Eighteen months in, your co-founder tells you they are done. They are not angry, they are tired, and they hold twenty-eight percent with fourteen months vested.
The conversation everyone has is about the twenty-eight percent. The number is the part you can compute, which is exactly why it absorbs all the attention it does not deserve.
Why these three models
The decision is what to settle, in what order, with someone who is leaving and will hold a claim on the company for years afterwards. The features that fire are a division where the two sides have very different alternatives, two reputations that will describe this event to other people, and a residual stake whose value depends on work only one of you will now do.
Three lenses. Bargaining produces an equilibrium answer about what actually determines the split. Reputation decay produces a cycle answer about the two standings this writes. Options produce a random answer about what the retained equity is and why it is so hard to price. The second is the one founders skip and the one that costs most, because it operates on everyone you have not spoken to yet.
1. The split is decided by the alternatives, not by the contribution
Every conversation about a departing co-founder is conducted in the language of fairness and settled by the structure of the alternatives. It helps to see that clearly rather than to resent it.
What each side can do if there is no agreement is what sets the outcome.1 So write both. Yours: continue running the company with an absent shareholder holding a large stake, which is survivable and makes your next fundraise materially harder. Theirs: walk away with whatever has vested, which is a real asset requiring nothing further from them.
Notice that the vesting schedule is not a detail of the negotiation, it is the negotiation. It is the only thing that has already decided what happens if you do not agree, which makes it the reference both sides are actually arguing from. If there is no vesting, there is no structure, and you are negotiating from equal claims with unequal futures, which is the situation that produces the worst outcomes and the longest ones.
Put numbers on it first, because the vesting position is usually less dramatic than either side assumes. On a four-year schedule with a one-year cliff, fourteen months vests roughly twenty-nine percent of the grant. Against twenty-eight points that is about eight points earned and twenty sitting unvested. The conversation you were dreading, the one about twenty-eight percent, is a conversation about eight. The other twenty return to the company under a document you both signed, and the only live question is whether you accelerate any of it as part of the settlement.
Two moves follow. Separate what is contractually settled from what is being negotiated, out loud, early: this much has vested and is theirs, and the conversation is about the rest. That single sentence removes most of the heat, because a large part of the tension is each side not knowing which parts are in play.
And find the trades that are not equity. A departing founder often wants things the company can give cheaply: a title for their next conversation, a defined story, an agreed reference, time to tell people themselves, a longer option exercise window. Those are worth a great deal to them and cost you little, and they are the only currency available for buying back equity the schedule has already granted.
Expect them to dispute the vested line anyway, and know in advance which disputes are real. Two are. Whether the cliff was met is a date question and settles in a minute. Whether acceleration was promised verbally is the one that does not, because founders routinely say things in year one that nobody papered. If you said it, honour it and price it into the trade. If you did not, say so once and move on. Three weeks of arguing about a remembered sentence costs more than the points in dispute.
The failure mode of writing out the alternatives is that you may discover yours is the weaker one. An absent shareholder holding twenty points is not merely awkward at the next raise. It is one of the first things a lead investor asks about, and some will pass rather than price it. If that is your position, do not open with a low offer you cannot defend. You are the party who needs the deal, and a leaver who works that out halfway through will reopen everything already agreed.
2. This settlement writes two reputations
The second lens covers what happens outside the room, and it is the one that operates for years.
Standing is not a balance that accumulates and then sits there. It is a flow: topped up by what people observe recently, decaying when they observe nothing, and under imperfect observation it is temporary rather than permanent.2 Which means the account of this departure is not written once and filed. It is re-told, and the version that circulates is the one that is easiest to tell.
If you do not agree the story, two versions will exist, and yours will not be the one told in rooms you are not in.
So the highest-value item on the agenda is not the percentage. It is a single agreed paragraph: what happened, why, and what each of you will say. Written down, agreed by both, used by both. It costs nothing, it takes twenty minutes, and it is the only mechanism that prevents the ambiguity being filled in by whoever is asked first.
