Export to Excel

Save Complete Model Results and Enable Reimport

Overview

Excel export creates a comprehensive workbook containing all model details, statistics, transformations, and parameters. These files serve as permanent records, can be shared with stakeholders, and - importantly - can be reimported to MixModeler to perfectly recreate the model.

Purpose: Save complete model analysis to Excel for archiving, sharing, and reimporting

Access: Model Library → Click Export button for any model

What Gets Exported

Complete Model Package

An exported Excel file contains multiple sheets with all information needed to understand and recreate your model:

Sheet 1: Model Summary

Model metadata:

• Model name and KPI

• Model type (OLS or Bayesian)

• Number of variables

• R² and Adjusted R²

• Observations used

• Creation and modification dates

Key statistics:

• F-statistic and p-value (OLS)

• Log-likelihood

• AIC and BIC

• Condition number

Sheet 2: Variable Coefficients

For each variable:

• Variable name

• Coefficient (β value)

• Standard error

• T-statistic

• P-value

• 95% Confidence intervals

Sorted by:

• Constant term first

• Then all variables alphabetically

Sheet 3: All Transformations

Critical sheet for reimport:

• Lists every transformed variable

• Transformation type (adstock, curve, weighted, etc.)

• All parameters (adstock rate, alpha, power, etc.)

• Base variable name

• Enables perfect recreation on reimport

Example rows:

• TV_Spend_adstock_70: type=adstock, rate=70, base=TV_Spend

• TV_Spend_adstock_70|ICP_ATAN_a0.6_power1.8: type=ICP, alpha=0.6, power=1.8, base=TV_Spend_adstock_70

Sheet 4: Model Data

Complete dataset used:

• Date/Observation column

• KPI column

• All independent variables (including transformed)

• Exactly what was used for regression

Use cases:

• Verify data correctness

• Perform external analysis

• Share with colleagues

• Archive data snapshot

Sheet 5: Diagnostics (if run)

Statistical test results:

• Residual normality tests

• Autocorrelation tests

• Heteroscedasticity tests

• Multicollinearity (VIF)

• Influential points

• Test statistics and p-values

Sheet 6: Decomposition (if run)

Contribution analysis:

• Date/observation column

• Contribution from each variable over time

• Baseline contribution

• Total predicted values

Use cases:

• ROI analysis

• Budget allocation

• Performance tracking

• Stakeholder reporting

Additional Sheets (Bayesian models)

If Bayesian inference was run:

• Bayesian Coefficients: Posterior means, std dev, credible intervals

• MCMC Diagnostics: R-hat, effective sample size

• Model Metrics: WAIC, LOO, convergence statistics

• Prior Settings: Prior distributions used

How to Export

Step 1: Set Export Path

Before exporting, specify where to save the file:

1. In Model Library table, locate your model

2. Click in the Export Path column

3. Enter full folder path (e.g., C:\Users\YourName\Documents\MMM\)

4. Press Enter to save

Path requirements:

• Must end with \ or /

• Folder must exist

• Must have write permissions

• Path persists across sessions

Filename: MixModeler automatically appends ModelName.xlsx

Step 2: Click Export Button

Click the 📊 Export button in the Actions column for your model.

For Bayesian models: A warning dialog appears:

"⚠️ Bayesian Model Export Warning You are about to export a Bayesian model. To ensure all your analysis results are properly saved, please make sure you have run the Bayesian inference before exporting. Do you want to continue with the export?"

• Click OK to proceed

• Click Cancel to go run Bayesian inference first

Step 3: Wait for Export

Progress indication:

• Loading indicator appears

• "Exporting..." message shown

• Usually takes 5-30 seconds depending on model size

Large models (>50 variables):

• May take up to 1 minute

• Be patient, don't click multiple times

Step 4: Confirm Success

Success message appears: "Model 'YourModelName' exported successfully to C:\Path\YourModelName.xlsx"

File is now available at specified location - ready to open, share, or archive.

Excel Export Use Cases

1. Archiving and Documentation

Purpose: Permanent record of model at specific point in time

Process:

1. Complete model development

2. Run all diagnostics

3. Run decomposition

4. Export with descriptive filename

5. Store in project archive folder

Filename convention: Sales_Model_Final_2024_12_15.xlsx

Benefits:

• Preserve exact model state

• Audit trail for compliance

• Historical reference

• Prevents data loss

2. Sharing with Stakeholders

Purpose: Present results to non-technical audience

Process:

1. Export model

2. Open Excel file

3. Add executive summary sheet

4. Format decomposition for readability

5. Create charts in Excel

6. Share file via email or shared drive

Key sheets to highlight:

• Model Summary (overall stats)

• Variable Coefficients (channel effects)

• Decomposition (contribution over time)

Benefits:

• Everyone can open Excel

• Stakeholders can explore at their own pace

• Supports data-driven discussions

3. Collaboration with Colleagues

Purpose: Enable colleague to continue your work

Process:

1. Export your model

2. Share file with colleague

3. Colleague imports file into their MixModeler

4. Model perfectly recreated with all transformations

5. Colleague can modify and iterate

Benefits:

• Perfect model replication

• No knowledge loss

• Enables teamwork

• Version control through files

4. Backup and Recovery

Purpose: Protect against data loss

Process:

1. Export all important models weekly

2. Store in cloud storage (OneDrive, Google Drive)

3. If local system fails, reimport from backups

4. Resume work immediately

Benefits:

• Disaster recovery

• Peace of mind

• No work lost

• Quick restoration

5. Compliance and Audit

Purpose: Meet regulatory or internal audit requirements

Process:

1. Export final production models

2. Add date to filename

3. Store in audit-compliant folder

4. Include in compliance documentation

5. Retain for required period

Benefits:

• Audit trail

• Reproducibility

• Regulatory compliance

• Legal protection

Reimporting Excel Files

Why Reimport?

