Commitment Devices and Credible Threats: Schelling's Playbook for Pricing and Competitive Strategy

TL;DR: Competitive strategy is usually taught as the art of keeping options open. Schelling (1960) argued the reverse: the player who can visibly destroy his own options often wins, because a threat you would never carry out deters no one. A large part of what operators call pricing policy is commitment in this sense. Capacity investments (Spence 1977; Dixit 1980), most-favored-customer and most-favored-nation clauses (Cooper 1986; Salop 1986), and price-match guarantees (Png and Hirshleifer 1987; Corts 1997; Moorthy and Winter 2006) are devices whose real product is not a lower price but a change in what rivals believe you will do. Many of them work by making tacit coordination easier to hold together, which is why regulators watch them: Apple and five publishers used agency-plus-MFN contracts to move e-book bestsellers from Amazon's $9.99 toward $12.99 and $14.99 (United States v. Apple, 2013), and the European Commission took commitments from Amazon over e-book MFNs in 2017. Yet the evidence cuts both ways. When Best Buy adopted a most-favored-customer clause, prices fell rather than rose (Chen and Liu 2011), and most advertised price-match guarantees do not fit the collusion story at all (Arbatskaya, Hviid and Shaffer 2004). Commitment works only when it is observable, costly to reverse, and still correct when the world changes. Before copying a rival's policy, an operator should ask which of those three conditions it actually meets.

---

The Meeting Where Someone Says Let Us Just Match Their Price

Every operator has sat in the version of this meeting. A competitor has cut price on the product that carries the category, the sales team is losing deals it used to win on reputation, and someone reasonable proposes the reasonable thing: publish a price-match guarantee. Match any competitor, no questions asked. The logic in the room is that a guarantee removes price as a reason to defect, reassures the nervous buyer, and costs almost nothing because few customers will ever file a claim.

The logic is half right, and the half it gets wrong is the important half. A price-match guarantee is not primarily a promise to your customers. A price-match guarantee is a message to your competitor, and the message is: cutting your price will no longer win you my customers, because I will match you automatically, so you will have started a fight that gains you nothing and costs us both margin. Read that way, the guarantee is not a concession. It is a threat, and its whole value depends on whether the competitor believes it.

That distinction, between a policy aimed at customers and a policy aimed at rivals, is the subject of this essay. My claim is that a large share of what firms file under pricing policy is better understood as commitment: capacity announcements, most-favored-nation clauses, public roadmaps, long-term supply contracts, and price-match and price-protection guarantees are all, in the technical sense, commitment devices. Each of them works by deliberately narrowing the firm's own future choices in order to change what competitors expect. And the resource each of them is really buying, I will argue, is credibility.

Credibility is scarce because flexibility is cheap. Any firm can announce that it will defend its market share to the last dollar. The announcement means nothing, because everyone knows that when the moment arrives, defending to the last dollar will be more painful than accommodating, so the firm will accommodate. What separates a threat that moves a rival from a threat that a rival ignores, as we will see repeatedly, is not conviction or tone. It is whether the threatening firm has arranged its affairs so that carrying out the threat is actually its best move when the time comes. Thomas Schelling built an entire theory of conflict on that single observation, and it remains the most useful lens I know for reading a competitor's pricing behavior and for auditing your own.

Schelling's Inversion: Weakness as Strength

Thomas Schelling published The Strategy of Conflict in 1960, drawing on nuclear deterrence, labor negotiation, and street-level bargaining rather than on any market. His central move was to invert the everyday intuition that having more options is always better. In strategic interaction, Schelling showed, the visible loss of an option can be an advantage, because it changes what the other side expects and therefore what the other side does.

His examples are physical and memorable, which is why they have survived. A general who burns the bridge behind his advancing army has thrown away the option to retreat. In a naive accounting that is pure loss. In strategic terms it is a gift, because the enemy now knows the army cannot retreat and must therefore fight to the end, which changes the enemy's own calculation about whether to attack. The driver in a game of chicken who conspicuously rips off his steering wheel and throws it out the window has destroyed his ability to swerve, and by doing so has forced the other driver to be the one who swerves. The kidnapper who cannot be trusted to release a hostage after payment sometimes wishes he could bind himself to release, because the binding is what makes payment worthwhile to the victim's family.

The power to constrain an adversary may depend on the power to bind oneself.

