Vary one assumption for explanation, several for a decision
A one-way change can show which input drives the result. A two-way table is useful when two uncertain assumptions interact, such as promotion volume and the expected no-promotion baseline. Choose ranges from the commercial evidence rather than making every slider move by the same arbitrary percentage.
Price and quantity are independently varied in the workbench. Raising a price slider does not automatically predict lower demand. Enter a coherent pair from a demand scenario, and document whether the two assumptions should move together.
Worked case: a positive center hides a negative range
At regular price €50, promotion price €40, variable cost €25 and €300 spending, 1,800 promotion units contribute €26,700. A 1,000-unit baseline contributes €25,000, leaving +€1,700. Change baseline or promotion volume and the conclusion can reverse.
| Baseline units | 1,620 promotion units | 1,800 promotion units | 1,980 promotion units |
|---|---|---|---|
| 800 | +€4,000 | +€6,700 | +€9,400 |
| 1000 | −€1,000 | +€1,700 | +€4,400 |
| 1200 | −€6,000 | −€3,300 | −€600 |
Turn the table into an evidence request
At 1,800 promotion units, the break-even no-promotion baseline is 1,068 units: €26,700 / €25 regular unit contribution. If your plausible baseline range is 800–1,200, the decision depends on whether a baseline above 1,068 is credible.
That is a concrete next question for a forecast review or experiment. It is more useful than reporting a positive central case while omitting the nearby negative cases. An agreed downside limit can also help choose a smaller, more informative campaign rather than an all-or-nothing rollout.
Map the sensitivity worksheet
The baseline-sensitivity template supplies the central case. The workbench saves scenario multipliers and report context in project backups, so the reviewed assumptions can be reopened. Preserve the original data alongside any scenario rather than overwriting it with an optimistic result.
| Assumption | Workbench field or control | Central case |
|---|---|---|
| No-promotion volume | baseline_units / baseline multiplier | 1000 / 100% |
| Promotion response | promo_units / volume multiplier | 1800 / 100% |
| Promotion price | promo_price / price multiplier | 40 / 100% |
| Campaign cost | promo_spend / spending multiplier | 300 / 100% |
Do not call a scenario range a confidence interval
A manually selected range has no coverage probability merely because its endpoints are called conservative and optimistic. A statistical interval needs a design or model and the relevant dependence assumptions. Likewise, assigning equal probability to every cell is a modeling choice, not a property of a sensitivity table.
The table also omits effects not represented by its inputs. If larger volume changes returns, funding eligibility or future displacement, vary those assumptions coherently instead of holding them fixed by convenience.
Frequently asked questions
Which assumption should I investigate first? Focus on uncertain inputs that can change the decision near the operating range. A very sensitive but already well-measured cost may deserve less attention than an uncertain baseline.
Can sensitivity analysis choose the optimal campaign? It compares supplied scenarios. Optimization requires an objective, constraints and a defensible response model; more slider positions do not supply those missing elements.
Put it to work
Source context
Guide text and examples were prepared independently for MarginTrial. Sources provide the stated methodological or product context. They do not validate the hypothetical commercial outcomes.