introduction is a communication rather than a transfer.

• Nobody uses it. The most likely outcome and the one to design against. Mitigated by the quarterly round, the commission, and a named owner. If two consecutive quarterly rounds produce fewer than five introductions, the network is theatre and should be either fixed or dropped.
• A bad introduction damages a customer relationship. Mitigated by the quality bar and by the referring company owning the decision. One member burning a customer on a weak referral will make no further introductions, and word travels.
• Attribution disputes turn into resentment. Mitigated by first-logged-wins and the fourteen-day objection window. Blunt rules beat fair processes here.
• It reads as extraction. A founder who feels their customer base is being farmed for other people's companies will leave and tell others. Mitigated by reciprocity, by Cupel taking nothing, and by the survives-exit clause, which is the clearest signal that the network exists for members rather than for the studio.
• Buyer segments drift. If cohorts stop concentrating, the customer base fragments and referrals stop being relevant. This is the failure mode that matters most, because it originates in intake rather than in the network, and by the time it shows up in referral volume it is two cohorts old.