What Sensitivity Means

Sensitivity describes responsiveness. If an input changes slightly, sensitivity tells us how strongly the output responds. A model with high sensitivity reacts strongly to small changes in a variable. A model with low sensitivity changes only slightly. In mathematics, this idea is closely connected to derivatives, slopes, marginal change, elasticity and local approximation.

For a function \(f(x)\), the simplest sensitivity measure is the derivative \(df/dx\). If \(df/dx=10\), then near that point, a small increase of 1 unit in \(x\) is associated with an increase of about 10 units in \(f(x)\). This is absolute sensitivity because it keeps the original units of output per unit of input.

Relative sensitivity adjusts for scale. A 10-unit change can be large or small depending on the size of the output. If the output is 20, a 10-unit change is large. If the output is 10,000, a 10-unit change is small. Relative sensitivity compares percentage change in output with percentage change in input so that different variables and models can be compared more fairly.

The sensitivity formula is useful in economics, engineering, data modeling, finance, science and optimization because decisions often depend on which input matters most. If a result is highly sensitive to one assumption, that assumption deserves careful measurement and monitoring. If the result is barely sensitive to another input, that input may be less important for decision-making.

Main Sensitivity Formulas

The three most common sensitivity formulas are absolute sensitivity, relative sensitivity and finite-change sensitivity. They answer related but different questions.

Absolute sensitivity is the derivative of the output with respect to the input. It gives the local rate of change in output units per input unit.

Relative sensitivity compares the fractional change in the output with the fractional change in the input. It is dimensionless when the fractions are computed consistently.

For very small changes, the finite-change relative sensitivity approaches the derivative-based formula. This form is also called elasticity in many economics and applied mathematics contexts.

Finite-change sensitivity uses observed differences rather than derivatives. It is useful when you have before-and-after data instead of an explicit function.

Absolute Sensitivity

Absolute sensitivity measures the direct change in output for a one-unit change in input. If the function is \(f(x)\), then absolute sensitivity at a point is the derivative \(df/dx\). In a graph, it is the slope of the tangent line at that point.

Suppose \(f(x)=3x^2\). The derivative is \(df/dx=6x\). At \(x=5\), absolute sensitivity is \(30\). This means that near \(x=5\), a small increase of 1 unit in \(x\) produces an approximate increase of 30 units in \(f(x)\). For a smaller input change, such as \(\Delta x=0.1\), the approximate output change is \(30(0.1)=3\).

Absolute sensitivity is useful when the units are meaningful. In a cost model, if \(dC/dx=20\), the interpretation might be 20 dollars per additional unit. In a physical model, it might be volts per degree, meters per second per kilogram, or another unit ratio. The derivative keeps the units, which can make it practical and concrete.

The limitation is that absolute sensitivity can be hard to compare across models with different scales. A derivative of 20 may be large in one model and small in another. That is why relative sensitivity is often used when comparing the importance of different variables.

Relative Sensitivity and Elasticity

Relative sensitivity measures percentage response. It asks: if the input changes by 1%, approximately what percentage change should we expect in the output? This is why the formula is dimensionless and useful across different scales.

If \(S_{\text{rel}}=0.5\), then a 1% increase in \(x\) produces about a 0.5% increase in \(f(x)\), for small changes near the point. If \(S_{\text{rel}}=2\), then a 1% increase in \(x\) produces about a 2% increase in \(f(x)\). If \(S_{\text{rel}}=-1.2\), then a 1% increase in \(x\) produces about a 1.2% decrease in the output.

In economics, relative sensitivity is often called elasticity. Price elasticity of demand, for example, measures the percentage change in quantity demanded divided by the percentage change in price. In business models, the same idea can be used to analyze how revenue, cost, profit or margin responds to changes in price, quantity, labor cost or material cost.

Relative sensitivity works best when \(x\) and \(f(x)\) are nonzero. If the input or output is zero, the percentage-change interpretation can break down. In those cases, absolute sensitivity, finite differences or a different normalization may be more appropriate.