Common scenarios:

• Restore from backup

• Move to different computer

• Share model with colleague

• Recover deleted model

• Replicate external analysis

Reimport Process

Step 1: Prepare Excel File

• Ensure file was exported from MixModeler

• Do not modify sheet structure

• Data modifications allowed (refresh data)

Step 2: Navigate to Import

• Go to Model Library or designated import section

• Click "Import Model from Excel"

Step 3: Select File

• Browse to Excel file location

• Select .xlsx file

• Click Open

Step 4: MixModeler Recreates Model

What happens automatically:

1. Reads Model Summary sheet

2. Loads data from Model Data sheet

3. Reads All Transformations sheet

4. Recreates ALL transformed variables:

   • Adstock variables

   • Curve variables

   • Weighted variables

   • AVO variables

   • Any other transformations

5. Sets up model specification

6. Runs regression

7. Model appears in Model Library

Result: Perfect recreation of exported model, ready to use.

Reimport Capabilities

✅ Recreates transformations perfectly: All adstock rates, curve parameters, weighted variable formulas restored

✅ Preserves model type: OLS or Bayesian setting maintained

✅ Restores variable selection: Exact same variables included in model

✅ Maintains metadata: Model name, KPI, creation info

Reimport Limitations

❌ Cannot import arbitrary Excel files: Must be MixModeler export format

❌ Historical diagnostics not restored: Must re-run diagnostics after import

❌ Decomposition not automatically recalculated: Must re-run decomposition (uses imported data)

❌ Bayesian posterior samples not included: Must re-run MCMC inference (priors are imported)

Export Best Practices

File Organization

Folder structure:

MMM_Project/
├── Data/
│   └── raw_data.xlsx
├── Models/
│   ├── Sales_Model_v1_2024_12_01.xlsx
│   ├── Sales_Model_v2_2024_12_05.xlsx
│   └── Sales_Model_Final_2024_12_15.xlsx
└── Reports/
    └── Executive_Summary.pptx

Benefits:

• Easy to find files

• Clear version history

• Organized workflow

Naming Conventions

Include in filename:

1. Model purpose: Sales_Model, Revenue_Analysis

2. Version: v1, v2, Final

3. Date: YYYY_MM_DD format

4. Optional descriptor: With_Curves, Digital_Focus

Examples:

• Sales_MMM_Final_2024_12_15.xlsx

• Revenue_Model_v2_Curves_2024_12_10.xlsx

• Conversions_Digital_Focus_2024_Q4.xlsx

Export Frequency

When to export:

✅ After major milestones:

• Completed variable selection

• Finalized transformations

• Passed all diagnostics

• Final production model

✅ Before risky changes:

• Major specification changes

• Trying experimental approaches

• Switching methodologies

✅ Regular backups:

• Weekly during active development

• End of each work session

• Before sharing with others

❌ Don't export:

• Every minor tweak

• Failed experiments (export once if needed, then delete)

• Models with errors

Version Control

Use export as version control:

1. Export v1 after initial build

2. Make changes in MixModeler

3. Export v2 if improvements successful

4. Keep both until certain v2 is better

5. Delete v1 file (or archive) once v2 proven

Alternative: Keep only "current" and "final"

• Current: Working version during development

• Final: Production model when complete

Troubleshooting

"Export path required"

Cause: No export path specified for model

Solution:

1. Click in Export Path column for model

2. Enter folder path: C:\Users\YourName\Documents\

3. Press Enter

4. Try export again

"Export failed - path not found"

Cause: Folder in export path doesn't exist

Solution:

• Create the folder first in Windows Explorer

• Or use existing folder path

• Verify path is correct (no typos)

"Export failed - permission denied"

Cause: No write permissions for specified folder

Solution:

• Choose different folder where you have permissions

• Check if file is open (close it)

• Try Desktop or Documents folder

Exported file shows different R² than app

Cause: Minor numerical differences in regression calculation

Explanation:

• Should be <0.1% different

• Due to rounding or computation order

• Not a problem if very close

Solution:

• If difference >1%, contact support

• Otherwise, ignore minor variation

Reimport fails with error

Possible causes:

• Excel file modified improperly

• Sheet structure changed

• File corrupted

Solutions:

• Use original unmodified export

• If you modified data, check column names match

• Verify All Transformations sheet intact

• Try reimporting original export first to test

Bayesian model export missing posterior results

Cause: Bayesian inference not run before export

Solution:

• Go to Bayesian Model Interface

• Run MCMC inference

• Wait for completion

• Export again

Key Takeaways

• Excel export creates complete, shareable model package

• Includes all sheets: summary, coefficients, transformations, data, diagnostics

• Set export path before clicking Export button

• Files can be reimported to perfectly recreate model

• Use for archiving, sharing, backup, and compliance

• Name files descriptively with date and version

• Export after milestones and before risky changes

• Reimport capability enables collaboration and recovery

• All transformations automatically recreated on reimport