K
KnowMBAAdvisory
FinanceIntermediate6 min read

Sensitivity Analysis

Sensitivity Analysis isolates the impact of changing ONE input variable (or two, in a two-way table) on a model output, while holding all other variables constant. It answers: 'For each 1% change in churn rate, how much does NPV / runway / valuation change?' This is fundamentally different from scenario planning โ€” scenarios change MULTIPLE correlated variables at once, sensitivity changes ONE at a time. The output is typically a tornado diagram (variables ranked by impact) or a sensitivity table (output values across a grid of input combinations). The PRACTICAL purpose is to identify the 2-3 variables that drive 80% of model variance โ€” those are the variables to manage, monitor, and stress-test in scenarios.

Also known asWhat-If AnalysisTornado DiagramOne-Way SensitivityTwo-Way Sensitivity TablesDriver Sensitivity

The Trap

The trap is confusing sensitivity analysis with scenario planning. A sensitivity table that shows 'if churn rises from 5% to 6%, valuation drops 15%' is NOT a Bear case โ€” Bear cases combine higher churn AND lower growth AND delayed funding simultaneously, because those variables correlate in real downturns. A second trap is sensitivity-testing variables that you can't actually influence (e.g., 'sensitivity to interest rates') while ignoring variables you can (sales rep ramp time, onboarding completion rate). Sensitivity is most valuable when applied to variables that the team can MOVE, not just observe.

What to Do

Build a tornado diagram for every major financial decision: list every input variable, flex each one ยฑ20% from base, plot the resulting change in the output (NPV, IRR, runway). The variables with the longest bars are your top sensitivities โ€” those become the focus of management attention. For two-variable interactions (e.g., growth rate ร— CAC payback), build a 2D sensitivity table. Refresh sensitivity analysis quarterly: variables that mattered last year may not matter now as the business changes. Always pair sensitivity output with the team's ability to influence that variable โ€” high-impact + high-controllability variables are where management leverage exists.

Formula

Sensitivity = (% Change in Output) / (% Change in Input); Tornado bar length = |Output(input + ฮ”) โˆ’ Output(input โˆ’ ฮ”)|

In Practice

When Datadog filed its S-1 in 2019, the prospectus included extensive disclosures around revenue sensitivity to net dollar retention (NDR) โ€” a 5-percentage-point change in NDR would shift forward revenue growth by ~12-15%. This wasn't theoretical: investors used sensitivity tables to model what Datadog would be worth at NDR of 110% vs. 130% vs. 150%. Datadog's actual NDR ran above 130% for years, validating the high end of the sensitivity. Compare to Casper Sleep's 2020 IPO: the prospectus had limited sensitivity disclosure, and the valuation collapsed when ad-spend efficiency (the unstated #1 sensitivity) deteriorated. Casper's IPO priced at $12 vs. an initial range of $17-19, then traded down 90% within 18 months.

Pro Tips

  • 01

    KnowMBA POV: every model should have a tornado chart at the top showing the 5 variables that matter most. If your CFO can't recite the top 3 sensitivities of next year's plan from memory, the company is being run on intuition, not analysis.

  • 02

    Sensitivity-test the variables your TEAM controls (sales rep ramp time, onboarding completion, feature adoption) more aggressively than the variables your team OBSERVES (Fed rates, FX, GDP). The first set is leverage; the second is weather.

  • 03

    Two-variable sensitivity tables (e.g., growth rate ร— discount rate) are the hidden gem of finance. They reveal interaction effects that one-way sensitivity misses โ€” for example, that a SaaS company's valuation is highly sensitive to growth WHEN discount rates are low, and barely sensitive WHEN discount rates are high.

Myth vs Reality

Myth

โ€œSensitivity analysis and scenario planning are the same thingโ€

Reality

They're complements, not substitutes. Sensitivity changes ONE variable to identify drivers; scenarios change a coherent SET of correlated variables to model plausible futures. You need both. Sensitivity tells you what to monitor; scenarios tell you what to plan for.

Myth

โ€œIf a variable has low sensitivity, you can ignore itโ€

Reality

Low sensitivity in normal ranges can mask catastrophic non-linearities. A loan covenant has zero sensitivity until you breach it, then total sensitivity. Always check for non-linear cliffs (debt covenants, churn thresholds, hiring step functions) before declaring a variable irrelevant.

Try it

Run the numbers.

Pressure-test the concept against your own knowledge โ€” answer the challenge or try the live scenario.

๐Ÿงช

Knowledge Check

Challenge coming soon for this concept.

Industry benchmarks

Is your number good?

Calibrate against real-world tiers. Use these ranges as targets โ€” not absolutes.

Top Sensitivity Variable in SaaS Valuations

Mid-stage to late-stage SaaS, public and pre-IPO

Net Dollar Retention

Typically #1 โ€” drives forward growth

Gross Margin

#2 โ€” compounds across all future cash flows

S&M Efficiency / CAC Payback

#3 โ€” controls growth investment

Discount Rate / WACC

Market-determined; high impact but low controllability

Source: Bessemer Cloud Index, OpenView SaaS Benchmark Reports

Real-world cases

Companies that lived this.

Verified narratives with the numbers that prove (or break) the concept.

๐Ÿถ

Datadog

2019 S-1 filing

success

Datadog's IPO prospectus disclosed extensive sensitivity to Net Dollar Retention. Investors and analysts modeled NDR scenarios from 110% to 145%. Sensitivity tables made clear that even a 5pp shift in NDR moved forward revenue projections by 12-15% per year โ€” compounding over 5 years. Datadog's actual NDR ran above 130% from 2019-2023, validating the high end of the sensitivity. The clarity of the sensitivity disclosure helped Datadog price its IPO at the high end of the range and trade up sustainably.

Disclosed NDR (S-1)

>130%

Forward Revenue Sensitivity

~12-15% per 5pp NDR

IPO Pricing

Above range

Stock Performance Year 1

+85%

Disclosing your top sensitivity (and the value of improving it) educates investors and earns valuation premium. Hiding sensitivities forces investors to assume worst-case behavior.

Source โ†—
๐Ÿ›๏ธ

Casper Sleep

2020 IPO

failure

Casper Sleep's S-1 disclosed limited sensitivity to its #1 variable: customer acquisition cost via paid digital advertising. The model relied on continued favorable Facebook/Google ad economics, but didn't sensitivity-test what would happen if CAC rose 30%, 50%, or 100% (which it did across DTC during 2020-2022). The IPO priced at $12 vs. $17-19 initial range. Within 18 months, Casper traded below $1.50 and was taken private at a fraction of IPO valuation. The unmodeled CAC sensitivity was the central failure.

Initial IPO Range

$17-19

Final IPO Price

$12

Stock Low (within 18 months)

<$1.50

Outcome

Taken private at distressed valuation

If your business model has a single dominant sensitivity (DTC = CAC), you MUST sensitivity-test it aggressively. Hiding the sensitivity doesn't make it go away โ€” it makes the eventual reckoning worse.

Source โ†—

Related concepts

Keep connecting.

The concepts that orbit this one โ€” each one sharpens the others.

Beyond the concept

Turn Sensitivity Analysis into a live operating decision.

Use this concept as the framing layer, then move into a diagnostic if it maps directly to a current bottleneck.

Typical response time: 24h ยท No retainer required

Turn Sensitivity Analysis into a live operating decision.

Use Sensitivity Analysis as the framing layer, then move into diagnostics or advisory if this maps directly to a current business bottleneck.