Keep it to four sentences and one true reason. A statement with no reason invites the ecosystem to supply one, and the supplied reason is always worse than the real one. A visibly stage-managed statement reads as a cover-up to exactly the people you were trying to settle. If the honest reason is that the two of you stopped agreeing on where the company was going, write that. It is common, it is survivable, and it closes the topic.
If you cannot get agreement on the account, get agreement on the smaller thing: what neither of you will say. One line of mutual non-disparagement, plus a named list of the people you will each tell first, does most of the work of a joint statement at a fraction of the negotiation. That is the version that survives an angry leaver.
The audiences are specific and should be handled deliberately. Your team, who will read the departure as information about the company and will ask each other rather than you. Your investors, who have seen this before and whose real concern is whether the cap table is now a problem. And your market, which is small enough in most African ecosystems that both of you will meet the same people for the next decade.
Handle the team first and in person, before the announcement. A team that hears it from the founder with a clear account behaves entirely differently from one that infers it from a calendar invitation, and the difference persists long after the facts stop being interesting.
Then assume the account travels faster than your plan for it. In a market where the whole funder pool is a few dozen people who see each other monthly, an investor will hear a version of this from someone other than you within a fortnight. That is the real deadline. Not the announcement date you set, the first conversation you did not control.
3. What the retained stake actually is
The third lens explains why the buyback price is so hard to agree, and it is not because either side is being unreasonable.
A retained stake is a claim on an outcome that will now be produced by work only one of you does. That makes it an option: the holder has upside with no further obligation, and the person carrying the obligation is you. The value of that option rises with uncertainty, which is why an early-stage stake is genuinely hard to price and why both sides can hold defensible numbers that are far apart.3
The departing founder is not wrong to value it highly, because the range of outcomes is wide and they are entitled to the upside of the period they worked. You are not wrong to value it lower, because you are the one bearing the years of effort that produce it.
The productive response is to stop arguing about today’s value and to make the payoff conditional, which is what both parties actually believe anyway. A buyback at a formula tied to the next priced round rather than to a number agreed today. A partial buyback now with the remainder retained. A right of first refusal so the stake cannot arrive on a competitor’s cap table.4
Each of those has a failure mode worth naming before you pick one. A formula tied to the next priced round hands the leaver a stake in your fundraising and no obligation to help it, and it stalls completely if you spend three years growing on revenue instead of raising. Put a fuse on it. If no priced round closes within twenty-four months, the price defaults to a stated multiple of trailing revenue, or to an independent valuation you both pre-agree to accept. A mechanism with no expiry is an argument you have merely scheduled.
A right of first refusal fails the other way. It is cheap to sign and hard to exercise, because it asks you to find the money on a thirty-day clock at a price a third party set. Pair it with a transfer restriction naming who cannot buy at any price, and it starts protecting you rather than merely notifying you.
The reason a mechanism succeeds where a number fails is worth stating plainly, because founders resist it as a fudge. Two parties who cannot agree on a value can very often agree on a rule, because a rule does not require either of them to concede that their estimate was wrong. You are not splitting the difference, you are agreeing what evidence would settle it.
All three convert an unresolvable disagreement about present value into an agreement about a future observable, which is the standard move whenever two parties disagree about a number and would agree about a mechanism.
What the three say together
- State what is already vested and therefore not in play. Do it first and out loud.
- Agree the paragraph before the percentage. One account, written, used by both.
- Tell the team in person before the announcement, and the investors before they hear it elsewhere.
- Convert the valuation disagreement into a mechanism: a formula tied to the next round, a partial buyback, a right of first refusal.
Where they disagree
The bargaining view and the reputation view give opposite advice about pace.
Bargaining says settle quickly. Every week the situation persists, the departing founder becomes less engaged and more anchored, other people get drawn in, and the zone of agreement narrows. The reputation view says the settlement is a durable public artefact, that a fast agreement is usually one side’s draft accepted under fatigue, and that the resentment produced by a rushed deal outlives the deal by years in a market where you will keep meeting.
