The worst case was not bad enough
Buehler, Griffin, and Ross asked 37 psychology students to predict when they would submit their honours thesis. The average prediction was 33.9 days.
They also asked for two other numbers. If everything went as well as it possibly could: 27.4 days. If everything went as poorly as it possibly could: 48.6 days.
The average actual completion time was 55.5 days. Reality was worse than the worst case the students could imagine. Only around 30% finished within their own predicted time.
The follow-up work is more troubling than the headline. In later experiments, students were asked to recall their track record with similar assignments — in writing — and were explicitly told to keep it in mind while predicting. Their estimates came out just as optimistic. In think-aloud studies, roughly three quarters of estimating thoughts were forward-looking scenarios of the plan going smoothly. About 7% touched relevant past experience. Around 3% considered anything that might go wrong.
This is not confined to undergraduates. Flyvbjerg, Holm, and Buhl's analysis of 258 infrastructure projects across 20 nations found that 90% experienced cost overruns, averaging 28% for rail, 34% for bridges and tunnels, and 20% for roads.
Estimation does not improve with awareness, experience, or seniority. Which raises an obvious question: why is delivery planning built on it?
The inversion
Shape Up, published by Ryan Singer at Basecamp in 2019, starts from that question and reverses the usual relationship.
Conventional planning fixes scope and lets time vary. The feature set is agreed, the estimate is produced, and the date moves when the estimate proves wrong — which the research says it will.
Shape Up fixes time and lets scope vary. The period is decided first. What fits inside it is discovered during the work.
The consequence is that the estimate stops being load-bearing. Nobody needs an accurate prediction of how long something takes, because the time is a constraint rather than an output.
Appetite, not estimate
The core vocabulary change carries most of the method.
An estimate answers: how long will this take? It is a prediction, and the research above describes what happens to it.
An appetite answers: how much time is this worth? It is a decision, made before the work is shaped, based on the value of the outcome rather than the difficulty of the task.
Two appetites in the original: a small batch of a week or two, and a big batch of six weeks. The number matters less than the discipline — deciding what a problem is worth before discovering what it costs.
This changes the design conversation entirely. Given six weeks, what version of this solves the problem? That question produces a different and usually better answer than: what would the complete solution require?
The other components
Shaping. Before a cycle begins, senior people define the work at the right altitude — concrete enough to be actionable, abstract enough to leave room for judgement. Shaped work includes the appetite, the rough solution, and the identified rabbit holes. It is not a specification.
The betting table. Shaped work is bet on rather than assigned. Leadership decides which shaped pitches get a cycle. Nothing is owed a slot.
Six weeks, then two weeks of cooldown. Long enough to finish something meaningful, short enough that a bad bet costs six weeks rather than a quarter. Cooldown handles bugs, exploration, and the next round of shaping.
No backlog. The most contentious element, and the most defensible. Pitches that do not get bet on are not stored — anything genuinely important resurfaces. A backlog is a list of decisions nobody made, and maintaining it consumes real time.
The circuit breaker. If work does not ship within the cycle, it does not automatically get an extension. It stops, and must be re-bet to continue.
The circuit breaker is what makes it real
Every other element is a preference. This one is the mechanism.
Without it, fixed time is a suggestion. Work overruns, the cycle extends, and the method has quietly become conventional planning with different vocabulary.
With it, the deadline exerts actual force on scope. Teams cut, simplify, and make trade-offs during the work rather than requesting more time at the end — which is the behaviour the framework exists to produce.
It also converts a failure into information at a known cost. Six weeks spent and nothing shipped is a real signal about whether the problem was understood, and the loss is bounded. Compare that to a project that has consumed nine months and cannot be stopped because of what has already been spent.
The evidence
Documented practitioner method, with a narrow base. Shape Up describes how one company works, published by that company, free to read. There is no independent research comparing outcomes against alternative approaches.
Basecamp's context is unusual and worth naming: a small, profitable, independent software company with no external clients, no investor-imposed deadlines, and control over its own roadmap. Much of what makes the method possible flows from that.
The underlying argument, however, does not depend on Basecamp. The planning fallacy is among the better-replicated findings in behavioural research. Any method that reduces reliance on estimation is responding to something real. That is the reason to take Shape Up seriously — not because it works at Basecamp.
What transfers, and what does not
Transfers well: appetite as a concept, the circuit breaker, fixed cycle lengths, and abandoning the permanent backlog. None require Basecamp's business model.
Transfers with adaptation: cycle length. Six weeks suits Basecamp's rhythm. Agencies working to client milestones, or businesses with seasonal commitments, will need different periods and should not treat six as doctrine.