Finite-Change Sensitivity

Not every problem gives a formula. Sometimes you only have observations: an input before and after, and an output before and after. In that case, use finite-change sensitivity.

This tells you the average change in output per unit change in input over the interval. It is not necessarily the same as the derivative at one point, especially for nonlinear functions, but it is often a practical estimate.

Finite relative sensitivity uses percentage changes:

For example, if price increases from 50 to 55 and demand decreases from 200 to 180, then \(\Delta x/x_1=5/50=0.10\), and \(\Delta f/f_1=-20/200=-0.10\). The finite relative sensitivity is \(-0.10/0.10=-1\). This means a 10% increase in price was associated with a 10% decrease in demand over that interval.

Finite-change sensitivity is useful for experiments, before-and-after data, spreadsheet models and simulation outputs. The main caution is that it depends on the size of the interval. A large interval may hide local behavior that a derivative would show.

Deriving the Relative Sensitivity Formula

The relative sensitivity formula starts with the idea of fractional change. The fractional change in output is \(\Delta f/f\). The fractional change in input is \(\Delta x/x\). The ratio of these two quantities is:

For very small changes, \(\Delta f\) is approximated by \(df\), and \(\Delta x\) is approximated by \(dx\). Then:

Rearranging gives:

This form combines the derivative with the ratio \(x/f(x)\). The derivative measures local change in original units, and \(x/f(x)\) normalizes the result so it can be interpreted as a percentage-response measure.

The approximation becomes more accurate as the input change becomes smaller and the function behaves smoothly near the point. For large changes or sharp nonlinear behavior, finite-change sensitivity or a full scenario analysis may be more reliable than a derivative-based local estimate.

Worked Example: Revenue Sensitivity

Suppose total revenue is modeled by:

where \(x\) is units sold. Find absolute and relative sensitivity at \(x=40\).

First compute the derivative:

At \(x=40\), the derivative is:

So the absolute sensitivity is 10 revenue units per additional unit sold, near \(x=40\).

Next compute revenue at \(x=40\):

Now apply relative sensitivity:

The interpretation is that a 1% increase in units sold is associated with an approximate 0.3333% increase in revenue near \(x=40\). This is a local result. At a different value of \(x\), both the derivative and revenue level may change.

Revenue, cost and margin sensitivity are common business applications. When a problem includes actual revenue values, a tool such as the revenue calculator can help compute the base output before sensitivity is interpreted.

Worked Example: Cost Sensitivity

Suppose a cost function is:

where \(x\) is units produced. Find sensitivity at \(x=50\).

The derivative is:

At \(x=50\):

The absolute sensitivity is 30 cost units per additional unit produced. This is closely related to marginal cost. For a focused review of that idea, see the marginal cost formula.

Now compute the base cost:

Relative sensitivity is:

This means a 1% increase in output quantity is associated with about a 1.11% increase in cost near \(x=50\). The sensitivity is greater than 1 because the quadratic term makes cost grow more than proportionally at that point.

Worked Example: Power Function Sensitivity

Power functions have a clean sensitivity pattern. Let:

The derivative is:

The relative sensitivity is:

So for a power function, relative sensitivity equals the exponent. If \(f(x)=x^3\), relative sensitivity is 3. A 1% increase in \(x\) produces approximately a 3% increase in \(f(x)\) for small changes. If \(f(x)=\sqrt{x}=x^{1/2}\), relative sensitivity is \(1/2\). A 1% increase in \(x\) produces approximately a 0.5% increase in \(f(x)\).

This is one reason elasticity is so useful. It reveals structure that may be less obvious from the raw derivative. For \(x^3\), the absolute derivative \(3x^2\) changes with \(x\), but the relative sensitivity remains constant at 3.

Partial Sensitivity with Multiple Inputs

Many models depend on more than one input. If \(y=f(x_1,x_2,\ldots,x_n)\), sensitivity can be measured with respect to each input separately. The absolute sensitivity with respect to \(x_i\) is the partial derivative:

The relative sensitivity with respect to \(x_i\) is:

This tells you how a small percentage change in \(x_i\) affects the output, holding other variables constant. The phrase "holding other variables constant" matters. A partial sensitivity is not the same as changing all inputs at once.

