Lesson 7 of 8
Reading a sensitivity grid
Twenty-five defensible cells spanning 66.32 to 215.57 a share for one company. The grid measures your uncertainty, and it still understates it.
By Javier Sanz
Founder of ValueScreener. Built Ninety Nine, a securities services provider, and sold it to a tier-one global bank; ran operations and internationalization at Alpaca, the brokerage API.
Who writes thisEditorial processLinkedIn (opens in a new tab)X (opens in a new tab)
In one minute
- A sensitivity grid is a statement about how little the model constrains the answer, not a range of likely outcomes.
- Two points either way on growth and the discount rate spanned 66.32 to 215.57 a share on one real company, a factor of 3.25.
- The grid holds three of the five decisions fixed, so the honest range is wider than the grid shows.
Every cell in a sensitivity grid is arithmetically correct. That is the point and the problem: the model cannot tell you which one is the company, because nothing in the model knows the company.
Value per share across the grid, Procter and Gamble
- Growth 1%, rate 10%66.32$
- Growth 3%, rate 9%92.02$
- Growth 3%, rate 8% · the centre cell110.81$
- Growth 3%, rate 7%137.95$
- Growth 5%, rate 6%215.57$
The whole grid
| Growth | 6% | 7% | 8% | 9% | 10% |
|---|---|---|---|---|---|
| 1% | 150.99 | 115.77 | 93.32 | 77.76 | 66.32 |
| 2% | 165.20 | 126.41 | 101.72 | 84.61 | 72.06 |
| 3% | 180.64 | 137.95 | 110.81 | 92.02 | 78.25 |
| 4% | 197.39 | 150.47 | 120.65 | 100.03 | 84.94 |
| 5% | 215.57 | 164.02 | 131.29 | 108.69 | 92.16 |
Read across a row: one point of discount rate is worth between 15 and 20 percent of the answer. Read down a column: one point of growth is worth about 8 percent. The rate axis moves the model roughly twice as hard as the growth axis, which is why the least verifiable input deserves the most argument.
The axis that is not hypothetical
The rate axis is not a thought experiment. Four ordinary ways of measuring this company's beta, all from the same stored closing prices, produce a cost of equity from 4.67 percent to 7.22 percent. The 2 point band on the grid is narrower than the measurement spread on one input.
Three ways to misread it
Taking the centre cell. It is the centre because it is where your assumptions were, not because it is more likely. The grid is built around your inputs by construction.
Averaging the cells. The average of 25 arbitrary points is a 26th arbitrary point with a false air of consensus.
Reading the colours as a verdict. ValueScreener tints each cell by its distance from the stored share price so the edges are visible at a glance. That is a comparison with a price, not a statement about the business. Cells where the terminal growth rate is not below the discount rate print n/m rather than a number, because the perpetuity is undefined there.
The grid understates the range
Five decisions go into the model and the grid moves two of them. Every cell in the table above holds the same cash flow definition, the same ten-year horizon and the same 2.5 percent terminal rate.
Move the cash flow definition and the centre cell alone runs from 104.90 to 110.81. Move the terminal rate by a point and it moves again. The real range around a DCF is wider than any grid a spreadsheet prints, which is the argument for writing a range and a list of assumptions rather than a number.
What to write down instead
A sentence with the shape: at 8 percent and 3 percent this business is worth about 111 a share; the answer holds above 92 only if the rate is 9 percent or better, and the whole exercise assumes a definition of cash flow and a terminal state I have written beside it. That sentence can be checked later. A single figure cannot.
Check yourself
What does a sensitivity grid actually measure?
- The probability of each outcome
- How much the answer moves when inputs you cannot verify move by a little
- The range of prices the share is likely to trade at
Show the answer
Answer: B.
Every cell is arithmetically correct. The grid shows how weakly the model constrains the answer, which is a fact about the model and about your uncertainty.On the grid above, which axis moves the answer harder?
- Growth, because it compounds over ten years
- The discount rate, worth 15 to 20 percent a point against about 8 percent for growth
- They move it equally, by construction
Show the answer
Answer: B.
The rate discounts every year and sets the terminal divisor, so it strikes the perpetuity as well as the forecast.Why is the real range around a DCF wider than the grid?
- Because grids exclude the terminal value
- Because the grid moves two inputs and holds the cash flow definition, the horizon and the terminal rate fixed
- Because the share price moves daily
Show the answer
Answer: B.
A two-axis grid is a slice through a five-dimensional set of choices. The other three are constant in every cell.Sources
- Procter and Gamble, Form 10-K, fiscal year ended 30 June 2026, filed 2026-08-04: the filed cash flow, debt, cash and share count behind every cell.
- Damodaran, Investment Valuation: sensitivity analysis as a statement of estimation uncertainty rather than a range of outcomes.
- Beta spread computed from stored daily closes for this company and SPY, 2021-08-13 to 2026-09-10, under four window and frequency choices.
What this leaves out
- The grid is arithmetic on one set of filed inputs for one company. It is not a view about that company, and no cell is a target.
- The colouring compares each cell with a stored closing price, which is not a live quote.
- A grid cannot include the risk that the business itself changes, which is the risk that actually decides outcomes.
- Nothing here is advice about any security.