Skip to content

Leadership

Decide when to stop a failing project

Assess a struggling project from today: separate past effort from remaining costs, test the central assumption and make an explicit next decision.

Illustration of a woman setting aside a paper model among other prototypes on a workbench.

A team has spent six months building an internal dashboard, but two pilots have produced little sustained use. The next proposal is another round of features. Before approving it, the manager needs to answer a harder question: what would make the next investment worthwhile from today?

A disappointing project is not automatically a project to cancel. Some worthwhile work needs time to mature. The decision should turn on remaining prospects, risks and alternatives, with a clear explanation of what new evidence would justify continuing.

Separate past effort from the next commitment

Start by listing what cannot be recovered: completed development, previous meetings and effort already spent. Acknowledge those investments without making them a reason to spend more. Then list costs that still depend on the decision, including further work, support, cancellation obligations and the consequences of withdrawing a service.

A 2025 Journal of Finance study by Marius Guenzel used variation in acquisition costs associated with market movements to examine later commitment to acquired businesses. It found evidence consistent with sunk costs influencing subsequent investment decisions. The corporate acquisition setting differs from a team project, but the distinction matters: a cost already incurred is different from a cost you can still avoid.

Time can create a similar pull. A 2018 Science study found sensitivity to time already invested in laboratory tasks involving humans and other species. Those tasks do not provide a project-cancellation rule. They are a reminder to examine whether the argument for continuing is about future value or reluctance to abandon past effort.

Identify what has actually failed

Return to the project’s purpose and the evidence available when it was approved. Was the dashboard meant to reduce repeated manual reporting, improve decisions or satisfy a requirement? A project can deliver its planned features while failing to solve the problem it was intended to address.

Distinguish an invalid assumption from an execution problem. If people want the output but cannot access it reliably, another bounded test may be useful. If the intended users say the dashboard answers a question they do not need answered, more features may not address the underlying issue.

Ask someone close to the affected work to explain the evidence, including contradictory results. Avoid using enthusiasm from a single sponsor as a substitute for actual use. Record uncertainty plainly: low adoption could reflect poor fit, poor introduction or an observation period that was too short.

Compare credible options from today

Consider stopping, reducing scope, pausing for a dependency or running a limited test. For each option, describe the remaining effort, expected benefit, main risk and what the team would be unable to do instead. Include transition work; stopping is rarely costless.

For the dashboard, a useful test might focus on one reporting decision with a named user group and a defined review date. Agree in advance what evidence would justify more work and what would end the test. Choose criteria that reflect the original purpose, rather than a convenient activity count.

Keep the next commitment small enough to review honestly. A pilot that quietly becomes another full implementation postpones the decision. If no feasible test can change your understanding, explain why continuing would still be justified before asking for further capacity.

Make it possible to change course without humiliation

Experiments reported by Charles Dorison and colleagues in 2021 found that people could see persistence with a failing course as a signal of trustworthiness. That does not mean persistence is always wise, or that every manager faces the same reputational pressure. It helps explain why a technically sound stop recommendation may still feel personally risky.

Recognize the team’s effort and the information it produced. Separate the decision about future work from a judgment about someone’s worth. Ask the sponsor to review the evidence alongside someone who did not make the original commitment.

Close with a decision and responsibilities

Record the choice, the evidence behind it and any review condition. If stopping, assign responsibility for notifying affected people, preserving useful work and retiring access or support arrangements appropriately. If continuing, name the next checkpoint. The useful outcome is a decision that can be explained from current evidence, rather than another installment of hope.

More to read

Browse insights