Sensitivity Analysis

What Is Sensitivity Analysis?

Sensitivity analysis is a way of asking "what if?" about a financial model. An analyst starts with a base case, such as a forecast of revenue, costs, or interest rates, and then changes one assumption at a time or several at once to see how much the result moves. The goal is to find out which inputs matter most and how fragile a conclusion is to being wrong about them.

How It Works

The process usually follows a few steps. First, the analyst identifies the key assumptions in a model, such as sales growth, profit margin, discount rate, or terminal value growth. Second, each assumption is varied across a reasonable range, often above and below the base case. Third, the model is recalculated for each change, and the results are recorded. Analysts often present the outcomes in a table or grid, where each row and column shows a different combination of inputs and the cell shows the resulting value.

An Example

Suppose an investor values a company using a discounted cash flow model that produces a fair value per share. The analyst might test discount rates of 8%, 9%, and 10% alongside long-term growth rates of 2%, 3%, and 4%. The resulting grid shows how the value per share shifts across those combinations. If small changes in the discount rate produce large changes in value, the valuation is highly sensitive to that input, and the analyst knows to scrutinize it more closely.

Why It Matters to Investors

Sensitivity analysis helps investors understand the range of plausible outcomes rather than relying on a single number. It can reveal when a investment case depends heavily on one optimistic assumption, and it makes clear which variables deserve the most research. It is also a common component of scenario analysis, where several assumptions are changed together to represent a coherent story, such as a recession. Its main limitation is that it only tests the inputs the analyst chose to vary. Results are only as reliable as the model structure and the ranges selected, and unusual events outside those ranges can still produce outcomes the analysis did not capture.