For example, profit might depend on selling price, units sold, unit cost and fixed cost. Sensitivity with respect to selling price is different from sensitivity with respect to unit cost. If a decision involves margin targets or selling prices, related tools such as the selling price calculator and profit margin calculator can provide base values before the sensitivity calculation.

In engineering, partial sensitivities help identify which parameter most affects performance. In finance, they help identify which assumption most affects a valuation or risk measure. In science, they help decide which experimental parameter needs the most accurate measurement.

Interpreting Sensitivity Values

The sign matters. Positive sensitivity means the output increases when the input increases. Negative sensitivity means the output decreases when the input increases. The magnitude tells you how strong the response is. A value of \(-2\) indicates a stronger response than \(-0.2\), but in the opposite direction.

Always interpret sensitivity near the point where it was computed. A nonlinear function may have low sensitivity at one point and high sensitivity at another. Sensitivity is not automatically constant across the full input range.

Sensitivity Analysis Workflow

A practical sensitivity analysis follows a repeatable workflow. First, define the output of interest. This might be revenue, cost, profit, risk, concentration, temperature, speed, error, yield or any model result. Second, identify the input variable or variables to test. Third, choose whether absolute sensitivity, relative sensitivity, finite-change sensitivity or scenario sensitivity is most appropriate.

Fourth, compute a baseline. You need a base input \(x\), base output \(f(x)\), and either a derivative or a before-and-after output. Fifth, compute the sensitivity value. Sixth, interpret it in context. A high sensitivity value may mean the input must be measured carefully, monitored closely or controlled tightly.

Finally, communicate the result clearly. State the input, output, point of evaluation and formula used. A sensitivity value without context is incomplete. "The sensitivity is 0.8" is less useful than "At \(x=50\), a 1% increase in \(x\) is associated with an approximate 0.8% increase in \(f(x)\)."

For business lessons on output, cost and revenue relationships, costs and revenues can provide the model context before sensitivity is applied.

Graphical Interpretation

On a graph, absolute sensitivity is slope. For a function \(f(x)\), the derivative \(df/dx\) is the slope of the tangent line at the point. If the graph is steep, absolute sensitivity is large in magnitude. If the graph is flat, absolute sensitivity is near zero.

Relative sensitivity adds scale. A steep slope may not mean high relative sensitivity if the output value is also very large. A moderate slope may produce high relative sensitivity if the output is small. This is why relative sensitivity uses \(x/f(x)\) as a scaling factor.

For nonlinear functions, sensitivity changes as you move along the graph. A curve may be nearly flat at low input values and steep at high input values, or the reverse. A single sensitivity value describes one point or one interval, not necessarily the entire curve.

Plotting sensitivity values alongside the original function can be useful. The original graph shows output behavior; the sensitivity graph shows responsiveness. In model analysis, this can reveal thresholds, turning points and regions where assumptions matter most.

Common Mistakes to Avoid

Sensitivity Formula in Decision-Making

Sensitivity analysis helps decide where attention should go. If a model output is highly sensitive to price, then price assumptions deserve careful testing. If the output is highly sensitive to material cost, procurement and supplier changes may matter more. If a variable has low sensitivity, it may be less important for the specific decision.

In finance and business, sensitivity analysis is used to test assumptions such as sales volume, discount rate, cost growth, price changes and margin targets. The discount formula and revenue-related tools can support examples where price reductions affect output.

In engineering, sensitivity helps identify design parameters that strongly influence performance or safety. In science, it helps determine which measurements require higher precision. In data modeling, it helps identify assumptions that drive predictions.

The key is to avoid treating sensitivity as a final decision by itself. Sensitivity tells you responsiveness. The decision still depends on goals, constraints, uncertainty, cost, risk and feasibility.

Diagnostic-Test Sensitivity: A Different Formula

The term sensitivity is also used in diagnostic testing and classification. That formula is different from model sensitivity. Diagnostic sensitivity measures the proportion of actual positives correctly identified by a test.