, Thomas Schelling, The Strategy of Conflict (1960)

The unifying idea is that a threat or a promise has force only when the person making it has no good way to back out. We tend to admire flexibility and distrust anyone who has boxed themselves in. Schelling's contribution was to see that the box is often the point. A firm that retains the freedom to abandon a market at the first sign of trouble has, by retaining it, told every rival that the market is contestable. A firm that has sunk enough into the market that leaving would be ruinous has told every rival the opposite, and the second message, the one we should attend to, is worth more even though it was bought by giving something up.

Commitment in this sense is not a personality trait and it is not a communication strategy. Commitment is a change in the payoffs. A press release announcing that a firm will never cede share is talk, and talk is reversible, so rivals discount it to zero. A half-built factory is not talk. The concrete has been poured, the money is spent whether or not it is ever used at capacity, and everyone can see it. The factory changes what the firm will rationally do next, and because it changes what the firm will do, it changes what rivals expect, and because it changes what rivals expect, it changes what rivals do. Contrary to the boardroom instinct that prizes keeping powder dry, the most powerful competitive moves I have watched are frequently the ones that spend the powder in public and make retreat impossible.

Why Empty Threats Fail

To see why credibility, and not conviction, is what matters, we need the one piece of formal apparatus this essay relies on. It is called subgame-perfect equilibrium, and it is the tool that separates threats a rival should believe from threats a rival should laugh at.

Start with the simplest competitive drama: an incumbent holds a profitable market, and a potential entrant is deciding whether to come in. The incumbent would love to prevent entry, and the way it tries is by threatening a price war: enter, and I will cut price so hard that you lose money and wish you had stayed out. If the entrant believes the threat, it stays out, and the incumbent keeps the market without ever having to fire a shot.

The sequence of moves is what matters, so draw it as a tree. The entrant moves first, choosing to stay out or enter. If it enters, the incumbent moves second, choosing to accommodate the newcomer or to fight a price war.

Loading diagram...

Now reason the way the entrant will. The threat to fight is only frightening if the incumbent would actually fight once entry has happened. But look at the incumbent's choice at that second node. Having been entered, the incumbent compares fighting, which pays it 1, against accommodating, which pays it 4. Accommodating is better. So a rational entrant predicts that the incumbent, whatever it said beforehand, will accommodate, because accommodating is the incumbent's best move once the entrant is already in the room. The threat to fight is an empty threat. The entrant enters, the incumbent accommodates, and the comfortable monopoly payoff of 10 collapses to a shared 4.

Reinhard Selten introduced this idea in 1965, in a study of oligopoly pricing published in Zeitschrift für die gesamte Staatswissenschaft, and refined it a decade later in "Reexamination of the Perfectness Concept for Equilibrium Points in Extensive Games" (Selten, 1975, International Journal of Game Theory 4(1), 25-55), work that contributed to his 1994 Nobel. The intuition is worth stating in plain language because it is the whole game. A Nash equilibrium only requires that each player's plan be a best response to the others' plans. Subgame perfection adds a demand that sounds obvious once said: the plan has to remain a best response at every point where a player actually has to act, including points that the player hoped never to reach. A threat that fails this demand, a threat the threatener would not want to execute if called upon, is not part of any subgame-perfect equilibrium. Rivals should, and eventually do, ignore it.

The condition for a threatened response to be credible can be written in one line. Let R be the aggressive response the incumbent threatens (fight), and let the alternatives be everything else it could do once entry has occurred. The threat is credible only if carrying it out is a best reply in that subgame:

$$\pi_I(R \mid \text{entry}) \;\ge\; \pi_I(R' \mid \text{entry}) \quad \text{for every alternative } R'.$$

In the tree as drawn, the condition fails, because π_I(fight) = 1 < 4 = π_I(accommodate). This is the formal statement of a fact every operator already half-knows: saying you will retaliate is worthless if everyone can see that retaliation would hurt you more than acceptance.

The lesson so far is negative: threats are cheap and mostly empty. The rest of the essay is about the positive move, which is how a firm turns an empty threat into a credible one. Look again at the bottom two rows of Table 1. If the incumbent can, before the entrant decides, take an action that reorders its own later payoffs so that fighting becomes better than accommodating, then the threat stops being empty. The entrant, now facing an incumbent who will rationally fight, stays out. The whole art of commitment is the art of finding that prior action.

