Sensitivity analysis asks: when we change one input, how much does the output change? It identifies which assumptions drive the result. If a 10% drop in conversion rate flips your unit economics from positive to negative, conversion is a critical variable — you are sensitive to it. If a 20% change in cloud cost barely moves the bottom line, that input is less critical. The method does not require probabilities; it only requires varying inputs one at a time (or in combination) and observing the effect. The output is a ranking: which inputs matter most?
Engineers and financiers use sensitivity analysis routinely. A DCF model has many inputs: growth rate, discount rate, terminal multiple. Sensitivity analysis shows how valuation changes when each input moves. The same logic applies to product decisions: if retention is the lever that moves LTV the most, focus there. If pricing has little effect on payback, it is less sensitive. The goal is to spend effort and attention on the variables that move the needle and to avoid over-precision on variables that don't.
Sensitivity analysis also supports robustness. If the conclusion (e.g. "this is a good investment") holds across a wide range of the key input, the decision is robust. If it reverses with a small change in one assumption, the decision is fragile. The discipline is to run the analysis before committing: which assumptions would have to be wrong for us to regret this? Those are the assumptions to stress-test and monitor.
Section 2
How to See It
Sensitivity analysis appears when decision-makers vary inputs and track output change. Look for "what if X is 10% lower?" or tornado charts that rank variables by impact. The absence of it is a signal: a model or plan with many assumptions but no test of which ones drive the result is under-defended.
Business
You're seeing Sensitivity Analysis when a CFO presents a budget model and a table showing how EBITDA changes when revenue is ±5%, margin is ±2%, or opex is ±10%. The variables that move EBITDA the most get flagged for monitoring. The board knows which assumptions to watch.
Technology
You're seeing Sensitivity Analysis when a product team models LTV and varies retention, ARPU, and payback period. Retention drives 70% of the variance in LTV; the team prioritises retention over marginal ARPU gains. The analysis tells them where to focus.
Investing
You're seeing Sensitivity Analysis when an investor builds a DCF and runs scenarios: growth 15% vs 25%, multiple 5× vs 8×. The valuation range is wide; the key insight is that the outcome is most sensitive to the terminal multiple. The investor then focuses on what could justify that multiple.
Markets
You're seeing Sensitivity Analysis when a policy model shows that the fiscal impact of a reform is highly sensitive to labour supply elasticity. The debate shifts to that parameter: what do we assume, and what does evidence say? Sensitivity analysis surfaces the critical assumption.
Section 3
How to Use It
Decision filter
"Before locking a decision that depends on a model or plan, run sensitivity analysis. Vary each key input (or combinations) and see how the output changes. Rank inputs by impact. Stress the ones that matter most; treat the rest as secondary. If the conclusion flips on a small change in one input, that input is critical — get more evidence or add margin."
As a founder
Use sensitivity analysis on unit economics, runway, and roadmap. Which lever — conversion, retention, ARPU, CAC — moves the result the most? Allocate experiments and effort to the sensitive variables. When raising or planning, show investors and the board which assumptions drive the outcome and how wide the range is. If one variable dominates, say so and explain how you're de-risking it.
As an investor
Require sensitivity analysis in memos. When a founder presents a model, ask: which inputs are you most sensitive to? Run the analysis if they haven't. Decisions that are robust across a range of key inputs are lower risk; decisions that hinge on one assumption need that assumption validated or the position sized accordingly.
As a decision-maker
Before approving a major initiative, ask for a sensitivity view. What would have to go wrong for this to fail? Which assumptions drive the payoff? Allocate monitoring and contingency to the sensitive inputs. Avoid committing when the case is fragile — when a small error in one assumption reverses the conclusion.
Common misapplication: Varying too many inputs at once. Sensitivity analysis is clearest when you vary one (or a small set) at a time. Multivariate variation is Monte Carlo territory; sensitivity is about isolating which variable matters.
Second misapplication: Ignoring correlation. If two inputs move together (e.g. revenue and margin in a downturn), varying them independently understates joint risk. For critical decisions, run both one-way and correlated sensitivity (e.g. "recession" scenario where revenue and margin both fall).
Buffett stresses margin of safety and the few key variables that determine an investment. In practice that is sensitivity thinking: which assumptions drive the outcome? He focuses on businesses where the range of outcomes is narrow or where the downside is bounded. Sensitivity analysis would show that his preferred investments are robust to variation in key inputs.
Hastings has described using data to identify which levers drive subscriber and engagement outcomes. Netflix runs experiments that effectively vary one input (e.g. content, algorithm, price) and measure the effect. That is sensitivity analysis in product and strategy — find what moves the needle, double down there.
