Commitment Device

A small product studio has six promising prototypes.

Each week brings another insight. The team changes direction, rebuilds the shared foundations to cover every possible future, and starts one more experiment before choosing which earlier one deserves to become a product.

The work is real and technically impressive. Nothing becomes dependable enough for a customer to adopt. One collaborator has to remember how all six prototypes work and can't plan around any delivery date because the priority changes each week. Eventually, she stops planning around any of them.

By refusing to choose, the studio keeps all six prototypes unfinished.

Reopening the choice each week keeps every prototype unfinished

Optionality is valuable while uncertainty is high and new information can change the decision. It becomes expensive when the repeated ability to reconsider prevents any option from receiving enough time, trust or effort to work.

A Commitment Device is a constraint you choose now so a future plan is easier to keep and harder to abandon when hesitation, temptation, or a changing mood arrives.

You choose it while the long-term reason is still clear.

The device changes which options are available later and what it costs to abandon the plan. It may remove an alternative, make progress visible to someone else, or reserve resources that can't be casually reassigned.

Timing matters. Install the constraint while you can see the whole decision, before the predictable moment when short-term comfort becomes unusually persuasive.

A six-week commitment with a real exit

The studio chooses one customer problem for a six-week delivery bet.

It schedules a demonstration with three prospective customers, assigns one product person and one engineer, writes down a narrow test for success and puts the other prototypes on a dated idea list. New ideas can be recorded, but they can't displace the active bet before the review date.

The team also defines evidence-based exits:

• Stop if the required technology can't meet a named safety condition.
• Re-scope if users repeatedly fail the same core task.
• Continue through ordinary boredom, a more exciting idea or discomfort about showing incomplete work.

The demonstration date makes the commitment visible, the assigned pair can't be moved to another prototype, and the review date prevents the choice from becoming permanent.

The collaborator can plan again. Customers receive something concrete enough to judge, and the studio gets evidence from a product customers can actually try.

The six-week limit gives the team enough time to finish one bet without letting each new idea restart the decision.

Use the smallest constraint that will hold

A useful device doesn't always require severe or irreversible consequences. Think of commitment as a ladder:

1. Reminder: write the decision and the reason it was made.
2. Friction: remove distracting tools, block another option or make switching require a deliberate process.
3. Visibility: tell a colleague, customer or group what will be delivered and when.
4. Reserved resources: allocate money, time, inventory or capacity that can't be casually redirected.
5. Stake: use a deposit, bond, contract or other consequence that makes abandonment costly.
6. Irreversibility: eliminate the alternative entirely.

Move only as high as the behaviour requires. If a calendar block is enough, a binding contract adds unnecessary risk. If you routinely ignore calendar blocks, the device needs more consequence.

The strongest device is rarely the best by default. Choose the least restrictive constraint that still works at the moment you usually abandon the plan.

How commitment works in different fields

Psychology and behavioural economics

People often value immediate comfort more heavily when it becomes immediate. Someone who calmly plans to exercise tomorrow may reverse the decision when tomorrow arrives. A pre-paid class, scheduled partner or blocked distraction changes that later choice. The earlier decision changes what is available when immediate comfort becomes more tempting.

The analogy between different versions of the self has a limit. New information may genuinely make the later self better informed. The device needs a way to distinguish predictable avoidance from evidence that the plan should change.

Strategy, law and credible promises

A promise changes what another person can safely plan only when they believe it will hold. A deposit that is lost, money held by a neutral third party, or a contractual penalty makes reversal costly. Governments also bind future choices through constitutional rules. These constraints can support cooperation that a statement of intent can't.

Credibility can also become coercion. A contract negotiated under unequal power or an exit penalty that traps someone in harm shouldn't be defended merely because it produces predictable behaviour.

Software and product architecture

Software and product teams make commitments too. They agree on a stable way for systems to talk, set a final date for moving old data, or freeze changes before an important release so one team's convenience can't keep disrupting everyone else. A product deadline can also force unresolved preferences to become real decisions.

Every hard limit needs a way out for genuine failure. Without a safe way to undo the change or an emergency exception, a rule that protects focus during normal work can make an unusual failure much worse.

In each case, the earlier decision removes an option or makes abandoning the plan more costly later.

Decide what still needs to stay flexible

Use a Commitment Device when:

• You keep reopening a decision without receiving meaningful new evidence.
• Short-term emotion predictably defeats a long-term preference.
• Other people can't plan because your commitment remains provisional.
• A project needs sustained effort before results become visible.
• Competing priorities repeatedly consume reserved capacity.
• Nobody can plan around a promise when abandoning it has no consequence.

Before installing one, write:

1. The action you want to protect.
2. The moment you usually abandon or renegotiate it.
3. The smallest constraint that would survive that moment.
4. The evidence that should legitimately release the constraint.
5. The date when the commitment must be reviewed.
6. The people who would pay for a wrong commitment.

Exit conditions should depend on evidence, safety, or an agreed change in circumstances. “I feel less excited today” is usually the fluctuation the device exists to absorb.

When commitment starts protecting the wrong thing

A Commitment Device can lock effort into a bad decision, turn an experiment into an identity or encourage people to interpret every warning as weakness. Public declarations can make changing course embarrassing even after the original assumptions fail.

A founder may announce an aggressive deadline while the team works the nights and weekends required to meet it. The founder chose the commitment, but the team pays for it.

There's also a selection problem: someone may commit strongly before doing enough exploration. Confidence doesn't improve the underlying option.

Use time limits, agreed reasons to stop, emergency exceptions and review from someone outside the decision. Preserve the ability to exit for reasons you chose in advance, while making it harder to abandon the plan only because today's mood changed.

Nearby models

Real Options preserve the ability to expand, delay, change or abandon a commitment while uncertainty is valuable. Commitment Devices deliberately narrow future choice when continued optionality is preventing execution. They often belong in sequence: preserve options while learning can change the decision, then commit when focused action becomes more valuable than another round of possibility.

Path Dependence explains how earlier choices can unintentionally constrain what remains possible. A Commitment Device creates a chosen form of dependence for a defined purpose. Both can produce lock-in, so the origin of the constraint doesn't guarantee that it remains useful.

Keeping possibilities alive can improve a future choice. Once there's enough evidence to choose, protect one option long enough to test whether it can actually work.