A Go/No-Go decision is a commitment of money, people and years. It is also a decision made under uncertainty, by people who have worked on the project and want it to succeed. That combination is where good teams still make avoidable mistakes. The habits below do not remove the uncertainty, but they make the reasoning easier to inspect.
Agree the criteria before the data arrives
Write down what a successful result looks like before the experiment is run: the target potency, the exposure needed, the safety margin, the size of effect that would matter to a patient. When thresholds are set after seeing the data, it is very easy to find a reason why a marginal result is good enough. A target product profile gives these criteria a home and keeps them tied to the clinical need rather than to whatever the compound happened to achieve.
Grade the evidence, not just the headline
Two results can both read "positive" and carry very different weight. Ask how many independent experiments support the claim, whether it has been reproduced by someone other than the original team, how relevant the model is to the human disease, and whether the effect is large compared with the variability. A single striking result from one model deserves less weight than a modest result that replicates across systems.
Separate the question of the science from the question of the project
"Is the biology right?" and "Can we make a medicine from this compound series?" are different questions. A project can fail because the target hypothesis was wrong, or because the chemistry cannot deliver the right properties. Treating them separately keeps a chemistry problem from killing a sound target, and keeps a weak target from being rescued by yet another round of optimisation.
Write kill criteria, and treat them seriously
For every stage, define the results that would stop the project. Then, at the decision point, ask whether any of them has been met. If the team wants to proceed anyway, make them state what has changed in the evidence, not just in the hope.
Watch for the usual biases
- Sunk cost: the money and time already spent are not a reason to spend more. Only the expected value of the next step is.
- Optimism and confirmation: we search harder for reasons a result is an outlier when it disappoints us than when it pleases us.
- Authority: the most senior person in the room should often speak last.
- Single-point stories: a compelling narrative built on one data point is not a plan.
Run a pre-mortem
Before committing, ask the team to imagine it is two years from now and the project has failed. Have everyone write down why. The reasons that surface are often the risks nobody wanted to raise in the meeting, and they point directly at the experiments to run next.
Record the decision and the reasoning
Keep a short decision record: the options considered, the evidence used, the criteria applied, the dissenting views, and what would change the decision. When results arrive later, you can see whether the call was good reasoning that met bad luck, or poor reasoning that happened to work. That is how a team gets better at decisions over time.
A short checklist for the next decision meeting
- Were the success criteria written before the data came in?
- How strong is the evidence, and has it been independently reproduced?
- Is this a target problem or a molecule problem?
- Has any kill criterion been met, and if so, what is the case for continuing?
- What would a failure look like, and what would we do about it today?
- Is the decision, and the reasoning behind it, written down?