MMM - Model Comparison

Updated by Eddie Chou

Overview

The "Compare Performance" page enables users to compare Return on Ad Spend (ROAS) or Cost Per Acquisition (CPA) results across different marketing mix models (MMM) to address a common question: How has performance changed across models over time?

With this new chart, users can evaluate the performance of various channels and campaigns across up to four selected models, helping them better understand model differences and refine their marketing strategies.

Key Features

  1. Performance By Feature By Model Chart
    • Compares ROAS or CPA for each feature across selected models.
    • Provides a clear visual representation of performance differences.
  2. Customizable Model Selection
    • Users can compare up to four models at a time.
    • By default, the page selects the most recent four models, but users can de-select and choose others.
  3. Chronological Sorting
    • Models are displayed chronologically from the least recent to the most recent.

How to Use the "Compare Performance" Page

  1. Navigate to the Page
    Access the "Compare Performance" page from the MMM section of the platform.
  2. Select Models to Compare
    • By default, the four most recent models are selected.
    • Use the dropdown menu to customize your selection.
  3. Review the Chart
    • Hover over each bar to see specific ROAS/CPA values for a feature within a model.
    • Use this to analyze feature-level performance trends across models.
  4. Interpret Results
    • Compare how performance metrics like ROAS vary for each feature.
    • Identify high-performing features evaluate shifts caused by shifts in marketing and model updates.

Use Case Examples

  • Assessing Model Changes: See how recent updates to your marketing mix have impacted channel-level ROAS or CPA.
  • Performance Diagnostics: Evaluate discrepancies between models to identify key drivers of performance changes.

Troubleshooting

If you notice that two separate features in the chart should actually be combined for an accurate comparison, this may be due to a mismatch in the way features were named between models.

To resolve this issue, please contact your Rockerbox team. We can update the feature mappings to ensure that these features are combined correctly for comparison purposes.


How did we do?