Your Scenario Model Changed the Decision. Which Assumptions Should You Stress-Test First?
Published August 2026 by Solarc Labs
Start with the assumptions the decision actually depends on
A scenario model can contain many editable inputs, but not every input deserves equal attention. Start with the variables that materially drive the decision: price, demand, conversion, unit cost, staffing, timing, adoption, churn or another assumption that changes the commercial choice you are considering. Write down the baseline value, why you chose it and what evidence supports it. If an input is merely a placeholder or management judgement, label it that way. The model becomes more useful when a decision-maker can see which values are observed facts, which are estimates and which are deliberate scenario choices.
Stress-test what is uncertain or susceptible to change
HM Treasury’s 2026 Green Book describes sensitivity analysis as testing how changes in key assumptions affect results and says practitioners should stress-test assumptions that are uncertain or susceptible to change. A small business model does not need to pretend it is a government appraisal to use the same discipline. For each important assumption, choose a defensible lower and upper case or another range that reflects the uncertainty you actually face. Change one input at a time first so you can see which variables genuinely move the outcome instead of hiding the effect inside a large bundle of optimistic or pessimistic changes.
Find the switching point, not only the best-looking scenario
The most useful question is often not “what happens in our upside case?” but “how far would this assumption have to move before we would choose differently?” The Green Book uses the idea of a switching value: the value at which a preferred option becomes less attractive than an alternative. For a commercial what-if model, the same idea can expose decision fragility. If a 2% change in demand reverses the choice, the decision is highly sensitive to demand evidence. If the decision survives a much wider range of plausible inputs, that is different information. Neither result is a prediction; both make the uncertainty visible.
Do not give every assumption a fake probability
A range is useful only when the team can explain what it represents. Do not assign precise probabilities merely to make a model look sophisticated. Where the evidence does not justify a probability distribution, show explicit cases or sensitivity ranges and keep the uncertainty visible. Current UK appraisal guidance similarly emphasizes identifying key uncertainties in data, assumptions and models, then testing the significant ones. The transferable lesson is transparency: record where the range came from and distinguish evidence-backed inputs from judgement calls.
Save the assumptions that supported the decision
If the model influences a real decision, keep the assumption set used at that moment rather than letting the next edit silently rewrite history. Record the baseline, alternative case, important sensitivity results, the switching point that mattered and the decision or next action. Later, compare the assumptions with what actually happened. That feedback can improve the next scenario and reveal which variables the team repeatedly estimates badly. A scenario calculator becomes a better decision tool when it preserves learning instead of producing a disposable number.
FutureCanvas is for transparent decision support, not prediction theatre
FutureCanvas is positioned as one bounded baseline-versus-alternative scenario workflow with visible assumptions, formulas and sensitivity logic. It should help a team inspect why a result changed and which assumption would change the decision again. It does not guarantee a forecast, assign certainty to unsupported inputs or promise that a scenario result will occur. The existing category comparison continues to answer the separate “scenario calculator vs forecast model” buying question; this article focuses on how to challenge the assumptions inside a scenario before relying on the decision.
Primary sources
Sources used for this article
Continue the job