Every competitor in your category offers three months free. None of you wanted to. The first one did it in a bad quarter, the second matched within six weeks, and now it is simply what the product costs.
You have been looking for who decided this. Nobody decided it, and that is precisely why you cannot argue your way out of it.
Why these three models
The decision is whether to hold a term that is costing you money, and what it would actually take to change it. The features that fire are several firms whose best move depends on what the others do, a market that behaves differently in different periods without anyone announcing the switch, and a structure that keeps producing the same outcome even as the people and the companies change.
Three lenses. Equilibrium produces an equilibrium answer about why the state is stable and what unilateral deviation costs. Regime states produce a cycle answer about which world you are in right now, since the same term is survivable in one and fatal in another. System traps produce a complex answer about the structure that generated the state in the first place, which is the only one of the three that suggests a way out.
1. Stable does not mean chosen
An equilibrium is a combination of strategies where no single player can do better by changing their own move while everyone else holds theirs. That is the whole definition, and two things follow from it that founders consistently get wrong.
The first is that an equilibrium can be worse for everyone than an available alternative and still be perfectly stable. Every firm in your category would earn more with a one-month trial. None of them can go first, because the one that shortens its trial alone loses deals to the four that did not. Stability is not agreement and it is not efficiency. It is only the absence of a profitable unilateral move.1
The second is that nobody has to understand any of this for it to hold. The state was reached by ordinary competitive responses, each one locally correct, and it persists because it keeps being locally correct. There is no conspiracy to find and no cartel to break. The founder searching for the person who set the market price is looking for an author that does not exist.
The practical consequence is uncomfortable. Strategies that assume the equilibrium is a habit will fail. Explaining to the market that everyone would be better off with shorter trials changes nothing, because each firm’s problem is not ignorance. It is that the first mover loses. Any real exit has to change what the deviation costs, and there are only a few ways to do that: change who you are competing with, change what the buyer is comparing, or make your deviation credible enough that others follow rather than take your customers.
Take the middle option seriously, because it is the only one of the three that needs nobody else to move. Changing what the buyer compares is available to you alone. A three-month free trial on a product at UGX 400,000 a month is UGX 1.2m given away per account that closes, and the same UGX 1.2m buys a paid onboarding sprint, a data migration your team performs, or a first quarter billed at half rate against a twelve-month commitment. Each costs you roughly what the trial costs. None of them is a trial, so the column your competitor fills on the buyer’s comparison sheet no longer exists. You have not left the equilibrium. You have declined to be scored on it.
The failure mode is worth naming before you try it. A repackaged term that is visibly the same thing under a new label invites the buyer to convert it back, and you end up granting the trial anyway plus whatever you added. The test is whether your version transfers a cost the trial did not. Onboarding performed by your staff carries an hourly cost the buyer would otherwise absorb themselves, so withholding it is a real decision rather than a formality. If your alternative costs you money and saves the buyer nothing, it is the same discount wearing a different name.
2. Which market are you in this quarter
Equilibrium analysis assumes the game is fixed. It usually is not. The same market alternates between states with different rules, and the switch is not announced.
In an abundant-capital regime, competitors are funded to buy market share and the equilibrium discount deepens, because losing money on acquisition is cheap. In a tight regime, the same discount becomes unaffordable and firms start quietly withdrawing it. The term did not become wiser. The payoffs changed, so the stable point moved.
Treating the state as persistent while you are in it, and as switchable at moments you can partly anticipate, is the right posture. Regime models formalise exactly this: a system that stays in one state with high probability and occasionally transitions, so the useful questions are which state am I in and what is the transition rate, rather than what is the long-run average.2 Averaging across regimes produces a number that describes no period that ever happened.
Do not plan against the average of two markets you have lived through, because you have never once traded in the average.
The founder-facing version is a watch list rather than a forecast. Write down the three observable things that would tell you the regime had switched: a competitor quietly removing the term, a funding round in your category that does not close, a buyer asking for annual prepayment. When two of the three appear, the cost of deviating has fallen and the window for changing your terms is open. Windows in this structure are short, because everyone else is watching the same signals.
The watch list has a failure mode of its own, which is acting on one signal. Each of the three appears regularly for reasons that have nothing to do with the regime. A competitor removes the term because a new head of sales wanted a clean page. A round fails to close because that company was weak. A buyer asks for annual prepayment because their own budget year is ending. Two signals inside a quarter is the threshold worth acting on, and the reason for pairing them is that a single event usually has a private explanation while two rarely share one.
The watch itself is cheaper than it sounds. Three named competitors and one visible source for each: the pricing page, the standard contract a shared customer will show you, what a lost deal reports the other side offered. Check on a fixed day each month rather than continuously. Ten minutes, against the alternative of learning about the switch from your own renewal numbers a quarter after the window opened.
3. The structure that keeps rebuilding it
The first two models describe the state. Neither explains why this particular state and not another, and that question has a structural answer.
Certain configurations of feedback produce the same pathological result reliably, whoever is in them. Two are directly relevant here. Escalation is a pair of reinforcing loops where each party’s position is set relative to the other, so any move by one triggers a matching move by the other and the whole system ratchets in one direction until an external limit stops it. Drift to low performance is the structure where the standard being measured against is itself based on recent performance, so a run of bad results lowers the bar, which lowers effort, which lowers results.3
Read your market’s terms against those two shapes. A trial length that only ever gets longer is escalation. A quality bar that has quietly moved because everyone benchmarks against last year’s competitors rather than against what the buyer needs is drift. Both are structures, not decisions, and both have known exits: escalation ends when one party refuses to respond and can survive the gap, or when an external constraint binds. Drift ends when the standard is fixed to something absolute rather than to recent performance.
