Underwriting
How To Read A Sensitivity Table
Sensitivity tables show how an outcome moves as two inputs change, and their real value is revealing which assumptions the result depends on rather than producing a better estimate.
A sensitivity table varies two inputs and reports the resulting outcome in a grid. It is one of the few modelling tools that makes uncertainty visible rather than hiding it.
What the grid is actually showing
Each cell holds the model rerun with one combination of the two inputs. Reading across a row or down a column shows how much a single change moves the result.
The steepness of that movement is the point. A grid where values change slowly describes a robust deal; one where they swing wildly describes a fragile one.
This is different information from a single headline return. The headline says what the model produces under one set of assumptions, while the grid says how much that depends on them.
Choosing the two axes
The axes should be the inputs with the greatest leverage and the least certainty. For most hold-period models that means exit pricing and rent growth, or purchase price and financing cost.
Putting a well-known input on an axis wastes the grid. Varying an expense you can verify from bills produces a table that confirms the obvious.
Some of the most useful grids pair an input the buyer controls with one they do not, which shows how much good execution can offset unfavourable conditions.
The ranges have to be plausible and symmetric
A grid centred on the base case with equal movement in each direction is readable. One that explores generous outcomes further than adverse ones produces a comforting and useless picture.
Ranges should also reflect how the input actually behaves. Some inputs move gradually and others jump, and a grid with even increments can imply a smoothness that does not exist.
Where an input has a hard boundary, that boundary belongs in the range. A grid that never tests the point at which debt coverage fails omits the outcome that matters most.
Two inputs is a real limitation
Grids hold everything else constant, but inputs are correlated. Weak rent growth and unfavourable exit pricing tend to arrive together, driven by the same underlying conditions.
Reading only the corners of the grid partly compensates, because the corner where both inputs are adverse approximates a coherent downside rather than an isolated one.
For anything more complex, scenario analysis is the better tool. A scenario moves a whole set of inputs together in a way that describes a recognisable state of the world.
What to do with the result
The output of a sensitivity exercise is not a number. It is a description of which assumptions the deal depends on, and therefore which ones deserve verification effort.
If the grid shows the result hinges on an input nobody can check, that is a finding about the deal's character rather than a reason to refine the model further.
Used this way, the table redirects attention. It moves work from polishing a projection toward testing the two or three claims that actually determine whether the projection holds.
