Sensitivity Analysis in Excel
Sensitivity Analysis in Excel
Procedure for performing sensitivity analysis when solving LPs using Excel Solver:
When you have run Solver and a solution has been offered you see the following dialog box
Note: you must select Assume Linear Model from the Solver Options in order to get the sensitivity analysis
information
This will produce a new worksheet in your workbook called “Sensitivity Report 1”
For example, in the case of IMC Investments the sensitivity report looks something like this:
Objective Function
Sensitivity Analysis
RHS of Constraints
Sensitivity Analysis
1
Interpreting the Report
As you can see, the sensitivity report is divided into two sections – the top half concerns changes to the
objective function, the bottom section with changes to the constraints.
The table above shows the four decision variables (EOIL, AOIL, ASTL, CLMB) listed.
The Final Value column gives the optimal values for each of the four variables ($60,000 invested in EOIL
and $40,000 in CLMB).
The Objective Coefficient column gives the coefficient value for each variable.
The Allowable Increase column shows how much the coefficient can increase without changing the
optimal result. Therefore the EOIL coefficient can increase by a huge amount (effectively infinite), the AOIL
coefficient can increase by 0.01, and so on, without the solution changing.
The Allowable Decrease column shows how much the coefficient can decrease without changing the
optimal result. Therefore the EOIL coefficient can decrease by 0.01, the AOIL coefficient can decrease an
infinite amount, and so on, without the solution changing. This is very useful when investigating whether
changes to the objective function will impact on the optimal solution.
Note: The Allowable Increase and Decrease only apply to individual coefficient changes. If two or more
coefficients are changed then the solution may well change. To investigate multiple changes you must go
back into the model and run Solver again with the new set of coefficients.
The lower half of the report shows the sensitivity ranges for the four constraints.
Final Value gives the values for the four constraints at the optimal solution. As you can see constraint 1
has a value of 40,000, constraint 2 has a value of 28,000 and so on.
Shadow Price gives the increase in the objective function for each unit increase in the RHS value of each
constraint. Therefore as we increase the RHS of constraints of 1 and 2 we get no increase in the objective
function value, whereas each increase in the RHS of constraint 3 gives us 0.03 dollars increase in our OF.
Similarly the shadow price for constraint 4 (the total amount invested) shows that our OF improves 0.078
dollars for every RHS increase.
This is important for judging whether it is cost-effective to alter the conditions of the problem, for
example investing more money, increasing working time or purchasing new machinery.
The Constraint R.H. Side gives the current RHS values of all the four constraints
The Allowable Increase shows how much the RHS value can increase such that the solution remains at the
intersection of the same constraints. In other words the upper limit of the sensitivity range. So, for
example, constraint 4 (the total amount invested) can increase by $78,571 and the solution will continue
to be situated at the intersection of the same constraints.
In other words, as we increase the total amount invested (our constraint 4 RHS) we can go as far as
178,571 dollars before we go beyond our sensitivity boundaries. And each one of those additional 78,571
dollars invested will yield an increase in our OF of 0.078 dollars (as this is our shadow price for constraint
4).
Note: the optimal solution can change as the RHS values are altered. The sensitivity range just gives the
limits where the solution lies at the intersection of the same constraints.
The Allowable Decrease gives the lower limit to the sensitivity range for each of the four constraints. For
example, constraint 2´s RHS can decrease by 22,000 (to a final value of 28,000) before the solution
changes to the intersection of a different set of constraints.
2