In a diligence process, the moment that reveals most about founder relations is rarely a moment of visible disagreement. It is the pause. A management presentation reaches a question about pricing authority, or about which of two product lines gets the next hiring allocation, and one founder answers while the other looks at the table; the answer is given, the meeting continues, and nothing about the exchange is minuted. Three weeks later, the data room produces a set of board minutes in which the same question appears on the agenda of four consecutive meetings, each time carried forward, each time without a recorded resolution. Nothing in that record says the founders are in conflict. Everything in it says the company has no channel through which a disagreement between them terminates.
The same pattern appears in the calendar rather than the minutes. Companies where founder disagreement has no exit tend to accumulate decisions that are technically pending but functionally abandoned — a supplier consolidation that has been under review for eleven months, a compensation framework that was drafted and never approved, a second-site lease that is renegotiated annually because the decision to commit was never taken. Each of these has a plausible individual explanation. Taken together, and cross-referenced against which founder sponsored which initiative, they map with uncomfortable precision onto the fault line between two people who have chosen avoidance over adjudication.
The mechanism underneath this is not dysfunction; it is a rational economy of relationship capital. In a founding team, the working relationship is the company's most load-bearing asset, and both founders understand this at a level below articulation. Escalating a disagreement to the point of formal resolution consumes that asset, and the consumption is immediate and certain while the cost of deferral is diffuse and deferred. Under those conditions, avoidance is the cheaper option on any given Tuesday. The problem is not that founders make this trade; it is that the trade remains attractive at every individual decision point while the aggregate cost compounds invisibly across the whole set of deferred decisions.
There is a second layer to it. Early-stage founding teams typically operate on undifferentiated authority — both founders decide everything, jointly, by convergence rather than by rule. This works, and works well, precisely because it produces decisions that both parties own and neither will undermine. But it is an operating model calibrated for a company small enough that the founders can converge on every material question within the time the question allows. Once the volume of material decisions exceeds that bandwidth, the same model that produced alignment begins producing latency, and the founders experience this as increased friction rather than as an outgrown structure. The instinct is to work harder at the relationship. The requirement is to redesign the decision architecture.
The balance-sheet expression of this is indirect, which is why it is frequently missed by the founders themselves and almost never by an experienced acquirer. Decision latency shows up as extended sales cycles when pricing exceptions require both signatures and the signatures are not co-located in time. It shows up in working capital as inventory positions that reflect a purchasing policy nobody has been authorized to change. It shows up in personnel data as elevated turnover in the second management layer, because a director who reports functionally to two founders with divergent views on their mandate will, within roughly a budget cycle, either learn to serve whichever founder is more recently annoyed or leave for a company with one boss.
In the transaction itself, the exposure is priced through structure rather than through headline value. An investment committee that identifies an unresolved founder dynamic without a governing mechanism will typically respond in one of three ways: it will extend the earn-out period so that a founder split occurring after closing does not fall entirely on the buyer; it will raise the escrow proportion and widen the warranty coverage around key-person and management-continuity representations; or it will introduce closing conditions requiring an executed deadlock provision, a defined casting mechanism, and sometimes a non-founder chair, before funds move. Each of these is a cost. None of them appears in the price line, which is why founders often conclude that the issue did not affect the deal.
The diligence question is more specific than most founders anticipate, and it is asked across six distinct planes. Does a defined mechanism exist at all, or is the answer a verbal assurance that the founders have always worked things out. Is it documented — is there a deadlock article in the shareholders' agreement, a board charter defining escalation, a written delegation of authority that names which decisions belong to whom. Has it been exercised in practice, and can the exercise be evidenced. Is the effectiveness of the arrangement observable in any data series, or is it asserted. Is there an accountable owner for the mechanism who is not one of the parties it governs. And, most decisively, does the arrangement survive the departure of either founder, or does it work only because these two particular people are willing to make it work.
The failure is almost never at the first plane and almost always at the fifth and sixth. Most companies of any institutional maturity have a deadlock clause somewhere in their constitutional documents, drafted at the last financing round by counsel and never read since. The clause is real. It is also, in the ordinary case, structurally inoperable — it triggers a shotgun mechanism or a forced sale, which is to say it resolves the deadlock by dissolving the partnership. A mechanism whose only setting is catastrophic is a mechanism nobody will invoke over a hiring dispute, which means it does not manage conflict at all; it terminates a company that failed to manage conflict.
What actually works occupies the space between informal convergence and the shotgun clause, and it has four separable components. First, a decision-rights map that assigns unilateral authority by domain, so that the majority of disagreements never become deadlocks because only one founder holds the pen. Second, a graduated escalation path with time limits attached — a defined number of days after which an unresolved question moves from the founders to the board, rather than remaining in the founders' inbox indefinitely. Third, a casting mechanism vested in a party who is not a founder, whether an independent director, a designated chair, or, in smaller structures, a nominated external arbiter named in advance rather than selected in the moment. Fourth, a decision register that records what was decided, by whom, on what date, and over what stated objection — the objection field being the component that most registers omit and the one that carries the evidentiary weight.
In the engagements we run, the work begins with the register rather than the agreement, because the register is what produces the evidence the other three components will later be judged against. We open the decision log at the point of proposal, not at the point of approval, so that the interval between the two becomes a measurable series — decision latency by domain, month over month — and we record dissent as a field rather than as a footnote, which means a founder who disagreed can be shown to have disagreed, been heard, and been overruled through a defined route. Two derived metrics come out of this almost immediately and both are legible to a diligence team: the share of material decisions resolved within their stated window, and the frequency with which a decision is reopened after being taken. Reversal frequency is the more diagnostic of the two, because a decision that keeps returning to the agenda is the signature of an authority question that was never actually settled.
The second half of the work is the ownership plane, and it is the part founders resist longest. Introducing a non-founder casting vote is experienced as a transfer of control, and it is one; but the control being transferred is control over a narrow class of questions the founders have already demonstrated they cannot close. We define that class explicitly rather than generally — the domains, the thresholds, the trigger conditions — and we rehearse the mechanism on a live but non-existential question before it is needed on an existential one, because a governance instrument invoked for the first time under real pressure will be contested on procedure rather than on substance. A mechanism that has been used twice on ordinary matters is a mechanism; one that has never been used is a clause.
The continuity test is what separates the two at exit. An acquirer's question is not whether these founders resolve their disputes; it is whether the company would resolve an equivalent dispute between the two people who hold these roles in three years, when neither of them is a founder and neither has the moral authority that founding confers. If the answer depends on the specific forbearance of specific individuals, the capability is personal and the buyer is acquiring a dependency. If the answer is a documented route with a named non-party decision-maker and a record of prior use, the capability is institutional and transfers with the shares. That distinction, rather than the presence or absence of conflict itself, is what the valuation is actually responding to.
Every company with more than one founder will generate disagreements it cannot resolve by convergence; this is a property of distributed authority, not a symptom of a bad partnership. The question a diligence process is asking has never been whether the founders get along. It is whether the company has built somewhere to put the disagreement when they do not.