Model audit

When running Network Optimization, information about potential issues with the model is gathered and provided in the Model Building node on the Launch Pad solve tracker. The model audit provides information about the following categories, along with recommendations for correcting the condition detected:

  • Model Complexity
    • ABC Analysis - This category identifies conditions where a large percentage of items constitutes a small percentage of input, such as "90% of customers represents only 10% of demand".
    • Scaling Ratio - This category identifies ratios between maximum and minimum values that exceed industry standards, such as the ratio between maximum and minimum demand quantities.
    • Complexity of Possible Sources - This category identifies cases where customers or sites have a large number of possible sources for location-product combinations.
    • Complexity of Variables - This category indicates how many binary or integer values are in the model. Large numbers of these variables can significantly increase model complexity.
  • Cost Structure
    • Steps in Step Costs - This category identifies the average number of steps in step costs where the number of steps exceeds the recommended maximum.
    • Negative Cost - This category identifies negative costs as inputs that can cause the model to be unbounded.
  • Period Ambiguity
    • Cross Period Flows - This category identifies conditions that can cause problematic cross period flows, such as periods with varying lengths and issues of maximum lead time.

    For each category where issues are detected, you can click View Details to see a report of the conditions that have been identified. You can then use this information to make corrections as needed.

    For the majority of categories, if a potential issue is identified, the category displays a warning icon. Two of the Model Complexity items (Complexity of Possible Sources and Complexity of Variables) will show success even if there are issues identified, since their thresholds need to be refined to prevent most models from displaying solver warnings.

Last modified: Wednesday May 15, 2024

Is this useful?