This is sometimes called the true positive rate or recall. For example, if a test correctly identifies 90 true positives and misses 10 actual positives, sensitivity is \(90/(90+10)=0.90\), or 90%.

This article focuses primarily on mathematical sensitivity analysis for functions and models, because that is the focus of the current page. The diagnostic-test formula is included only to prevent confusion. Diagnostic interpretation, medical screening and classification model evaluation require additional context, including specificity, prevalence, predictive values, thresholds and the consequences of false positives and false negatives.

Rounding and Reporting Sensitivity Results

Sensitivity values should be reported with enough precision to be useful, but not so many digits that they imply false accuracy. If inputs are approximate, sensitivity results are also approximate. A result of \(0.333333333\) may be better reported as \(0.333\) or \(0.33\), depending on the context.

When reporting relative sensitivity, include the percentage interpretation. For example, \(S_{\text{rel}}=0.333\) means a 1% input increase is associated with about a 0.333% output increase, near the evaluation point. This is clearer than simply writing the number.

When reporting absolute sensitivity, include units. If cost is measured in dollars and quantity is measured in units, \(dC/dx=30\) means 30 dollars per unit. Without the unit, the number is incomplete.

If you need a refresher on precision, the rounding and significant figures guide is useful for deciding how many digits to keep in final answers.

Practice Problems

When solving practice problems, always compute the base output before applying the relative sensitivity formula. The denominator \(f(x)\) is what turns a derivative into a percentage-response measure.

Local Sensitivity vs Global Sensitivity

Local sensitivity measures responsiveness near one point. The derivative-based formulas on this page are local formulas because they describe behavior at a particular value of \(x\). If \(x=40\), the sensitivity value tells you how the model behaves near \(40\), not necessarily at \(10\), \(100\), or across the entire domain.

Global sensitivity asks a broader question: across a wider range of possible inputs, which variables have the greatest influence on the output? Global methods may use scenario tables, simulation, sampling, variance-based measures, tornado charts, or other model-testing tools. The mathematical formulas are different, but the goal is related: understand which assumptions drive the result.

For smooth functions, local sensitivity is often a good first step. It is quick, interpretable, and connected to calculus. If the model is strongly nonlinear, has thresholds, includes discontinuities, or depends on interactions among variables, local sensitivity may be incomplete. In that case, evaluate sensitivity at several points or run finite-change scenarios.

For example, the revenue function \(R(x)=50x-0.5x^2\) has derivative \(50-x\). At \(x=10\), the derivative is 40. At \(x=40\), it is 10. At \(x=50\), it is 0. The same model has very different local sensitivity at different production or sales levels. A single value does not summarize the whole curve.

A practical workflow is to begin with local sensitivity at the current operating point, then test nearby scenarios, then test larger ranges if the decision is important. This gives both mathematical precision and practical robustness.

One-Way Sensitivity Analysis

One-way sensitivity analysis changes one input at a time while holding other inputs constant. It is common in spreadsheets, business models, engineering design and risk analysis because it isolates the effect of one variable. If the output changes strongly when one input changes, that input is influential under the assumptions of the model.

Suppose profit depends on price, quantity and unit cost. A one-way analysis might increase price by 5% while keeping quantity and cost fixed, then increase quantity by 5% while keeping price and cost fixed, then increase cost by 5% while keeping price and quantity fixed. The relative sensitivity formula is the calculus version of this idea for small changes.

One-way analysis is easy to explain, but it has limits. In real systems, inputs may not be independent. Price changes may affect demand. Material cost changes may affect production choices. Weather assumptions may affect both demand and cost. If variables move together, one-way sensitivity can understate or overstate the real effect.

Still, one-way analysis is useful for screening. It tells you where to look first. If a model is barely affected by a 10% change in one input but strongly affected by a 1% change in another, the second input needs more attention. After identifying the sensitive variables, you can move to two-way sensitivity tables or scenario analysis.