Capacity as a Signature in Concrete

The classic prior action is capacity. Michael Spence, in "Entry, Capacity, Investment and Oligopolistic Pricing" (Spence, 1977, The Bell Journal of Economics 8(2), 534-544), and Avinash Dixit, in "The Role of Investment in Entry Deterrence" (Dixit, 1980, The Economic Journal 90(357), 95-106), formalized the idea that an incumbent can deter entry by building productive capacity it does not currently need.

The mechanism is precisely the payoff-reordering of Table 1. Building excess capacity is expensive and much of the cost is sunk, meaning it cannot be recovered by selling the plant. Once the capacity exists, however, the marginal cost of actually using it to flood the market during a price war is low. So the sunk investment does two things at once. It lowers the cost of fighting, because the ammunition is already bought, and it raises the return to fighting, because idle capacity earns nothing while capacity used to defend share earns something. Both effects push the incumbent's post-entry payoff from fighting above its payoff from accommodating. The threat that was empty becomes credible, and precisely because it is now credible, it never has to be used. The entrant, forecasting a fight, stays out, and the capacity sits partly idle, having earned its keep by deterring rather than by producing.

Dixit's insight refined a subtlety we keep relearning. What deters is not capacity as such but the portion of it that is a sunk, irreversible commitment. Reversible capacity, the kind you could quietly resell if entry happened, deters nothing, because the rival knows you will resell rather than fight. The commitment value lives entirely in the irreversibility. This is Schelling's burned bridge wearing an industrial-economics suit.

We can write when the whole maneuver is worth it. Let π_I^M be the incumbent's profit if it deters entry and keeps the market to itself, let π_I^D be its profit if it accommodates a rival, and let K be the sunk cost of the capacity that makes the fighting threat credible. Committing pays when the deterred-monopoly profit, net of the capacity bill, beats the accommodated-duopoly profit:

$$\pi_I^{M} - K \;\ge\; \pi_I^{D}.$$

But the inequality is necessary, not sufficient, and the second requirement is the one firms forget. The capacity only deters if, having been built, it genuinely makes fighting the incumbent's best reply, that is, if π_I^F(K) ≥ π_I^A(K) once K is sunk. A firm that builds a factory too small to actually win a price war has paid K and bought no credibility, because the rival can still see that fighting would lose. Commitment that stops short of the threshold is the worst of both worlds: the cost of the bridge without the deterrent value of having burned it.

The same logic runs, unlabeled, through two industries I get asked about constantly, where capacity is the daily subject. In semiconductors, a leading-edge fabrication plant costs on the order of ten to twenty billion dollars and takes years to build, and firms preannounce these buildouts long before the wafers ship. A press release about future capacity is cheap talk in Schelling's sense, but a fab under construction is not, and rivals treat the two very differently. The half-built fab is a commitment; the roadmap slide is a Nash threat. In cloud infrastructure, providers preannounce regions, capacity expansions, and multi-year price commitments, and the credible ones are backed by data centers already leased and power already contracted. In both industries the observable, sunk, irreversible part of the announcement is the only part that moves a competitor. The rest is atmosphere.

Tough, Soft, and the Four Animals

Capacity deterrence makes it sound as if commitment always means being big and aggressive. Drew Fudenberg and Jean Tirole showed that the truth is more interesting, and occasionally the reverse. Their 1984 paper carries one of the best titles in economics: "The Fat-Cat Effect, the Puppy-Dog Ploy, and the Lean and Hungry Look" (Fudenberg and Tirole, 1984, The American Economic Review 74(2), 361-366). It asks a question the capacity story skips: when a firm makes an early, observable investment, should it invest a lot or a little, and does the investment make the firm behave more aggressively later or less?

The answer turns on a distinction that Jeremy Bulow, John Geanakoplos, and Paul Klemperer made precise in "Multimarket Oligopoly: Strategic Substitutes and Complements" (Bulow, Geanakoplos and Klemperer, 1985, Journal of Political Economy 93(3), 488-511). Two firms' strategies are strategic substitutes when one firm getting more aggressive makes the other want to pull back, which is the texture of quantity competition: if I flood the market, your best response is to produce less. They are strategic complements when one firm getting more aggressive makes the other want to match it, which is the texture of price competition: if I cut price, your best response is to cut yours too. The sign of that cross-reaction decides whether commitment should make you look tough or soft.

