A what-if analysis changes one or more inputs in a model to see the effect on the outcome. It is used to test decisions before they are made, to size risk, and to identify which single assumption the answer depends on most.
The plain definition
You have a model. It takes inputs and produces an outcome. A what-if analysis asks what the outcome becomes if you change an input, and it comes in three sizes. Change one input across a range and you are doing sensitivity analysis. Change a coherent set of inputs together, a good case and a bad case, and you are doing scenario analysis. Fix the outcome you need and solve backwards for the input that delivers it, and you are doing a goal seek.
The word to notice is model. A what-if analysis is only as honest as the thing it is run on. If the underlying model is a plug or a hardcoded number, the scenario is a picture of nothing.
What it is used for
The everyday uses in a growing company are unglamorous and high value: what happens to runway if the round closes three months late, what happens to gross margin if we lift price 8 percent and lose 5 percent of customers, what happens to cash if the two engineers start in March rather than January.
It is also how a finance function handles what it cannot control. Finance teams cannot control external factors like tariffs or the weather, but they can control the response: tightening reporting, stress-testing scenarios, and keeping the board aligned with what is really happening.[1] That is the honest framing. Scenario work does not reduce uncertainty. It converts uncertainty into a set of decisions you have already thought about.
The output format matters as much as the analysis. The clearest example I use outside of forecasting is equity: the best way to present an equity offer is a table of economic outcomes tied to the grant structure, showing a range of exit scenarios from zero to a billion-dollar exit.[2]
The techniques and tools
Most what-if analysis in Australian scale-ups still happens in a spreadsheet, and that is fine at this size. The three built-in Excel tools map directly onto the three sizes above.
| The technique | When to reach for it | |
|---|---|---|
| Sensitivity analysis | Excel data tables, one or two inputs across a range | When you want to know which single assumption the answer actually depends on. Usually there are only two. |
| Scenario analysis | Excel scenario manager, or a scenario switch in the model | When several inputs move together and have to stay internally consistent, such as a downside case where growth falls and churn rises. |
| Goal seek | Excel goal seek, solving backwards from the outcome | When the outcome is fixed, such as reaching breakeven by a date, and you need to know what has to be true to get there. |
Dedicated planning tools do this more cleanly at scale, and AI assistance is now genuinely useful in the setup. The limit is well understood: AI can comfortably handle basic checking, reconciliation and producing first-pass analysis, forecasts or models, but the final stage still needs a qualified human who understands the challenge and what a good outcome looks like.[3] A scenario deck that nobody has sanity-checked is a very fast way to be confidently wrong in front of a board.
A what-if analysis is only useful on top of a model somebody owns, which is the actual constraint in financial forecasting.
Who should be running it
Running the mechanics is not hard and plenty of people can do it. Choosing which scenarios are worth modelling, and being trusted when you present them, is the part that is scarce.
In early-stage CFO and head of finance roles the attributes founders consistently look for are direct ownership and operational experience alongside the strategic work, someone who can act as their commercial and strategic eyes and ears, with demonstrated credibility in front of banks, funders or investors.[4] That last clause is the scenario job in one line. A what-if analysis is an argument about the future, and an argument needs a credible person attached to it.
If scenarios are currently being run by whoever has time, the fix is structural, so start with how a startup finance team should be structured.
Common questions
What is a what-if analysis?
A what-if analysis changes one or more inputs in a model to see the effect on the outcome. It comes in three common forms: sensitivity analysis, which moves one input across a range; scenario analysis, which moves a consistent set of inputs together; and goal seek, which fixes the outcome and solves backwards for the input required.
What is the difference between sensitivity analysis and scenario analysis?
Sensitivity analysis isolates one variable to find out how much the answer depends on it, which tells you where to spend your research time. Scenario analysis moves a coherent group of variables together to describe a plausible state of the world, such as a downside where growth slows and churn rises at the same time. The first finds the fragile assumption, the second tests a story.
What tools are used for what-if analysis?
In most Australian scale-ups it is still a spreadsheet, and Excel has three purpose-built features: data tables for sensitivity, scenario manager for scenarios, and goal seek for solving backwards. Dedicated planning tools handle this more cleanly at scale. AI assistance is useful for producing first-pass analysis, but a qualified person still has to check it before it goes anywhere.
Who should run scenario analysis in a startup?
Mechanically, anyone competent in a spreadsheet. Practically, it should sit with whoever owns the forecast, because choosing which scenarios matter is the difficult part and it requires knowing the business rather than the model. Founders looking at first finance leaders consistently ask for operational ownership alongside strategic work and credibility in front of investors or funders, which is exactly what scenario work demands.
References
- What I tell finance teams in volatile industries: they cannot control external factors like tariffs or rain, but they can control their response by tightening reporting, stress-testing scenarios, and keeping boards aligned with what is really happening.
- How I describe it: the best format is a table of economic outcomes, closely linked to the founder's equity grant structure, showing a range of exit scenarios from zero to a billion-dollar exit.
- Tom Hunter on the limits of AI in finance, speaking on a podcast interview about finance careers and technology: AI can easily handle basic tasks like checking, reconciliation, or producing basic level analysis, forecasts or models, but you still need a qualified human in the loop for the final stages who understands the challenge and what a good outcome looks like.
- How this plays out by stage: direct ownership and operational experience alongside strategic work, someone who can act as their commercial and strategic eyes and ears, with demonstrated credibility with banks, funders or investors.