Two-Way Sensitivity and Interaction Effects

Two-way sensitivity analysis changes two inputs at the same time. This is important when variables interact. An interaction exists when the effect of one input depends on the value of another input. In business, price and quantity often interact because a higher price can reduce quantity sold. In engineering, temperature and material properties may interact. In finance, interest rates and growth assumptions may interact.

For a two-variable function \(f(x,y)\), the local absolute sensitivities are partial derivatives:

These measure the effect of changing one variable while holding the other constant. Interaction can be studied using a mixed partial derivative:

You do not always need advanced calculus to understand the idea. A two-way table can show how output changes for combinations of two inputs. For example, a profit table can show selling price along one axis and unit cost along the other. The table reveals whether a favorable price can offset a higher cost, or whether margin collapses under certain combinations.

When communicating two-way sensitivity, avoid claiming that one variable is always most important unless you have tested the relevant range. The most important driver may change depending on the starting point and the values of other variables.

Scenario Sensitivity in Spreadsheets

Many users meet sensitivity analysis through spreadsheet scenarios. A base case uses the most likely assumptions. A low case tests unfavorable assumptions. A high case tests favorable assumptions. The sensitivity formula helps interpret why the output changes across those cases.

A simple spreadsheet might have inputs for unit price, quantity sold, unit cost and fixed cost. Revenue is price times quantity. Profit is revenue minus total cost. If price changes by 5% and profit changes by 12%, profit is more sensitive than price in that range. If cost changes by 5% and profit changes by 20%, cost may be the larger risk driver.

When building scenarios, keep the original inputs visible and put changed inputs in separate columns. Label each scenario clearly. Do not overwrite the base case. A clean model lets you compare the base case, downside case and upside case without losing the assumptions behind each result.

Use formulas consistently. If relative sensitivity is calculated as \((\Delta f/f)/(\Delta x/x)\), use the same base values for every scenario. Mixing base denominators and changed denominators can make sensitivities inconsistent. For small changes, the difference may be minor. For large changes, it can be significant.

Spreadsheet sensitivity is most useful when paired with interpretation. A table of numbers is not enough. State which input changes the output most, whether the change is realistic, and what decision follows from that insight.

Sensitivity and Elasticity in Economics

In economics, relative sensitivity is often called elasticity. Price elasticity of demand measures the percentage change in quantity demanded divided by the percentage change in price. Income elasticity measures the percentage change in demand divided by the percentage change in income. Cross-price elasticity measures the sensitivity of demand for one good to the price of another good.

If demand elasticity with respect to price is \(-1.5\), then a 1% increase in price is associated with an approximate 1.5% decrease in quantity demanded, near the point of measurement. The negative sign shows the inverse relationship. The magnitude shows responsiveness.

Revenue sensitivity can be subtle because revenue depends on both price and quantity. Raising price increases revenue per unit but may reduce quantity sold. Whether total revenue increases or decreases depends on demand sensitivity. This is why sensitivity analysis is valuable for pricing decisions, discounts and sales forecasts.

For business calculations, base values such as revenue and margin must be computed correctly before sensitivity is interpreted. The revenue calculator, profit margin calculator and selling price calculator can support those base calculations before you analyze responsiveness.

Sensitivity in Engineering and Science

Engineering models often include parameters such as dimensions, material properties, loads, temperatures, flow rates, voltages, concentrations and time constants. Sensitivity analysis helps identify which parameter most affects performance. This can guide design tolerances, measurement priorities and safety margins.

For example, if a beam deflection model is highly sensitive to length, then small errors in length measurement can produce large output errors. If the same model is less sensitive to a material parameter, that parameter may need less precise measurement for the specific calculation. Sensitivity does not tell you whether a variable is important in every context; it tells you how important it is for the model and point being studied.

Scientific experiments also use sensitivity thinking. If an output is highly sensitive to temperature, the experiment may require careful temperature control. If it is highly sensitive to concentration, measurement accuracy for concentration becomes critical. Sensitivity analysis helps allocate effort to the variables that most affect the result.