The four animals repay slow reading, because each corresponds to a real competitive posture. The top dog over-invests to be big and aggressive, the capacity story of the previous section. The puppy-dog ploy is the surprising one: under price competition, a firm that plans to coexist with a rival may deliberately stay small and unthreatening, because looking tough would only provoke the rival into a mutual price war that hurts everyone. In a market of strategic complements, aggression is contagious, so the profitable commitment is to visible restraint. The fat cat over-invests in something, often a large installed base or a brand, that makes it disinclined to fight on price later, and that very softness keeps prices high for both firms. The lean and hungry look is under-investment that keeps a firm aggressive and nimble, useful when the investment on offer would have made it fat and complacent.

What I take from the grid is not a table to memorize. It is to notice that whether your competitive commitment should be a show of strength or a show of weakness depends entirely on whether your rivalry runs through quantities or through prices. In practice most digital and retail competition runs through prices, which means it is a game of strategic complements, which means the naive instinct to commit to toughness is frequently wrong. The puppy-dog ploy, deliberately declining to build the capacity or make the aggressive move that would trigger a price war, is often the better commitment, and it is almost never the one that gets proposed in the meeting.

Pricing Policies Are Commitment Devices

Everything so far has been about capacity and posture. The richest and most regulated commitment devices, though, are contractual, and they hide in plain sight inside pricing policy. Three families matter: most-favored clauses, price-match guarantees, and the long-term or agency contracts that bind them together. Each works by removing the firm's own freedom to do something, and each thereby changes what rivals expect.

Most-favored clauses

A most-favored-customer clause promises a buyer that it will receive the best price the seller gives to anyone; if the seller later offers someone a lower price, the favored customer gets a refund of the difference. A most-favored-nation clause in a supply or distribution contract does the same across trading partners. Both sound pro-buyer, and that is how they are sold. Their strategic effect can be the opposite.

Thomas Cooper worked this out in "Most-Favored-Customer Pricing and Tacit Collusion" (Cooper, 1986, The RAND Journal of Economics 17(3), 377-388). A most-favored-customer clause is a commitment device pointed at the firm's own future self. By promising every current buyer a refund if it ever cuts price, the firm has made cutting price expensive in a new way: a cut no longer costs only the margin on new sales, it also triggers retroactive refunds to the entire existing base. The firm has, in Schelling's terms, burned the bridge to future discounting. And a firm that has credibly bound itself not to discount is a safer partner in a tacit truce, because its rivals no longer have to fear a surprise cut. The clause that looks like a gift to customers is, functionally, a promise to competitors that the truce will hold.

Steven Salop set this in a wider frame in "Practices that (Credibly) Facilitate Oligopoly Coordination" (Salop, 1986, in Stiglitz and Mathewson, eds., New Developments in the Analysis of Market Structure, MIT Press). His term of art, facilitating practices, names exactly the class of moves this essay is about: contractual and informational devices that make a coordinated high-price outcome easier to sustain without any explicit agreement. Meeting-competition clauses, advance price-change announcements, most-favored-customer terms, and price-match guarantees are his canonical examples. What makes them work, and what makes them legally interesting, is the credibly in his title. A facilitating practice only facilitates if it is a real commitment, which is why regulators look past what a clause says to whether it actually binds.

Price-match guarantees

Return to the guarantee from the opening meeting, now that we have the theory in hand. A price-match guarantee is a commitment that neutralizes the return to undercutting. I. P. L. Png and David Hirshleifer, in "Price Discrimination through Offers to Match Price" (Png and Hirshleifer, 1987, The Journal of Business 60(3), 365-383), gave the first careful treatment, and Kenneth Corts, in "On the Competitive Effects of Price-Matching Policies" (Corts, 1997, International Journal of Industrial Organization 15(3), 283-299), and Sridhar Moorthy and Ralph Winter, in "Price-Matching Guarantees" (Moorthy and Winter, 2006, The RAND Journal of Economics 37(2), 449-465), worked out when the effect softens competition and when it does not.

