Scenario Modeling Tip #1: Include only constraints that are real business constraints in a marketing optimization scenario.
Common misconception: Adding constraints makes the optimization problem easier. In fact, the opposite is true.
Example
Suppose you made 5,000 offers from Campaign A last month, and you decide that you again want to make about 5,000 offers from the same campaign this month. Should you add constraints to SAS Marketing Optimization to require that you must make at least 4,900 offers and at most 5,100 offers from Campaign A?
The incorrect assumption is that specifying a range of only 200 makes it easier to solve the problem because the optimization must search for solutions only within this smaller range. However, the optimization algorithm that is used by SAS Marketing Optimization searches for the best solution that lies within the feasible region, which is the set of solutions that satisfies all the constraints.
So, adding constraints to the problem makes the feasible region smaller and, when the feasible region becomes smaller with respect to the aggregate constraints, the optimization problem becomes more difficult to solve.
Do not include a constraint that requires at least 4,900 offers unless you are unwilling to accept a solution with fewer than 4,900 offers--that is, unless the constraint is a real business constraint.
To determine whether a constraint is a real business constraint, ask yourself, “If I could improve my objective value, but doing so would require me to violate this constraint, would I prefer the solution with the better objective, or would I prefer the solution with the worse objective that satisfies the constraint?” If you would prefer the solution with the better objective, then do not include the constraint in the optimization problem because it artificially confines you to suboptimal solutions. You might then decide to remove the constraint from the scenario, or you might decide to decrease the constraint limit to a lower value that you are willing to accept, thus trading off the possibility of being able to improve the objective value. As you decrease the constraint limit, you can keep asking yourself, “Am I willing to accept a lower objective in order to ensure that this constraint is satisfied?” Set the limit when the constraint satisfaction is more important than an improved objective.
Similarly, you might want to ask yourself the same question about the constraint that requires at most 5,100 offers from Campaign A. Making 5,000 offers from Campaign A last month is not necessarily a compelling reason to require the same number of offers this month. For example, if you were able to increase your objective value by 10%, but doing so would require 10,000 offers from Campaign A, would you be willing to make 10,000 offers from Campaign A? If the answer is yes, you should not include the constraint in the scenario, but rather let the optimization decide the best allocation of offers among campaigns in order to maximize your objective.
Note: This post was adapted from Michelle Opp’s “Scenario Modeling Tips” document. For more information, see the SAS Marketing Optimization User’s Guide.
Registration is now open for SAS Innovate 2025 , our biggest and most exciting global event of the year! Join us in Orlando, FL, May 6-9.
Sign up by Dec. 31 to get the 2024 rate of just $495.
Register now!
Data Literacy is for all, even absolute beginners. Jump on board with this free e-learning and boost your career prospects.