What makes escalation hard to see from inside is that every individual move in it is defensive. Nobody in a trial-length ratchet believes they are escalating. Each firm believes it is responding to a competitor who moved first, and all of them are correct, which is the signature of a reinforcing loop rather than of a strategy.4 If every party believes they are only matching, you are inside the loop, not managing it.
The value of naming the structure is that it tells you which intervention is wasted. In an escalation, matching faster is the failure mode dressed as competitiveness. Unilateral restraint only works if you can survive the period where you look worse, which returns you to the runway question and not to the strategy question.
One distinction the shape-matching blurs. Not every downward move is escalation. A competitor with a genuinely lower cost base is not ratcheting, they are pricing where they can profitably sit, and your refusal to respond changes nothing about their economics. The test is whether the mover is worse off at the new level than at the old one. If they are, it is a loop and someone declining to answer can break it. If they are not, it is a cost position, and the response is a different buyer or a different unit of sale rather than a stare-down you will lose slowly.
Prepare too for the competitor who does the opposite of what restraint assumes. The escalation exit assumes that when you stop responding the others eventually stop as well, because they are also bleeding. Some will read your restraint as weakness and push further, to take the deals your hold has released. That does not make the decision wrong, but it changes the horizon: you are now outlasting one firm rather than resetting a market. Decide before you begin how many quarters of that you will fund, write the number down, and treat reaching it as a plan rather than a defeat.
What the three say together
- State the equilibrium plainly. What term is universal in your category, and what would happen to you alone if you dropped it tomorrow.
- Name the regime. Is capital in your category abundant or tight this quarter, and what three signals would tell you it had switched.
- Identify the structure. Is this escalation, drift, or neither. The three have different exits and only one of them is patience.
- Then decide whether you are changing the term or changing the game. Most founders attempt the first and only the second works.
Where they disagree
Equilibrium logic and system-trap logic give opposite instructions about unilateral action.
The equilibrium view says do not deviate alone, because the deviation is by construction unprofitable while everyone else holds. The escalation view says someone has to stop responding or the ratchet continues to the limit, and the only exit that does not require an outside force is exactly the unilateral one. Both are correct. They differ on whether you survive the interval.
The resolution is not clever. It is your balance sheet. A firm that can absorb two quarters of losing deals on terms can end an escalation and will usually be followed, because the others are also suffering. A firm that cannot should not attempt it, and should instead compete somewhere the equilibrium does not bind: a different buyer, a different unit of sale, a different comparison set. Calling that a repositioning rather than a retreat is accurate, since the point is to enter a game with different payoffs rather than to lose the current one slowly.
What none of them contain
None of the three prices your reputation for holding terms, which behaves differently from the terms themselves. A firm known never to discount faces a different negotiation from a firm with the same list price and a history of exceptions, and the equilibrium description treats them identically.
None of them handles the buyer who is also a player. All three treat demand as the environment. In many markets a single large buyer sets the term by refusing to sign anything else, and then the analysis is a bargaining problem with one dominant party rather than a symmetric equilibrium.
And none of them tells you whether the equilibrium is about to be broken by something outside the market. A regulatory change, a platform policy, a new entrant with a structurally different cost base. Those arrive from outside the model and the model gives no warning, which is a reason to keep a list of them rather than to trust the analysis.
The one action that survives the ignorance: write one sentence describing exactly what happens to your pipeline in the ninety days after you unilaterally drop the term. If you cannot write it with a number in it, you do not yet know whether you are trapped or merely uncomfortable, and finding out costs one week of pipeline analysis rather than a quarter of revenue.
Who has to move
This is a founder decision and it cannot be pushed to sales, because sales is the part of the company most exposed to the deviation and will correctly resist it. The instinct under margin pressure is to hold the term and cut cost elsewhere, which preserves the equilibrium and funds it out of your own operations. The cheapest first test is a single segment: hold the shorter term for one buyer type where you have the strongest differentiation, and watch what your competitors do rather than what your prospects say. Competitor response is the only reliable evidence about whether the equilibrium binds, and it arrives within a quarter.
Sources and notes
- Jeffrey Carpenter and Andrea Robbett, Game Theory and Behavior, MIT Press. Nash equilibrium, the distinction between stability and efficiency, and the standard examples in which the equilibrium outcome is worse for every player than an available alternative are developed in the opening chapters on simultaneous-move games. Used in section 1 for the definition and for the two consequences drawn from it. The book also documents the experimental record, which matters here: real players do not always land on the equilibrium immediately, and convergence depends on repetition and feedback.
- Sanjit S. Dhami, The Foundations of Behavioral Economic Analysis, Oxford University Press, and the regime-switching literature it draws on. The claim used in section 2 is structural rather than empirical: where a system alternates between persistent states with occasional transitions, the average across states describes no individual period. No transition probability is claimed for any real market here, because these are not estimated for private-market terms and any number offered would be invented.
- Donella H. Meadows, Thinking in Systems: A Primer, Chelsea Green Publishing. The system traps, including escalation and drift to low performance, are set out with their structures and their documented exits in the chapter on system traps and opportunities. Used in section 3 for both archetypes and for the specific point that escalation ends either through an external limit or through one party refusing to respond.
- John D. Sterman, Business Dynamics: Systems Thinking and Modeling for a Complex World, McGraw-Hill. The treatment of reinforcing loops and of the conditions under which competitive matching behaviour ratchets a system in one direction supports the reading of trial-length escalation in section 3. Cited for the loop structure, not for any measured magnitude.
A note on a number this article does not give. There is no figure for how long a unilateral deviation costs you before competitors follow. It depends on how visible your terms are, how fast your category’s sales cycle runs, and how much slack your competitors have. What transfers is the question, which is whether you can survive the interval, and that number is one you already have.
Joshua Agonya Pi’Rwot, Founder.