The softening mechanism is clean. Suppose two firms sit at a common price p*, and firm i posts a guarantee to match any lower price from firm j. Now consider firm j thinking about a cut to p* - ε. In a world without the guarantee, the cut would steal firm i's customers, which is the whole point of cutting. With the guarantee, the customers do not move, because firm i automatically matches, so the customer who would have switched simply gets the lower price from i instead. The gain from undercutting evaporates while the cost, lower margin on your own existing sales, remains. Firm j's deviation payoff becomes

$$\Delta\pi_j(p < p^{*}) \;=\; \underbrace{0}_{\text{no share won, }i\text{ matches}} \;-\; \underbrace{(p^{*}-p)\,q_j}_{\text{margin given up}} \;<\; 0,$$

so no cut is worthwhile and the common price holds. Read this against the repeated-game logic of the folk theorem: sustaining a high price normally requires patient firms and fast detection of cheating, and the discipline collapses when discounting is heavy. A universal price-match guarantee relaxes that constraint from a different direction. Instead of relying on the threat of future punishment, the guarantee makes the deviation unprofitable in the very period it happens, because the matching is automatic and instantaneous. Detection lag, the thing that usually breaks a tacit truce, goes to zero.

Png and Hirshleifer added the second face of the same policy. A guarantee can also be a screening device that separates informed, price-sensitive shoppers from uninformed ones. The shoppers who will actually invoke the match are the ones who found the lower price and bothered to bring proof; the guarantee lets the firm charge a high list price to everyone and quietly discount only to the segment that would otherwise have walked. Seen this way, the guarantee is simultaneously a threat to rivals and a form of price discrimination against your own customers, which is a lot of strategic work for a policy the meeting treated as a low-cost reassurance.

The Apple e-books case

The most instructive real episode I know is the one the United States government eventually prosecuted. When Amazon dominated e-book retailing around 2009, it sold many new releases and bestsellers at $9.99, a price the major publishers hated because it anchored customers below the hardcover economics they depended on. In 2010, as Apple prepared to launch the iPad and its iBookstore, Apple and five of the six largest publishers restructured the business around two commitment devices working together: the agency model, under which the publisher sets the retail price and the retailer takes a fixed 30 percent commission, and a most-favored-nation clause requiring that Apple never be undersold, so that any lower price elsewhere had to be matched in Apple's store.

The MFN was the load-bearing piece, and it was a commitment device aimed at a specific rival. Because Apple's contract guaranteed it the lowest price, each publisher had bound itself: if it let Amazon keep selling at $9.99, it would have to give Apple that price too and eat the loss. The only way out was to move Amazon onto agency as well and take control of the retail price away from the discounter. Which is what happened. Bestseller and new-release e-book prices moved off Amazon's $9.99 and toward agency caps of $12.99 and $14.99.

The district court, in United States v. Apple Inc. (952 F. Supp. 2d 638, S.D.N.Y. 2013), found that Apple had orchestrated a horizontal price-fixing conspiracy among the publishers, a ruling the Second Circuit affirmed in 2015. Richard Gilbert's survey "E-Books: A Tale of Digital Disruption" (Gilbert, 2015, Journal of Economic Perspectives 29(3), 165-184) sets out the price and market-share aftermath: prices of the participating publishers' e-books rose by roughly eighteen percent in the months around the switch, and Amazon's share of US e-book sales fell from about ninety percent before agency toward the low-to-mid sixties over the following two years. Apple ultimately paid $450 million to settle, of which $400 million went to consumers. The point for our purposes is mechanical, not moral: the MFN clause was the commitment that made the coordinated price increase self-enforcing. Once every publisher was bound to deny Amazon a price advantage, none of them had to trust the others to hold the line, because the contracts held it for them.

The European Amazon commitments

The mirror image arrived in Europe, with the roles reversed. By 2015 the European Commission had opened an investigation into Amazon's own e-book distribution contracts, which contained a battery of parity, or most-favored-nation, clauses requiring publishers to offer Amazon terms and features at least as good as those offered to any rival, and to tell Amazon when they gave a competitor something better. The concern was the same commitment logic seen from the buyer side: the clauses removed publishers' freedom to give a new or smaller e-book retailer a better deal, which blunted any rival's ability to compete for supply and helped entrench the incumbent.