Section 6
Visual Explanation
Sensitivity analysis: vary each input and measure change in output. Rank by impact. The inputs that move the output the most are critical; stress-test them and monitor.
Section 7
Connected Models
Sensitivity analysis connects to scenario analysis (multiple futures), Monte Carlo (full distribution of outcomes), and the idea of critical assumptions and margin of safety. These models either extend sensitivity or use its output.
Reinforces
Scenario Analysis
Scenario analysis defines a few discrete futures (base, bull, bear). Sensitivity analysis identifies which inputs, when varied, would move you from one scenario to another. The two reinforce: scenarios frame the set of futures; sensitivity shows which levers define the boundaries.
Reinforces
Critical Assumptions
Critical assumptions are the ones that must hold for the plan to work. Sensitivity analysis is how you find them: the inputs that move the output the most are the critical assumptions. List them, stress them, and monitor them.
Tension
Monte Carlo Simulation
Monte Carlo varies many inputs randomly and aggregates outcomes. Sensitivity varies one (or few) at a time. The tension: one-way sensitivity can miss interactions and correlation. Monte Carlo captures them but is heavier. Use sensitivity to find key drivers; use Monte Carlo when you need the full distribution.
Tension
Confounding Factor
Confounding factors are variables that influence both an input and the output, creating spurious sensitivity. The tension: sensitivity analysis can highlight a variable that is actually a proxy for something else. Validate that the input you're varying is causal, not just correlated.
Section 8
One Key Quote
"The key is to figure out which assumptions are critical and which are not, and to focus on getting the critical ones right."
— Howard Marks
Sensitivity analysis is the method for finding those critical assumptions. Vary each input; the ones that move the output the most are critical. Spend your limited attention and research there. Leave the rest at a reasonable default.
Section 9
Analyst's Take
Faster Than Normal — Editorial View
Run it before you commit. Every material decision that rests on a model or plan should have a sensitivity pass. Which inputs drive the result? If you don't know, you're flying blind. A one-hour sensitivity table can change the whole conversation.
Present the tornado. A tornado chart or table that ranks variables by impact is worth a thousand assumptions. It tells the board or investor exactly what to watch. "We're most sensitive to retention; here's our plan to de-risk it."
Fragility is a red flag. If a small change in one assumption flips the conclusion, the decision is fragile. Either get better data on that assumption, add margin, or pass. Don't commit when the case is one assumption away from breaking.
Combine with scenarios when inputs correlate. When revenue and margin move together (e.g. in a recession), one-way sensitivity understates risk. Run a "downside" scenario where the sensitive inputs move together. That is where fragility hides.
Section 10
Test Yourself
Is this mental model at work here?
Scenario 1
A team builds a unit-economics model and finds that a 1% change in retention moves LTV by 8%, while a 1% change in ARPU moves LTV by 2%. They prioritise retention initiatives over ARPU experiments.
Scenario 2
An investor asks a founder: 'If your conversion rate is 10% lower than plan, does the round still work?' The founder runs the numbers and says no — the runway falls short. The investor flags conversion as a critical assumption.
Scenario 3
A CFO presents a single budget with one set of assumptions. No table or chart shows how the result changes when key inputs vary.
Scenario 4
A product memo says: 'We're most sensitive to activation; if we hit 60% Day-7 activation instead of 50%, LTV goes up 40%. We're running three experiments on activation this quarter.'
Section 11
Further Reading
Sensitivity analysis is standard in finance and engineering. These sources cover the method and how to present it.
Damodaran's valuation texts include extensive treatment of sensitivity analysis in DCF models. How to vary growth, margin, and cost of capital and present the range.
Bank stress tests are sensitivity analysis at the tail: key inputs (rates, unemployment, defaults) are pushed to extreme values. The framework applies to any decision with key drivers.
Klarman on margin of safety in investing. Sensitivity analysis tells you how wrong your assumptions can be before the thesis breaks — and thus how much margin to demand.
Summary: Sensitivity analysis varies key inputs and measures how much the output changes. It ranks assumptions by impact and identifies critical variables. Use it before committing: stress the inputs that matter most, add margin where the conclusion is fragile, and focus effort on the levers that move the result.
Leads-to
Stress Testing
Stress testing is sensitivity analysis at the extremes: what happens when we push key inputs to worst plausible values? Sensitivity identifies which inputs to stress; stress testing applies the push. Banks stress capital; companies can stress cash, demand, or cost.
Leads-to
Margin of Safety
Margin of safety is the buffer so that the decision still works when inputs are worse than expected. Sensitivity analysis shows how much worse each input can be before the conclusion flips. That defines how much margin you need. The two are paired: sensitivity informs the required margin.
