Almost every business plan shares one weakness: it rests on a single number. A sales forecast, an exchange rate, a raw-material cost — you pick the "most likely" value, build the budget on top of it and present the result as if it were a certainty. Then the future does what it always does: it fails to match the plan.
The answer to this uncertainty is not to forecast better. It is to stop depending on a single forecast. Scenario analysis and sensitivity analysis exist precisely for that: instead of asking "what will happen?", they ask "what happens to our result if things play out in various ways?" — and turn that answer into a more robust decision.
This article explains what these two tools are, how they differ, how to build them with the data you already have and which traps to avoid so they do not become a decorative exercise.
What scenario analysis is
A scenario is an internally coherent story about how the future might unfold, translated into numbers. It is not a loose guess: it is a set of assumptions that support one another. If the scenario assumes a recession, then sales fall, but perhaps the cost of financing also changes and demand shifts towards cheaper products. The strength of a scenario lies in that internal coherence.
Analysing scenarios means building several of these possible futures and seeing what each does to the result that matters — margin, cash, the return on an investment. The goal is not to guess which one comes true, but to know the range of outcomes the company is exposed to and to prepare responses for each.
Scenarios are not just "optimistic, base and pessimistic"
The most common way to build scenarios is also the most limited: three columns — one optimistic, one central and one pessimistic — each adding or subtracting a percentage from everything. It is better than nothing, but it misleads, because it treats every variable as if they all moved together and in the same direction, which rarely happens.
Well-built scenarios start from distinct narratives, not percentages. "What if a competitor enters with aggressive prices?" is a scenario. "What if the raw material rises 30% but we manage to pass half of it on to the price?" is another. Each combines variables in a specific way, rather than turning one global dial. That specificity is what makes them useful for deciding.
Sensitivity analysis: what really moves the result
Before imagining futures, it is worth understanding what the result is sensitive to. Sensitivity analysis answers a surgical question: if I change only one variable, holding everything else equal, how much does the final result change?
The value of this exercise is revealing the two or three variables that dominate everything. Often you discover that the result barely reacts to things you spend huge energy debating, and is extremely sensitive to an assumption no one had questioned. Knowing where that lever is changes the conversation: it focuses attention — and data collection — where it pays off.
One-way, two-way and the tornado chart
There are three classic ways to look at sensitivity. One-way sensitivity changes one assumption at a time and records the effect. Two-way sensitivity crosses two variables in a table, useful when you suspect they interact — for example, price and volume.
The tornado chart sums it all up in one image: for each variable, it shows the swing in the result between its low and high value, ordering the bars from the most influential to the least. The name comes from the shape — wide bars at the top, narrow ones at the bottom. At a glance, you see which assumptions deserve care and which are noise. It is probably the tool with the best ratio of effort to clarity.
How to build useful scenarios, step by step
A simple process that avoids the most common mistakes:
- Start from the result that decides: first define the metric that matters — margin, cash flow, return — before touching the assumptions.
- List the assumptions and their sources: each number enters with an origin (history, contract, estimate) and a plausible range, not a single value.
- Do sensitivity first: identify the few variables that dominate the result and build scenarios around them, ignoring the irrelevant ones.
- Write the narrative of each scenario: a paragraph explaining the story before the spreadsheets. If you cannot tell the story, the scenario is not coherent.
- Attach a signal and a response: for each scenario, define the indicator that would show it is happening and what the company would do in that case.
From deterministic to probabilistic: a note on Monte Carlo
Scenarios and sensitivity are deterministic: they test a few hand-picked futures. When the variables are many and interact, you can take a step further with Monte Carlo simulation, which assigns each assumption a distribution of possible values and generates thousands of futures at once, producing not a single number but a distribution of results.
The advantage is answering questions like "what is the probability the project makes a loss?" instead of only "how bad is the worst case?". The downside is that it demands more care: a simulation fed by made-up distributions gives a false sense of rigour. For most decisions, well-thought-out scenarios and a good tornado chart are enough; Monte Carlo is the tool for when uncertainty is the centre of the decision.
Common mistakes that make scenarios useless
These tools almost always fail for the same reasons:
- Moving everything by the same percentage: adding 10% to all revenue and costs ignores that variables rarely move together.
- Scenarios with no narrative: three columns of numbers with no story behind them help decide nothing.
- Mistaking the base case for a promise: the central scenario is a working hypothesis, not a commitment to defend.
- Stopping at the analysis: a scenario is only worth it if it is tied to a decision or a response plan. Without that, it is a spreadsheet exercise.
Mini-case: an industrial company
A mid-sized industrial company prepared its annual budget as always: a single forecast, approved in committee and treated as a target. In a year of volatile raw materials, they decided to do it differently. They started with sensitivity analysis and found that the result depended overwhelmingly on two variables — the price of one raw material and the exchange rate — and was almost indifferent to things they argued about for hours.
They then built three scenarios with a narrative: one of stability, one of a cost shock with partial pass-through to price, and one of recession with falling demand. For each, they defined an early signal and a prepared response — from freezing hiring to triggering price-revision clauses with customers. When, halfway through the year, the raw-material price jumped, the team did not lose weeks debating what to do: it recognised the scenario, activated the plan already thought through and protected the margin. The gain was not predicting the rise — it was being ready for it.
In practice
Planning under uncertainty is not about guessing better; it is about no longer betting everything on a single forecast. Start by using a sensitivity analysis to identify the two or three variables that truly move your result. Build a handful of scenarios with a coherent narrative around them and, above all, attach to each one a signal that announces it and a response already thought through. You do not need expensive tools: a well-organised spreadsheet and a tornado chart take most companies further than any single forecast. The ultimate goal is not to be right about the future, but never to be caught off guard by it.
