What is the difference between a sensitivity and a scenario?

A sensitivity changes one variable and holds everything else fixed. A scenario changes many variables together in a way that tells a coherent story about a different state of the world. The distinction matters because variables are linked in reality. When demand softens, revenue does not politely fall while pricing, collections, customer behaviour and cost flexibility all stay exactly on plan. Discounting rises, customers stretch payment terms, bad debt creeps up, and fixed costs prove sticky. A sensitivity that holds all of that constant is describing a world that cannot exist.

Why is a revenue down 10% case not really a scenario?

Because it flexes one number and freezes the reality around it. Turning a single knob three times produces three numbers, not three scenarios, and calling them scenarios creates the impression the business has stress-tested its plan when it has tested one assumption. In the worked example, a 15 percent volume decline modelled as a sensitivity reports 825,000 dollars of operating profit. The same event modelled with its linked drivers, deeper discounting, bad debt rising from 1 to 3 percent, and fixed costs that barely move, lands at 556,000 dollars, roughly a third lower.

Should the upside and downside cases be symmetric?

Rarely. Reality is not symmetric, because the downside usually carries compounding effects that the upside does not mirror. Softer demand brings discounting, slower collection and sticky costs all at once, while a comparable upside does not produce equivalent reinforcing gains. A tidy plus or minus 10 percent pair looks balanced on a slide and systematically understates how much worse the down case really is. Build the downside from what would actually happen, not from the mirror image of the upside.

How many scenarios should we build?

Two or three named worlds, each tied to a decision. A dozen scenarios is not rigour, it is paralysis, and a wall of variants nobody acts on is worse than a small set that genuinely pressure-tests a choice. Each scenario should be stated as a situation rather than an input, such as demand softens through the second half rather than revenue down 15 percent, and every linked driver inside it should move together consistently. If a scenario changes no decision, it has not earned its place in the pack.

What are the limits of scenario planning?

Scenarios rest on judgments about how drivers correlate, and those correlations are estimated rather than known. A scenario is not a probability forecast, and it is easy to overfit a vivid narrative and then treat an invented story as fact. Sensitivities remain genuinely useful as diagnostics, since they tell you which single input the answer is most fragile to and therefore where estimation effort belongs, but they are not a picture of a different future. The goal is not prediction. It is to replace the false comfort of a one-variable flex with a small set of coherent worlds you could actually face.

Scenarios Are Not Sensitivities

Core Argument

A planning team presents three scenarios to the board: base case, revenue up 10%, revenue down 10%. Everyone calls it scenario planning. It is not. It is a sensitivity on a single input, dressed in the language of scenarios, and the difference is why plans that look robust on the slide fall apart the moment the world actually moves. The world never moves one variable at a time.

The core insight

A sensitivity changes one variable and holds everything else fixed. A scenario changes many variables together, in a way that tells a coherent story about a different state of the world.

That distinction is everything, because in reality variables are linked. When demand softens, revenue does not politely drop 10% while your costs, your pricing, your collections and your customer behaviour all stay exactly on plan. Discounting rises as the sales team chases a thinner pipeline. Customers stretch their payment terms and bad debt creeps up. Fixed costs prove sticky because you cannot shed them as fast as revenue falls. A revenue down 10% sensitivity that holds all of those constant is describing a world that cannot exist. It flexes one number and freezes the reality around it.

A scenario refuses that fiction. It asks: if this different world came to pass, what would move, and how would those movements interact? The answer is almost always worse, or at least different, than the single-variable flex suggests, because the linked effects compound.

"A sensitivity asks what if this one number is wrong. A scenario asks what if the world is different."

The framework: the coherent scenario

A real scenario is built as a story, not a slider. Four steps.

1

Name the world.
State the scenario as a situation, not an input. "Demand softens through the second half" is a world. "Revenue down 15%" is a knob.
2

Map the linked drivers.
Identify what moves with what. In a downturn, revenue, discounting, bad debt and cost flexibility do not move independently. Write down the chain: softer demand leads to more discounting, which lowers effective price, while slower payment raises bad debt, while fixed costs resist cutting.
3

Quantify consistently.
Move every linked driver together, in the same direction the story implies. Not one input down and the rest on plan, but the whole coherent set.
4

Read the decision under each world.
A scenario exists to inform a choice: how much cash to hold, whether to commit to the expansion, when to act. If it does not change or pressure-test a decision, it is an exercise, not a plan.

The same event, modelled both ways

Driver	Base plan	Naive sensitivity	Coherent scenario
Units	100,000	85,000	85,000
Effective price	$50	$50	$48
Revenue	$5,000,000	$4,250,000	$4,080,000
Bad debt	1%	1%	3%
Fixed costs	$1,800,000	$1,800,000	$1,780,000
Operating profit	$1,200,000	$825,000	$556,000

The scenario column carries the linked effects the sensitivity holds constant: deeper discounting on price, customers stretching payment into bad debt, and fixed costs that will not fall as fast as revenue.

Both start from the same event: units fall 15%. The naive sensitivity flexes only volume and reports profit of $825,000, a manageable dip. The coherent scenario lets the linked drivers move as they actually would and profit lands at $556,000. That is a third lower than the sensitivity suggested. A company that planned to the $825,000 figure walked into the downturn believing it had far more cushion than it did. The gap between the two columns is not a modelling detail. It is the margin of error in the board's confidence.

Why experienced managers get this wrong

1

They label single-variable flexes as scenarios.
Turning one knob three times produces three numbers, not three scenarios. Calling them scenarios creates the impression the business has stress-tested its plan when it has only tested one assumption.
2

They move correlated inputs independently.
Even when several inputs are flexed, they are often flexed one at a time or in unlinked combinations, producing worlds that are internally inconsistent: revenue collapsing while collections and discounting stay pristine. The model runs, but the world it describes is impossible.
3

They build symmetric up and down cases.
Reality is rarely symmetric. The downside usually carries compounding effects, discounting, bad debt, sticky costs, that the upside does not mirror. A tidy plus or minus 10% pair understates how much worse the down case really is.
4

They build too many scenarios, or ones tied to no decision.
A dozen scenarios is not rigour, it is paralysis. Two or three coherent, decision-linked worlds beat a wall of variants nobody acts on. If a scenario changes no decision, it is not earning its place.

Putting it to work tomorrow

Build 2 or 3 named worlds, not a dozen inputs. A realistic downside, a realistic upside, and the base. Each stated as a situation, each with its linked drivers moving together consistently.
Make each scenario internally coherent. Before you trust a scenario, check that everything in it could be true at the same time. If revenue is down hard but discounting, collections and costs are all still on plan, the scenario is broken.
Keep sensitivities, but use them for what they are. A sensitivity is a diagnostic, not a plan. It tells you which single input the answer is most fragile to, which is genuinely useful for knowing where to focus your estimation effort. Just do not mistake that diagnostic for a picture of a different future.

Know the limits

Scenarios rest on judgments about how drivers correlate, and those correlations are estimated, not known. A scenario is not a probability forecast, and it is easy to overfit a vivid narrative and treat a made-up story as fact. The goal is not prediction. It is to replace the false comfort of a one-variable flex with a small set of coherent worlds you could actually face, so the decision holds up in more than one of them.

The principle worth keeping

You need both a sensitivity and a scenario, but only one of them is a plan.

"When the world moves, it moves every number at once, and a plan built on flexing them one at a time was never really tested."

Free Excel Model: Forecast Scenario & Sensitivity

A driver-based forecast with 3 named scenarios and a sensitivity ranking, built so the linked drivers move together rather than one at a time.

Download Free ↓