Strategic Commitment

SKILL.md

---
name: strategic-commitment
description: "Activate when: someone says 'they don't believe we'll follow through', 'how do I make my threat credible?', 'how do we deter a competitor from entering?', 'we need to lock in this customer/partner', 'should we burn our bridges?', or when a stated threat or promise is being discounted by the other side.
  Do NOT activate when: the problem is capability (you cannot execute the threat, not just credibility that you will), or when the environment is changing so fast that any lock-in creates more rigidity than credibility gain."
---

# Strategic Commitment

## Overview

Strategic commitment converts "I might do X" into "I will do X" — not through rhetoric but by *changing payoff structure* so follow-through is the rational move. Schelling (1960, Nobel 2005): keeping options open undermines you when opponents predict you'll rationally back down. A commitment device removes that prediction by making retreat more expensive than execution.

Composes with: [`prisoners-dilemma`](../prisoners-dilemma/SKILL.md) (diagnose defection traps) · [`zero-sum-game`](../zero-sum-game/SKILL.md) (confirm game structure) · [`batna-zopa`](../batna-zopa/SKILL.md) (BATNA sets outside option; commitment governs how credibly you communicate it).

## When to Use

Apply when:
- Designing a competitive deterrence posture — making a potential entrant believe you will fight, not accommodate
- Making a promise or pledge credible to a counterparty who doubts your follow-through
- In negotiation where the other side exploits your visible flexibility
- Signaling strategic intent to multiple audiences with a single costly action
- Evaluating whether a competitor's stated threat is real or theatrical

**When NOT to use:** commitment is unobservable to the target audience; environment changes so fast that lock-in outweighs credibility gain; cost of following through exceeds the value being protected; genuine one-shot anonymous game with no reputational audience.

## Coaching Novices (Adaptive Front Door)

- **Engine mode:** user has a concrete case → run The Process directly.
- **Coach mode:** user is unfamiliar → guide step by step.

In Coach mode, respond one step at a time. Each [WAIT] is a hard stop — output only that step's question, then stop.

1. What-it-is: commitment makes threats credible by locking yourself in so follow-through is the rational choice.
2. Check fit: does the other side disbelieve your threat? If problem is *capability* (can't execute), not credibility, stop here.
3. Elicit their case: specific threat, audience, why they doubt follow-through.
> **[WAIT — do not advance until user responds]**
4. Run CAVE audit. Ask: "What happens if you don't follow through? What is the current cost of retreat?"
> **[WAIT — do not advance until user responds]**
5. Name which device fits and why it changes payoffs — not just rhetoric.
> **[WAIT — do not advance until user responds]**

## The Process

Run the **Credibility Audit (CAVE)**, then design the **Commitment Device**.

1. **State the threat precisely:** action · trigger · audience. Example: "If X enters at price Y, we match within 30 days."
2. **CAVE audit** (all four required or the commitment is theatrical):
   - **C — Capability:** resources to execute?
   - **A — Action taken:** irreversible action making retreat costly?
   - **V — Visibility:** can the audience observe the commitment?
   - **E — Enforcement:** third party (contract/regulator/reputation) penalizing non-execution?
   Weakest element = design target.
3. **Cost of retreat now:** if retreat is cheap, threat is not credible. Device must raise this cost until retreat is no longer dominant.
4. **Choose device:** sunk-cost · reputation · third-party constraint · organizational lock-in.
5. **Test post-commitment game tree:** re-draw payoffs — is follow-through non-dominated at execution? Stop-rule: if retreat still dominant, device is insufficient.
6. **Rigidity risk:** (a) scenarios where follow-through becomes irrational? (b) exit clauses for material-change? (c) does conditionality undermine deterrence?

### Output: Commitment Design Report

```
Threat/promise: <action · trigger · audience>
CAVE: C=<present/absent> · A=<irreversible action> · V=<observable?> · E=<enforcement mechanism> · Weakest=<letter>
Cost of retreat now: <what happens if you don't follow through?>
Device chosen: <sunk-cost / reputation / third-party / org lock-in — reasoning>
Post-commitment payoffs: before=<retreat dominant?> · after=<retreat dominant?> · deterrence=<yes/no/partial>
Rigidity risks: <scenario> · exit clauses: <...> · conditionality: <undermines deterrence?>
Recommendation: <device · implementation · what opponent should now believe>
```

*→ Method in Action: [Amazon's Everyday Low Price Commitment (1994–present)](examples/amazon-everyday-low-price-commitment-1994-present.md)*

## Commitment Packs

**Entry Deterrence:** Irreversible capacity investment lowers marginal cost of price war. Without sunk cost, backward induction shows accommodation — threat incredible.
**Negotiation:** Credibly narrow your BATNA. Walk-away must be verifiable or reputation-backed; unverifiable claims are cheap talk.
**Brand Guarantees:** "Lifetime warranty" works when honoring costs are sunk into brand value. Startups need contract-based enforcement instead.

## Applying It Well

- Distinguish capability from credibility — commitment devices address credibility only; solve capability first.
- Announcement without action is the most common error — change *payoff structure*, not rhetoric.
- Rigidity is a real cost; design exit clauses for material-change conditions before committing.
- Third-party enforcement is most reliable; reputation-based fails when audience is inattentive or future is heavily discounted.
- Apply CAVE to the *opponent's* threats too — many are theatrical.

*→ Primary sources: [references/sources.md](references/sources.md)*

## Common Rationalizations

**[D] = designed upfront | [O] = observed in real use. [O] entries are more valuable.**

| Fake move | Reality |
|---|---|
| [D] "We announced publicly — it's credible now" | Raises reputational cost of retreat but not material payoff at execution. If retreat is still cheaper, opponents call the bluff. |
| [D] "Our track record proves we follow through" | Track records work in *similar* situations. A new game with new stakes requires commitment re-established in that context. |
| [D] "The contract has a penalty clause" | Only if penalty exceeds breach gain and enforcement is fast. Clauses set below defection gain are not commitment devices. |
| [D] "Burning bridges is too risky — we need flexibility" | The cost of flexibility is reduced credibility. Requires explicit comparison, not a default preference for optionality. |
| [D] "Our CEO's word should be enough" | CEO declarations carry reputational cost for the individual but no structural lock-in. Contracts and sunk investments are stronger. |
| [D] "The threat will deter — they're rational and know we'll fight" | Rational opponents reason backward: if fighting is costly at execution, they will enter. Rationality is *why* commitment devices are necessary. |
| *→ Add [O] entries here after each real use — paste the actual failure pattern* | *What went wrong and why* |

## Red Flags

- Announcement only — no structural payoff change · CAVE not fully run
- Post-commitment game tree not drawn — retreat may still be dominant
- Rigidity risk not assessed · opponent's threat taken at face value without CAVE
- Device depends on a single individual · penalty clause below defection gain

## Verification

- [ ] Threat stated precisely: action, trigger, audience named
- [ ] CAVE applied: all four elements evaluated; weakest identified
- [ ] Current cost of retreat quantified — if cheap, commitment is theatrical
- [ ] Device type chosen with reasoning
- [ ] Post-commitment game tree drawn: retreat no longer dominant at execution
- [ ] Rigidity risk assessed: at least one costly follow-through scenario addressed

---

*Part of **deciqAI Knowledge Skills** — open-source thinking skills that make rigor executable for AI agents. Built by deciqAI · https://deciqai.com · Contributions welcome — see the template at the repo root.*