Question
How do you determine if a business problem is a good fit for the Optimizer, and what prerequisites are required before starting?
Answer
You must first clearly define your business case by identifying your decision variables, constraints, and a single, clear objective. Mathematically, you must ensure that all equations are linear or can be linearized, as the solver cannot process non-linear equations directly. If the model is large or complex, you can utilize AI tools, Anaplan partners, or Professional Services to help structure and decompose the problem.
Question
How do you manage settings when dealing with multiple parameters, over ten variables, and numerous constraints? Is there an automated tool to determine parameter ranges, or is it a manual process?
Answer
Anaplan automatically maps and distributes your defined variable types across your dimensions in the backend, meaning you do not have to configure millions of individual combinations yourself. Powered by Gurobi, the platform easily scales to handle millions of variables and constraints. While setting variable bounds (such as a minimum limit of zero) is currently a manual process, the product team is actively researching dynamic bounds as a potential roadmap feature.
Question
Can the Co-Modeler AI tool automatically generate the actual optimizer actions in the backend when prompted?
Answer
This capability is not yet available, but it is actively on the product roadmap for a future release to allow the AI to handle both the model construction and the action generation.
Question
Does a model builder need a background in optimization or data science to build these models?
Answer
No background in data science is required for the majority of use cases. Most common business problems are straightforward enough to build as a proof-of-concept using standard Anaplan platform knowledge and basic math with the help of Co-Modeler. For massive, highly complex global supply chains that require clustering or advanced decomposition, you can leverage Anaplan’s professional services or partner network.
Question
Is the Optimizer suitable for handling discrete data sets containing over a thousand items?
Answer
Yes, a model of that size is considered very small and is easily solved by the Optimizer. While the Gurobi backend can scale to tens of millions of variables, builders should keep in mind that there is a strict 60GB memory limit on the optimization server for extremely large enterprise models.
Question
Does enabling new platform features alter existing optimizer models or actions, and is testing in production recommended?
Answer
New features and backend solver updates (such as upgrades to the Gurobi engine) are rolled out automatically and will not break existing setups. However, standard Application Lifecycle Management (ALM) guidelines apply: you should always build and test new configurations in a development model before deploying them to production. Once deployed, these actions integrate natively as machine tasks within Anaplan Workflow.
Question
Why might a user be unable to locate the Optimizer option in their Actions list, and does the engine currently work in Polaris workspaces?
Answer
First, ensure you are logged in with Workspace Administrator rights, as standard end-users cannot see or configure backend actions. Second, the Optimizer is not currently supported in Polaris workspaces, though it is actively on the roadmap. If you are in a Classic workspace and cannot find it, have your administrator contact your Customer Success Business Partner (CSBP) to confirm the feature is provisioned on your tenant.
Question
Is there an efficient way to identify which specific constraints or variables are causing performance and calculation delays?
Answer
Isolating performance bottlenecks mathematically is a complex task across all optimization platforms. The most practical approach in Anaplan is to build interactive Boolean checkboxes on your UX pages. This allows you to dynamically enable or disable specific constraint classes during testing to easily isolate which rules are slowing down the solver.
Question
When building multi-objective models, how should you choose objective and penalty weights (e.g., setting a priority value to 1 vs. 100) to avoid endless trial and error?
Answer
Currently, calibrating abstract weights requires manual trial and error. To simplify this, the best practice is to translate all metrics and penalties into monetary values (costs/profit) so they share a common unit. For a more robust solution, Anaplan is developing a Multi-Objective Function feature for next year's roadmap, which will allow builders to natively define hierarchical priority levels for goals instead of balancing manual weight formulas.