Getting started with Cassandra
      Back to home
      1. Knowledge base
      2. Getting started with Cassandra

      How to Refresh Your Model in Cassandra

      Regularly refreshing your model in Cassandra ensures that your predictions remain accurate, incorporating the latest data and adapting to changing market conditions. We recommend a monthly model refresh to maintain optimal performance.

      This guide walks you through the steps to refresh your model efficiently.

      1. Why Refresh Your Model?

      • Keeps Predictions Accurate – Ensures that forecasts reflect the most recent market conditions.

      • Incorporates New Data – Uses the latest historical data to improve modeling precision.

      • Enhances Budget Allocation – Adjusts recommendations based on updated performance insights.

      • Identifies Shifts in Marketing Effectiveness – Helps track trends in diminishing returns and channel efficiency.

      2. Steps to Refresh Your Model

      Step 1: Uploading New Data

      1. Navigate to the Model Overview page.

      2. Click on Refresh Model.

      3. Upload the latest historical data covering the previous month.

      4. Verify that all required fields are correctly mapped before submission.



      Step 2: Retraining the Model

      • Cassandra will automatically retrain the model using the newly uploaded data.

      • This process typically takes up to 6 hours, depending on the dataset size and complexity.

      Step 3: Reviewing Updated Results

      1. Once the refresh is complete, review the updated model performance.

      2. Compare the new ROI projections against the previous month to assess any significant changes.

      3. Identify potential optimization opportunities based on the latest performance trends.

      3. Adjusting Model Settings (Optional)

      • If needed, modify your model’s configuration before retraining:

        • Add or Remove Channels to refine marketing mix analysis.

        • Adjust Prior Constraints for Bayesian models to fine-tune attributions.

        • Update External Data Inputs if new calibration data is available.

      4. Best Practices for Model Refreshing

      • Refresh Monthly – A consistent update cycle ensures the most accurate insights.

      • Validate Data Before Uploading – Ensure no missing or inconsistent data points.

      • Monitor Changes in Model Output – Look for shifts in attributions or confidence intervals.

      • Use Incrementality Testing for Calibration – Apply experimental findings to refine predictions.

      • Compare Model Versions – Track how performance metrics evolve over time to adjust strategies accordingly.

      5. Need Assistance?

      If you have questions or need help with your model refresh, reach out to the Cassandra team.

      Keeping your model up to date ensures your marketing strategies remain data-driven and optimized for maximum impact.