The cost of not building
Budgets record what projects cost. They rarely record what it costs to leave an idea untested. A practical way to make “not now” a real decision.
In this story (6 sections)
General edition. , or use the switcher above.
When an organisation declines to build something, nothing visibly breaks. That is exactly why the cost is so easy to ignore.
This perspective argues that inaction deserves the same scrutiny as action, and proposes four small habits that make the cost of waiting visible without encouraging reckless building.
The short version
- Built things have owners, numbers and post-mortems. Unbuilt things don’t, so the cost of waiting is systematically under-counted.
- Decisions come in two sizes: hard to reverse and easy to reverse. Treating small, reversible bets with migration-grade process starves an organisation of learning.
- Put a price next to “not now”: what stays untested, what depends on it, what changes while you wait.
- An experiment is safe to approve when it has a question, a budget, a stop rule and an owner.
The expense nobody can see
Every budget meeting has a column for what a project will cost. Almost none has a column for what it costs to leave the project undone.
That asymmetry isn’t a conspiracy; it’s bookkeeping. A built thing has an owner, a number and a post-mortem if it fails. An unbuilt thing has none of those. Over time a sensible-looking habit forms: when in doubt, wait. And waiting never shows up as a loss.
“Doing nothing is also a decision. It just doesn’t get a meeting.”
Not every “yes” is the same size
In his 2016 letter to Amazon shareholders, Jeff Bezos separated decisions into two kinds. Some are consequential and nearly irreversible, “one-way doors”. Most are changeable and reversible, “two-way doors”. His observation was that as organisations grow they tend to apply the heavyweight process meant for the first kind to the second.
The practical consequence is easy to see. A three-week experiment with a capped budget is a two-way door. If approving it requires the same review as a platform migration, the organisation will run few experiments, and will mistake that scarcity for prudence.
| Kind of decision | Typical example | Sensible process |
|---|---|---|
| Hard to reverse | Replacing a core platform; a multi-year commitment | Slow, evidence-heavy, many reviewers |
| Easy to reverse | A time-boxed experiment with a fixed budget | Fast, one accountable owner, a written stop rule |
Put a price next to “not now”
The simplest intervention is one line in every proposal: what does it cost if we wait a quarter? The answer will often be small, and that is fine. The point is that it becomes an explicit estimate with an owner rather than an absence.
Good answers are specific. “We lose learning” is vague. “We keep shipping on an unverified assumption about checkout, and the next three features depend on it” is a cost someone can argue with.
- The decision or assumption that stays untested.
- What depends on it.
- What changes in the market, the team or the technology while you wait.
- The point at which waiting stops being cheap.
What makes an experiment safe to approve
Reviewers are right to be nervous about vague experiments. They tend to become projects with a friendlier name. A safe-to-approve experiment can be described in five lines:
- The question it will answer.
- The budget, in time and money, and the date it ends.
- The stop rule: what result ends it early.
- One accountable owner.
- What will be done with each possible outcome.
When not building is the right call
None of this argues for building more. Capacity is finite, maintenance is a real cost, and many ideas deserve a no. The claim is narrower: a no should be chosen with its cost in view, not inherited from a default.
What to change on Monday
- Add a “cost of waiting a quarter” line to every proposal, including the ones you expect to decline.
- Label each decision as one-way or two-way before choosing the review it gets.
- Give experiments a standing budget and a lightweight approval path, so approval does not depend on the roadmap’s mood.
- Keep a visible list of deferred ideas with their stated costs, and review it quarterly.
Key takeaways
- 01Treat “not now” as a decision with a written cost and an owner.
- 02Separate one-way doors from two-way doors before deciding how much process each deserves.
- 03Approve experiments by their question, budget, stop rule and owner, not by their ambition.
Sources
- 2016 Letter to Shareholders · Amazon · accessed 1 Oct 2026
- Site Reliability Engineering, “Embracing Risk” · Google · accessed 1 Oct 2026
Your perspective
A better read starts with knowing what matters to you.
Three quick questions. A publication that adapts to your perspective.
Keep reading
All stories →What happens before a good product decision
Framing, evidence, options, trade-offs and a pre-mortem: the unglamorous work that makes a product decision look obvious afterwards. Illustrated with a clearly hypothetical team.
Why promising AI pilots never become real products
A pilot proves something is possible. A product has to keep working. The gap between the two is predictable, and it can be closed before the money is spent.
What does it really take to bring AI into production?
Make four decisions about a hypothetical AI project (scope, data, oversight, rollout) and see what you’d be trading away. A simulation, clearly labelled.