Does not transfer cleanly: the no-external-commitments premise. A business that has contractually promised a specific scope on a specific date cannot let scope vary, and pretending otherwise creates a different problem. In those cases the useful move is applying appetite thinking to internal work while handling committed work separately.
Worth being honest about: Shape Up assumes a level of trust between leadership and delivery teams that many organisations do not have. Betting rather than assigning requires leadership willing to be wrong in public every six weeks.
Where it goes wrong
Adopted as a calendar change. Six-week cycles introduced without shaping, without betting, and without a circuit breaker produce longer sprints and nothing else.
Shaping done by people without authority. Shaped work has to come with a real appetite attached. Shaping performed by people who cannot decide what something is worth produces specifications with a new name.
The circuit breaker is never used. The first time an overrunning project is extended, the method has ended. Everyone learns the deadline is negotiable, and behaves accordingly.
Appetite set from the estimate. Asking the team how long it will take and calling the answer the appetite reintroduces the original problem. Appetite is set from value, before shaping.
Applied to work that cannot be scope-varied. Compliance work, migrations, and contractual commitments have fixed scope by definition. Forcing them into the model produces a bet that cannot be lost, which is not a bet.
Diagnostic: is time actually fixed?
Six tests.
The appetite for current work was set before anyone estimated it.
At least one piece of work has been stopped at the end of a cycle rather than extended.
There is no permanent backlog being groomed.
Shaped work includes the rabbit holes that have been ruled out, not just the solution.
Scope was cut during a cycle, by the team, without escalation.
Leadership bets rather than assigns, and some pitches do not get bet on.
Test two is the only one that matters. Without it, the rest is vocabulary.
What this produces
A known and bounded cost of being wrong.
That is the honest description of what fixed-time working buys. It does not make teams faster, and it does not improve estimates — the research says nothing does. What it does is cap the exposure on any single decision at one cycle, which changes the risk profile of the entire portfolio of work.
For a business with limited runway, that cap is the valuable part. The alternative is discovering at month seven that a project needed twelve, having committed capital on the assumption of four.
The constraint on scope is also a design constraint, and design constraints reliably produce better decisions than open-ended ones. Given six weeks, teams build the version that solves the problem. Given as long as it takes, they build the version they imagined.
Frequently asked questions
What is the difference between an appetite and an estimate?
An estimate predicts how long work will take. An appetite decides how much time the outcome is worth, set before the work is shaped. One is a forecast that research shows is unreliable; the other is a decision.
Why are software estimates so consistently wrong?
The planning fallacy is well replicated. In the foundational study, students predicted 33.9 days, gave a worst case of 48.6, and actually took 55.5. Reminding people of past prediction failures does not improve subsequent estimates.
Does Shape Up require six-week cycles?
No. Six weeks suits Basecamp's context. The principle is a fixed period long enough to finish something meaningful and short enough that a bad bet is affordable. Agencies and seasonal businesses should set their own.
What is the circuit breaker?
Work that does not ship within its cycle stops rather than being extended, and must be re-bet to continue. It is the mechanism that makes fixed time real; without it, the deadline is a suggestion.
Is abandoning the backlog realistic?
It is more defensible than it sounds. A backlog is a list of decisions nobody made, and grooming it consumes real time. Genuinely important work resurfaces. What is lost is mostly a record of things the organisation declined to prioritise repeatedly.
Can this work in an agency or client-services business?
Partially. Where scope is contractually fixed, it cannot vary, and forcing the model creates problems. Appetite thinking, fixed cycles, and the circuit breaker still apply to internal and discretionary work.
Is there evidence that Shape Up produces better outcomes?
No independent research compares it against alternatives. It is one company's documented method. The argument for it rests on the planning fallacy being real, and on any method that reduces reliance on estimation responding to something genuine.
Sources
Singer, R., Shape Up: Stop Running in Circles and Ship Work that Matters, Basecamp (2019)
Buehler, R., Griffin, D. and Ross, M., Exploring the "Planning Fallacy": Why People Underestimate Their Task Completion Times, Journal of Personality and Social Psychology 67(3), 1994
Kahneman, D. and Tversky, A., Intuitive Prediction: Biases and Corrective Procedures (1979)
Flyvbjerg, B., Holm, M. and Buhl, S., analysis of cost overruns across 258 infrastructure projects in 20 nations (2002)
Structure your next phase
Zerologic works with leadership teams to scope work against what an outcome is worth, and to cap what being wrong costs.
Talk to us: partners@zerologic.io · zerologic.io
