The ceiling on this relationship is pre-commitment

You negotiate every deal individually. Each one is a fresh conversation, each price is defensible, and you believe the flexibility is an asset. Your margin has not moved in two years.

There is a ceiling on what this can produce, and you are under it. The most you can earn against a stream of one-shot counterparties is what you would get by committing publicly, first.

Why these three models

The decision is whether to keep negotiating case by case or to publish a policy and hold it. The features that fire are one long-lived party facing many short-lived ones, an irreversible move whose option value you would be destroying, and a benefit that arrives well after the cost.

Three lenses that genuinely disagree. The ceiling says commit. Real options says committing destroys something valuable and prices it. Delay says the payoff arrives after the pain, which is why almost nobody holds the policy long enough to see it work. The disagreement between the first two is the actual decision.

1. The bound, and why cleverness cannot beat it

Set the situation up properly, because the asymmetry is what drives everything.

You are long-lived. You will be here next year and the year after. Your counterparties are short-lived in the relevant sense: each customer, each small buyer, each one-off purchaser is deciding once and is myopic about the future of the relationship.

In that structure there is a bound on what you can earn. A patient long-lived player facing short-lived opponents is limited by the pure-action Stackelberg payoff, which is what you could guarantee yourself by committing to a public action before the other side moves.¹ That is a ceiling, not a target, and the important part is what it rules out.

Mailath and Samuelson work through why the mixed-action version, which looks higher on paper, cannot be reached in any equilibrium regardless of how patient you are.¹ The implication for an operator is blunt: no amount of skilful case-by-case negotiation gets you above what visible pre-commitment would have earned. If you are below the ceiling, the lever is commitment. If you are at it, further negotiation effort is pure cost.

The mechanism is worth stating plainly because it is the reason the ceiling exists at all. Reputation works by putting probability mass on your being a type that cannot do otherwise, not a type that chooses well.¹ Every discretionary exception you grant is evidence that you can do otherwise, and it moves that probability the wrong way. Flexibility is not free even when each individual exception is correct on its own terms.

Stackelberg ceiling, the bound lens

• Assumes: you are long-lived and patient, your counterparties are short-lived and myopic, and a public commitment is observable.
• Fits because: you face many small buyers who each decide once.
• Breaks when: your counterparties are also long-lived, where the reputation results are considerably weaker.
• Evidence: grade B plus. Formally established under stated conditions, and the conditions are strong.
• Counteracts: the belief that better negotiation raises the ceiling.
• May reinforce: rigidity in situations where the counterparty is not short-lived at all.

2. What the commitment costs you

The second lens is the honest objection, and it is a real one rather than a formality.

Committing publicly destroys an option. The ability to price case by case has value precisely because the world is uncertain: a new segment appears, a competitor moves, a large buyer arrives with a genuinely different cost to serve. Flexibility is the right to respond to information you do not have yet, and that right has positive value which rises with volatility and with how long you would hold it.

So the decision is not commitment against flexibility in the abstract. It is the ceiling gain against the option value destroyed, and both sides are real. Where your market is stable and your counterparties are small and many, the option is worth little and the ceiling gain is worth a lot. Where the market is volatile and a single deal can be a large share of revenue, the option may well dominate.

That gives a usable test. How often in the last two years did a discretionary exception turn out to be correct in a way a published policy would have prevented? If you can name three, your flexibility is earning its cost. If you cannot name one, you have been paying option premium for an option you never exercised.

Real options, the flexibility lens

• Assumes: the right to decide later has positive value, rising with uncertainty and with the time you hold it.
• Fits because: committing is irreversible in the way that matters, since a policy you break was never a policy.
• Breaks when: the flexibility is nominal, exercised only under pressure rather than on information.
• Evidence: grade A. Formally established, and the hard part is estimating the volatility rather than the logic.
• Counteracts: committing to a policy in a market that is still moving.
• May reinforce: indefinite deferral, since there is always more information coming.

3. Why nobody holds the policy long enough

The third lens explains the specific way this fails in practice, and it is not a failure of conviction.

The costs of commitment arrive immediately. The first buyer who asks for an exception and is refused may walk, and that loss is visible, attributable and this week. The benefits arrive later and diffusely: buyers stop asking, the discount conversation disappears, the average price drifts up across a book you cannot easily attribute to the policy.

That is a delayed system with a fast-arriving cost, and correcting against it produces the standard failure. Three weeks in, with one lost deal and no visible gain, the policy gets its first exception, which is granted quietly and for good reasons. The exception is then observed, and the probability mass the whole thing depended on moves back.

The repair is the same as in any delayed system: decide the review date before you start, and do not read the result before it.³ Anything else is correcting at full strength against a signal that has not arrived.

Expect to underrate the lag specifically. Reasoning about delayed consequences defeats highly educated adults in controlled conditions, and the failure is not attributable to graph literacy, contextual knowledge, motivation or cognitive capacity.²

Delay, the sequencing lens

• Assumes: the cost of commitment lands before its benefit, and the benefit is diffuse.
• Fits because: the first refused exception is visible and this week.
• Breaks when: the policy is genuinely wrong, where early pain is real information and holding is stubbornness.
• Evidence: grade A. Structural, and the human failure to reason about it is well replicated.
• Counteracts: abandoning a policy during the interval where it can only look expensive.
• May reinforce: holding a bad policy past the point of evidence.