In a commitments decision of 4 May 2017 (Case AT.40153), the Commission accepted Amazon's offer not to enforce the parity clauses, and not to include them in new e-book contracts, for a period of five years across the European Economic Area. No fine, no finding of infringement, just the removal of the commitment device. What both the American and European cases share is the recognition that the interesting object is not the price and not the intent but the clause, because the clause is the mechanism that turns a wish for coordination into a binding fact. Regulators, unlike the meeting that proposes the guarantee, read pricing policy as commitment by default.

Where Commitment Fails

The theory is elegant, which is exactly the reason to be careful with it. Commitment devices fail in specific, predictable ways, and the empirical record on pricing policies is far messier than the models. A skeptic reading only the theory would conclude that price-match guarantees and most-favored clauses reliably soften competition. In practice, the data do not say that, and we should follow them where they point.

Consider what happened when Best Buy adopted a most-favored-customer clause in consumer electronics between April 2003 and March 2004. Jihui Chen and Qihong Liu, in "The Effect of Most-Favored Customer Clauses on Prices" (Chen and Liu, 2011, The Journal of Industrial Economics 59), measured the outcome and found the opposite of the collusion prediction. Best Buy's own prices fell by about 1.6 percent after adoption, and its competitors cut prices too rather than settling into a comfortable truce.

The advertised price-match guarantee fares no better as a reliable collusion device once you look at how the guarantees are actually written. Maria Arbatskaya, Morten Hviid, and Greg Shaffer, in "On the Incidence and Variety of Low-Price Guarantees" (Arbatskaya, Hviid and Shaffer, 2004, The Journal of Law and Economics 47(1)), collected 515 low-price guarantees advertised across 37 US metropolitan areas and found that the majority did not fit the facilitating-practice profile at all. Most applied only to rivals' advertised prices, or carried enough hassle, restrictions, proof requirements, and delay, that they could not plausibly deter a determined competitor. A guarantee laden with exceptions is a guarantee that no rival needs to fear.

Pulling the failure modes together, commitment devices break for three reasons, and each maps onto one of the conditions in the theory.

The first is that the rival cannot observe the commitment. A commitment that changes your payoffs but that no competitor can see does not change anyone's beliefs, so it deters nothing. Schelling's driver has to throw the steering wheel out where the other driver can watch it go. A price-match guarantee buried in the terms of service, or a capacity expansion no analyst has noticed, is a bridge burned in the dark.

The second is that reversal turns out to be cheap. The whole force of commitment comes from irreversibility, so any escape hatch destroys it. A guarantee with generous exclusions, a capacity investment that could be resold, a public roadmap the market knows will slip, all of these look like commitments and function like talk. Rivals price in the escape hatch and behave as if the commitment were never made.

The third and most dangerous failure is committing hard to a position that becomes wrong when the world moves. Commitment is the deliberate destruction of your own flexibility, and flexibility is precisely what you want when conditions change. A firm that sank a fab to deter entry into a product category that then gets disrupted has a burned bridge leading to a battlefield everyone else has left. The strength of a commitment in a stable world is exactly its fragility in a shifting one, which is why the most durable commitments are the ones whose underlying bet you would still make after the facts change.

What This Changes for an Operator

The practical upshot, in my experience, is a change in how we read pricing policy, our own and everyone else's. Stop asking what a price-match guarantee, an MFN clause, or a capacity announcement does to customers, and start asking what it does to competitors' beliefs. Most of these policies are commitment devices whose customer-facing story is a cover for their strategic function, and once you see them that way the design questions change.

The commitment lens also runs inward, which is the part I find operators most often miss. The same logic that binds a firm against its rivals can bind a firm, or a team, against its own future self. A future self, like a competitor, has different incentives from the one making the plan, and sometimes the useful move is to remove its options in advance. Deadlines that carry a public, reputational cost if missed are commitment devices against a team's own tendency to slip. Default enrollment and opt-out contract terms bind a future decision-maker toward the choice the present one prefers. Escrow and staged, milestone-gated payments bind both sides of a deal to perform, substituting a mechanical commitment for the trust that neither party fully has. Vendor contracts with termination penalties, and the outside options and deadlines that set the terms of any negotiation, are commitment devices wearing procurement clothes.