In nonlinear systems, sensitivity can change across operating ranges. A stable system may be insensitive near one equilibrium and highly sensitive near another. For that reason, engineering and scientific sensitivity results should state the operating point, assumptions and range tested.

Sensitivity and Uncertainty

Sensitivity and uncertainty are closely related but not identical. Sensitivity tells how strongly an output responds to an input. Uncertainty describes how much the input might vary or how uncertain its true value is. A variable can be highly sensitive but well known, or weakly sensitive but highly uncertain.

The effect on output depends on both. A useful approximation is:

If \(df/dx\) is large but \(\Delta x\) is tiny, the output uncertainty may still be small. If \(df/dx\) is moderate but \(\Delta x\) is large, the output uncertainty may be important. Sensitivity alone does not measure risk unless input uncertainty is also considered.

For relative changes, the same idea applies:

If relative sensitivity is 2 and the input is uncertain by 3%, the output is approximately uncertain by 6%, assuming the local linear approximation is valid. This is a practical way to translate input uncertainty into output uncertainty.

Using Sensitivity to Prioritize Inputs

When a model has many inputs, sensitivity analysis can help prioritize them. Inputs with large relative sensitivity values are often the strongest drivers of output. Inputs with small sensitivity values may be less important, at least near the baseline. This ordered view helps decide where to spend time, money and attention.

For example, a business model may depend on sales volume, unit price, cost per unit, fixed cost and discount rate. If profit is most sensitive to sales volume, market demand assumptions deserve extra attention. If profit is most sensitive to unit cost, supplier contracts and production efficiency may matter more. If the output is sensitive to the discount rate, financing assumptions deserve careful review.

However, the ordered view should be interpreted carefully. A variable with low sensitivity but very high uncertainty can still matter. A variable with high sensitivity but tight control may be less risky than it looks. Sensitivity-based prioritization is a guide for investigation, not a substitute for judgment.

A good sensitivity report lists the top drivers, explains the tested range, states whether values are local or scenario-based, and connects the findings to practical decisions. This turns formulas into usable analysis.

Model Audit Checklist

Before trusting a sensitivity result, audit the model. First, confirm the output definition. Sensitivity of revenue is different from sensitivity of profit, margin, cost or cash flow. Second, confirm the input definition. Price, quantity, cost and rate variables must be labeled clearly.

Third, confirm the evaluation point. A derivative at \(x=40\) does not automatically describe the model at \(x=100\). Fourth, check units. Absolute sensitivity should have meaningful units, while relative sensitivity should be dimensionless. Fifth, confirm that denominators are not zero in relative formulas.

Sixth, check whether the change is small enough for derivative-based interpretation. If the tested change is large, use finite-change sensitivity or scenario analysis. Seventh, check whether variables are independent. If changing one input also changes another, one-way sensitivity may miss part of the real effect.

Eighth, report rounding carefully. If a sensitivity value is based on approximate inputs, do not display excessive digits. If the result is used in a decision, show enough detail for someone else to reproduce the calculation.

More Worked Examples

Linear function

Let \(f(x)=20+5x\). Then \(df/dx=5\). At \(x=10\), \(f(10)=70\). Absolute sensitivity is 5, and relative sensitivity is:

A 1% increase in \(x\) produces an approximate 0.7143% increase in \(f(x)\).

Inverse function

Let \(f(x)=100/x\). Then \(df/dx=-100/x^2\). At \(x=20\), \(f(20)=5\), and \(df/dx=-0.25\). Relative sensitivity is:

A 1% increase in \(x\) produces an approximate 1% decrease in \(f(x)\).

Exponential function

Let \(f(x)=e^{0.2x}\). Then \(df/dx=0.2e^{0.2x}\). Relative sensitivity is:

At \(x=5\), relative sensitivity is 1. At \(x=10\), it is 2. The sensitivity grows with the input.

When the Sensitivity Formula Is Not Enough

The sensitivity formula is powerful, but it is not a complete model validation method. It does not prove that the model is correct. It does not show whether inputs are realistic. It does not automatically account for uncertainty, correlations, constraints, thresholds or real-world behavior outside the tested range.

