Optimizer Predictability & Explainability Challenge
We tried leveraging Optimizer to resolve issue for our business requirement. It did not work well for us due to below reasons. We are opening this ticket asking Anaplan to look into enhancing the Optimizer Predictability & Explainability.
Here is our note from the development team:
Regarding the optimizer predictability challenge: We were trying to optimize the allocation for multiple products at a time. Simply changing the product mix (adding/removing a product into the optimizer request) would change the optimizer suggestion for a product from run to run.
Regarding the optimizer explainability challenge: We couldn’t identify a consistent pattern in the results. When trying to explain the suggestion and results, one suggestion we heard was that the optimizer would prefer the first member in a list; however, we couldn’t consistently demonstrate that pattern or figure out why the optimizer preferred one solution over another.
Another challenge we had was with applying a “fair share” concept. We couldn’t quite get the optimizer to suggest splits across equally rated customers and alternatives. Some customers with the same rating (e.g., everything the same) would unfairly be allocated proportionately more.
We would like to see more explainability that shows why Optimizer generate the number that it did. or What was the most contributing factor that drive the optimization.
Get Started with Idea Exchange
See our Submission Guidelines and Idea Evaluation Criteria, then start posting your own ideas and showing support for others!