The resolution is to separate the two clocks, because they are attached to different items. Settle the story and the announcement fast, in days, because ambiguity there compounds immediately and nothing about it improves with waiting. Settle the equity slowly enough to take advice, because it is durable, it is documented, and no part of it degrades in three weeks. Founders usually invert this, delaying the announcement while ambiguity spreads and rushing the equity because the conversation is uncomfortable.
What none of them contain
None of the three prices the possibility that they are right to leave and you should too. All of them assume the company continues and one person exits. Sometimes a co-founder’s departure is the first honest assessment anyone has made, and every lens here will help you negotiate efficiently past a conclusion that deserved examining.
None of them handles the co-founder who wants out and will not say so. The version in this article is the tractable one, because it was stated. The common version is months of reduced engagement that nobody names, and none of these lenses fires until someone says the sentence.
And one property the ensemble will not produce: your own account of the relationship is unreliable at exactly this moment. Both of you will construct a version in which the other changed, and both versions will be sincere. The written paragraph is valuable partly because it forces one account to exist while you can still both be in a room, which is a narrow window and it closes quietly.
The one action that survives the ignorance: before any negotiation, write down what is already vested and what happens if you never reach agreement. Give that to them as well. Most of the fear in the conversation is each side not knowing the other’s fallback, and it is a document rather than a discussion.
Who has to move
The remaining founder, and the sequencing is the whole intervention: the paragraph and the team conversation in days, the equity over weeks with a lawyer. The instinct is the reverse, holding the announcement until the equity is agreed, which lets ambiguity run for a month while everyone constructs their own account. The cheapest first test costs one hour: draft the paragraph, send it to them, and see whether you agree on what happened. If you do not, you have found the real negotiation and it was never the percentage.
Sources and notes
- Jeffrey Carpenter and Andrea Robbett, Game Theory and Behavior, MIT Press. Bargaining, disagreement points and the result that the division reflects the parties’ alternatives rather than their contributions are developed in the chapters on bargaining. Used in section 1. The corollary drawn here, that a vesting schedule functions as the disagreement point and therefore as the reference for the whole negotiation, follows from the standard treatment rather than being stated in it.
- Martin W. Cripps, George J. Mailath and Larry Samuelson, Imperfect Monitoring and Impermanent Reputations, Econometrica 72(2), 2004, pages 407 to 432, developed at book length in George J. Mailath and Larry Samuelson, Repeated Games and Reputations: Long-Run Relationships, Oxford University Press, 2006. Under imperfect monitoring a reputation is temporary and must be replenished rather than accumulated once. Used in section 2 for treating both standings as flows that this settlement writes into, and for the point that the account which circulates is the one easiest to re-tell.
- Sanjit S. Dhami, The Foundations of Behavioral Economic Analysis, Oxford University Press, and the real-options literature it references. The property used in section 3 is that the value of a claim carrying upside with no further obligation rises with the uncertainty of the underlying outcome, which is why two parties can hold defensible and widely separated valuations of the same early-stage stake. No option value is computed, since the inputs do not exist for a private company at this stage.
- Jeffrey Carpenter and Andrea Robbett, Game Theory and Behavior, MIT Press. Mechanism design, and the general move of resolving a disagreement about a value by agreeing instead on a rule that maps a future observable to an outcome, underpins the recommendations at the end of section 3. This is the same structure as an earn-out or a ratchet, and it works here for the same reason: both parties can accept a mechanism whose output neither of them can predict.
A note on a number this article does not give. There is no standard buyback discount and no fair percentage for a departing co-founder. It depends on the vesting position, the stage, and what the leaver still wants from the relationship. The transferable part is the sequence: what is already settled, then the story, then the mechanism, and only then the number.
Joshua Agonya Pi’Rwot, Founder.