If a decision is high-stakes, use sensitivity analysis as one part of a broader review. Check the model assumptions, data quality, units, range of validity and reasonableness of outputs. Compare model behavior with known examples. Test scenarios beyond the base case. Document the limitations.

For classroom work, the formula and interpretation are usually enough. For professional work, sensitivity should be paired with validation, uncertainty analysis and clear communication. The value of sensitivity analysis comes from asking better questions, not just producing a number.

Choosing the Right Sensitivity Formula

The best sensitivity formula depends on the question being asked. If the question asks how many extra output units are produced by one extra input unit, use absolute sensitivity. If the question asks how strongly the output responds in percentage terms, use relative sensitivity. If the model is given as a table, simulation, spreadsheet or before-and-after observation, use a finite-change formula instead of forcing a derivative.

Use absolute sensitivity when the original units carry the meaning of the decision. In a production setting, the units might be dollars per unit, kilograms per degree, metres per second, marks per hour of study or joules per kilogram. A result such as \(df/dx=8\) is only clear when the units are included. If \(f(x)\) is cost in dollars and \(x\) is quantity, \(df/dx=8\) means the local marginal cost is 8 dollars per extra unit. If \(f(x)\) is distance in metres and \(x\) is time in seconds, the same number means 8 metres per second.

Use relative sensitivity when inputs and outputs have different scales or when a percentage comparison is more useful than a unit comparison. Relative sensitivity is especially helpful when two inputs use different units. For example, price might be measured in dollars, quantity in units, and a rate in percent. Their absolute sensitivities cannot be compared directly without context. Their relative sensitivities are dimensionless, so they can be placed beside one another more fairly:

Use finite-change sensitivity when the change is already known. If a spreadsheet shows that increasing a price from 20 to 22 changes profit from 5000 to 5400, you can compute the average sensitivity over that interval. The result is not the exact derivative at a single point, but it is often the most honest result when the model is not expressed as a clean formula.

Building a Sensitivity Table

A sensitivity table makes the formula easier to audit because it separates inputs, baseline values, tested changes and output responses. Instead of reporting a single number without context, a table shows what was changed and how the result was produced. This is useful for school solutions, business cases and technical reports.

A simple one-way table begins with a baseline row. The baseline row states the original input value \(x_0\), the original output \(f(x_0)\), and the formula or spreadsheet model being used. The next rows test alternative input values. For each row, calculate the new output, the absolute change \(\Delta f\), the percentage output change \(\Delta f/f\), and the sensitivity measure that best matches the task.

This table shows why the interval matters. The first change gives an average sensitivity of 10, while the wider change gives an average sensitivity of 9. The values are close, but not identical. That difference may happen because the model is nonlinear, because rounding is present, or because the tested interval is too wide for a single local derivative to describe it well.

When a table is used in a report, include the direction of the test. A 5% increase and a 5% decrease do not always produce symmetric output changes. Many real models have thresholds, minimum order sizes, fixed charges, tax brackets, capacity limits or nonlinear formulas. A useful sensitivity table therefore tests both upward and downward movement when the decision depends on both risk and opportunity.

Writing a Clear Sensitivity Interpretation

A sensitivity result should be written as a sentence, not only as a formula. The formula proves how the number was found, but the interpretation explains what it means. A clear interpretation should include the input, the output, the baseline point, the direction of the effect, and whether the result is absolute or relative.

For absolute sensitivity, a strong sentence has this form: at the baseline \(x=a\), increasing the input by 1 unit changes the output by approximately \(df/dx\) output units. For example: at \(x=40\), an extra unit sold increases revenue by approximately 10 dollars, assuming the local model remains valid. This is clearer than saying only \(S=10\).

For relative sensitivity, a strong sentence has this form: near the baseline, a 1% increase in the input produces an approximate \(S_{\text{rel}}\)% change in the output. If \(S_{\text{rel}}=1.4\), a 1% input increase produces an approximate 1.4% output increase. If \(S_{\text{rel}}=-0.6\), a 1% input increase produces an approximate 0.6% output decrease. The sign matters because it tells whether the relationship is direct or inverse.