The levers, cheapest first

• Count your exceptions. How many were granted in the last year, and how many turned out to be right? Two numbers, one afternoon, and they usually settle the argument.
• Check the counterparty type. If your buyers are few and long-lived, this whole analysis does not apply and you should keep negotiating.
• Publish the smallest policy that binds. One term, stated publicly, not the whole price list. Commitment works by being observable, not by being comprehensive.
• Remove your own discretion. A policy you can quietly waive is not a commitment. Put the exception behind a second signature or a published process.
• Write the review date before you publish. Long enough for the benefit to arrive, decided while you are calm.
• Keep the option where it is worth something. Commit on the terms that are stable, stay flexible on the ones that genuinely move.

What to do this quarter

Do now, sized at one afternoon, effect immediate. Count the discretionary exceptions of the last twelve months and mark each as correct or not in hindsight. Reversible, free, and dominant across every scenario about whether flexibility is earning its keep.

Hedge, where the premium is the whole loss, live before the next cycle. Publish one term and hold it for a stated period. If commitment was the wrong call you have lost some flexibility on a single term for one cycle, and that is the entire downside.

Defer and trigger, size fixed now. Do not publish a full policy. Pre-commit the trigger: if the exception count for that one term drops to zero over the stated period, the policy extends to a second term automatically. Decide the period and the second term now, because a decision made after one lost deal is made under exactly the conditions section three describes.

Note the arrivals, because they are the whole difficulty. The cost of the first refusal arrives in days. The benefit arrives across a quarter and cannot be attributed to any single deal. Anyone reading the result early will read a loss.

What usually happens next

Run the break test first. Has a rule changed, has an actor entered or left, has a measurement become a target? A new competitor offering the flexibility you just gave up changes the calculation genuinely, and holding the policy on principle at that point is the error this article would otherwise cause.

If nothing broke, the pattern is consistent. The policy is published, held for six weeks, breached once for a large buyer with a good reason, and the breach becomes known. Within two months the negotiation conversation has returned in full, and the conclusion drawn is that customers in this market simply expect to negotiate.

There is a second regularity worth expecting, and it is about who asks. Published policies are tested hardest by your largest counterparties, because they have the most to gain from an exception and the most leverage to request one. So the first serious challenge arrives from the account you can least afford to lose, which is precisely the case the policy exists to handle and precisely the one where holding it feels most reckless.

Subtract the counterfactual before crediting a margin improvement. Prices that rose in the quarter you published a policy may have risen because a competitor exited. The test is whether buyers stopped asking, which is the mechanism, rather than whether the number moved.

What this ensemble cannot see

All three lenses assume your commitment is observable. Most are not.

The ceiling result depends on the counterparty seeing your public action before they move. A policy that exists in your internal documentation and is communicated deal by deal is not a public commitment, it is a negotiating position, and it earns none of the benefit while paying the full flexibility cost. That is the worst square of the whole grid and it is where most companies sit.

There is also a strong condition in the central model. The result holds for a patient long-lived player facing short-lived opponents, and the same source is explicit that when the counterparty is also long-lived there is no direct link from their belief to their best response, which leads to considerably weaker reputation results.¹ Enterprise sales to a handful of large accounts is that case, and this article does not apply to it.

And one property none of these models contains: the exception you refuse is a specific person having a bad day, and the benefit is an abstraction spread over people you will never meet. The models are symmetric about that and nobody experiences it symmetrically.

The one action that survives the ignorance: this week, count last year’s exceptions and how many were right. If you cannot name three that a published policy would have wrongly prevented, publish one term, tell your buyers, and put the waiver behind someone else’s signature.

Who has to move

The person who needs this is whoever grants the exceptions, and each one is defensible in isolation, which is why the pattern survives. The cheapest first test is the count: exceptions granted, exceptions vindicated. If the second number is near zero, the flexibility was never an asset, and you have the argument in two numbers rather than in principle.

Sources and notes

1. George J. Mailath and Larry Samuelson, Repeated Games and Reputations: Long-Run Relationships, Oxford University Press, 2006. Long-lived and short-lived players, including the minmax construction and the constraints on achievable payoffs, are sections 2.7 and 3.6. The Stackelberg payoff, defined as what a player can guarantee by committing to a public action before the other moves, together with the result that the mixed-action Stackelberg payoff cannot be achieved in any equilibrium and that no equilibrium gives the long-lived player more than the pure-action Stackelberg payoff independently of the discount factor, is developed at section 2.7 and revisited in section 15.4.1. Commitment types and the adverse-selection basis of reputation, in which reputation rests on probability mass attached to a type that cannot deviate, are section 15.2. The weakening of reputation results when the second player is also long-lived is noted at the end of section 15.1 and developed in chapter 16.
2. Matthew A. Cronin, Cleotilde Gonzalez and John D. Sterman, Why don’t well-educated adults understand accumulation? A challenge to researchers, educators, and citizens, Organizational Behavior and Human Decision Processes 108(1), 2009, pages 116 to 130. Author copy: https://www.mit.edu/~jsterman/CroninGonzalezSterman061210.pdf. The abstract states that highly educated people are often unable to infer the behaviour of simple stock-flow systems, and that persistent poor performance is not attributable to an inability to interpret graphs, contextual knowledge, motivation, or cognitive capacity.
3. John D. Sterman, Business Dynamics: Systems Thinking and Modeling for a Complex World, McGraw-Hill. The treatment of delays between action and response as the standard source of oscillation, and the requirement to damp the correction or shorten the delay rather than respond at full strength to a lagged signal, is chapter 17. Used in section 3 for the rule that the review date is fixed before the policy is published.

A note on the tension this article does not resolve. The first model says commit and the second says commitment destroys something valuable. Both are correct and the balance depends on your volatility and your counterparty mix, neither of which this piece can measure for you. What it does supply is the count of vindicated exceptions, which is the closest thing to a measurement available and is usually decisive once someone actually produces it.

Joshua Agonya Pi’Rwot, Founder.