Set alongside the other moats a firm can build, commitment is a distinct and often cheaper thing. Customer switching costs, the subject of switching cost engineering, bind your customers to you; commitment devices bind you, visibly, in order to move a rival. The two are complements, not substitutes, and they fail differently. A platform choosing whether to enter a complementor's category, the problem in platform cannibalization dynamics, is making a commitment decision about its own future conduct whether or not it frames it that way. And the repeated-price-war logic that a price-match guarantee is designed to short-circuit, the discounting trap where each firm's rational cut leaves everyone worse off, is treated on its own in the companion essay on the prisoner's dilemma of discounting; commitment devices are one of the few durable exits from that trap, precisely because they change the one-shot payoffs rather than relying on everyone staying patient.

The deepest change the commitment lens made for me is to how I treat my own flexibility, and how I coach operators to treat theirs. The reflexive corporate value is to keep options open, and in operational matters that instinct is usually right. In competitive strategy it is frequently wrong, because the flexibility you keep is flexibility your rival prices in. Schelling's inversion is that the visible, credible, costly destruction of your own options can be the strongest move on the board. The discipline is knowing which options are worth destroying, making sure the destruction is one your competitors can see and believe, and keeping enough judgment to recognize the rare, dangerous case where the bridge you are about to burn is the one you will need.

Key Takeaways

1. Credibility, not flexibility, is the scarce resource in competitive strategy. A threat or promise moves a rival only when the firm has arranged its payoffs so that carrying it out is its best move when the moment comes; Schelling (1960) showed that visibly destroying your own options is often how you arrange them.
2. Empty threats fail by subgame perfection, not by insincerity. Selten's (1965, 1975) refinement says a plan must be a best response at every point a player actually has to act. A price-war threat the incumbent would not want to execute, because 1 is worse than 4 in Table 1, is discarded by any rational entrant; a sunk capacity investment that flips the ranking to 5 over 3 is what makes the same threat credible (Spence 1977; Dixit 1980).
3. Whether commitment should look tough or soft depends on the type of rivalry. Fudenberg and Tirole (1984), with the strategic-substitutes and strategic-complements distinction of Bulow, Geanakoplos and Klemperer (1985), show that under price competition, a game of strategic complements, the profitable commitment is often the puppy-dog ploy of visible restraint, not the top-dog show of force that firms instinctively reach for.
4. Much of what firms call pricing policy is commitment aimed at rivals. Most-favored-customer clauses (Cooper 1986), most-favored-nation clauses, and price-match guarantees (Png and Hirshleifer 1987; Corts 1997; Moorthy and Winter 2006) work by removing the firm's own freedom to discount, which can make a high-price outcome self-enforcing, exactly the facilitating practices Salop (1986) described and regulators police.
5. The cases show the mechanism, not a verdict. Apple and five publishers used agency-plus-MFN contracts to move e-book bestsellers from $9.99 toward $12.99 and $14.99, prices up roughly eighteen percent, ending in a $450 million settlement (United States v. Apple, 2013; Gilbert 2015); the European Commission took five-year commitments from Amazon to drop e-book MFNs in 2017. The clause, not the price or the intent, is the object of interest.
6. The empirical record is mixed, so commitment is a mechanism, not a law. Best Buy's most-favored-customer clause coincided with price cuts of 1.6 percent for itself and up to 3.5 percent for rivals (Chen and Liu 2011), and most advertised low-price guarantees are too hedged to function as a truce (Arbatskaya, Hviid and Shaffer 2004). A device only binds if it is universal, observable, and hard to escape.
7. Commitment fails in three ways, so test for all three. A device does nothing unless the rival can observe it, unless reversal is costly enough to be believed, and unless the committed position is still right when the world changes. The same logic, turned inward, makes deadlines, default terms, and escrow commitment devices against a firm's own future self, with the same failure modes.

Further Reading

• The Prisoner's Dilemma of Discounting: The Promotion Trap, the repeated-game price war that commitment devices are built to escape, treated in full where this essay only gestures at it.
• Signaling Theory and Advertising: Costly Signals, the sibling mechanism, where a firm spends observably to convey private information rather than to bind its own hands; guarantees and warranties sit on the border between the two.
• Switching Cost Engineering and Interoperability, the complementary moat that binds customers to you, where this essay's devices bind you in order to move a rival.
• Platform Cannibalization Dynamics, a platform's decision to enter a complementor's category is a commitment problem about its own future conduct, whether or not it is framed that way.
• Discount Engineering: When Promotions Destroy LTV, the margin cost of the discounting reflex that price-match guarantees and most-favored clauses are designed to discipline.