When the sensitivity value is based on a finite change, say so directly. A suitable sentence is: over the tested interval from \(x=a\) to \(x=b\), the average change in output was \(\Delta f/\Delta x\) output units per input unit. This avoids overstating the result as an exact derivative. It also helps readers understand that a different interval could produce a different sensitivity.

Good interpretation also avoids false precision. A sensitivity value of 0.333333333 is usually better reported as 0.333 or about one third unless the input data justify more digits. If the result comes from measured values, estimated values or rounded model outputs, the written conclusion should reflect that uncertainty.

Sensitivity Formula for Classroom Questions

In mathematics and exam-style questions, the most reliable approach is to define the variables before calculating. Write the function, state the point, differentiate if required, substitute the point, and then interpret. This order prevents common mistakes such as substituting before differentiating, confusing \(f(x)\) with \(df/dx\), or treating a relative sensitivity as a percentage without multiplying by the input-output ratio.

For a differentiable function, the working often follows this pattern:

The first result is the absolute sensitivity at the point. The final result is the relative sensitivity at the same point. Keeping both values visible helps because many questions ask for one but expect you to understand the other. If the problem says "per unit change," absolute sensitivity is usually the target. If the problem says "percentage change," "proportional change," or "elasticity," relative sensitivity is usually the target.

For functions that include powers, the relative sensitivity can often be simplified. If \(f(x)=kx^n\), then \(df/dx=knx^{n-1}\), so:

This means the exponent \(n\) gives the percentage response. If \(f(x)=5x^3\), a 1% increase in \(x\) produces an approximate 3% increase in \(f(x)\). If \(f(x)=4x^{-2}\), a 1% increase in \(x\) produces an approximate 2% decrease in \(f(x)\). This shortcut is useful, but it should still be supported by the formula when showing full working.

Sensitivity Formula for Business Models

Business models use sensitivity analysis to identify which assumptions deserve the most attention. A sales forecast might depend on price, volume, conversion rate, refund rate and advertising cost. A project model might depend on initial investment, operating cost, discount rate and growth assumptions. A pricing model might depend on cost, margin, markdown and customer demand.

For revenue and pricing work, sensitivity is closely related to marginal thinking. The revenue calculator is useful when the model is \(R=PQ\), while the profit margin calculator helps connect selling price, cost and margin. When the question is how much a result changes as cost changes, the same idea appears in marginal formulas such as the marginal cost formula.

In a business report, sensitivity should not be presented as a mechanical list of numbers. The useful conclusion is operational. If profit is highly sensitive to unit cost, the practical response may be supplier review, waste reduction or process improvement. If profit is highly sensitive to price, the response may be price testing, discount control or value communication. If profit is highly sensitive to sales volume, the response may be demand forecasting, marketing analysis or capacity planning.

The formula also helps avoid overconfidence. A model can show a profitable base case while still being fragile. If a small input change produces a large output swing, the decision depends heavily on that assumption. In that situation, sensitivity analysis should lead to better evidence, not just a more polished spreadsheet. The analyst should ask whether the assumption is supported by data, whether the tested range is realistic, and whether the decision still works under less favorable conditions.

Frequently Asked Questions

Final Sensitivity Formula Checklist

Start by identifying the output \(f(x)\), the input \(x\), and the point where sensitivity is being evaluated. Use \(df/dx\) for absolute sensitivity. Use \((x/f)(df/dx)\) for relative sensitivity. Use \(\Delta f/\Delta x\) or \((\Delta f/f)/(\Delta x/x)\) when working with finite changes rather than derivatives.

Interpret the sign, magnitude, units and point of evaluation. Report relative sensitivity as a percentage-response statement whenever possible. Use rounding that matches the precision of the inputs. Most importantly, connect the sensitivity result back to the decision or model question: which input matters, how strongly does it matter, and what should be checked next?