Model Reimport

Overview

Model reimport allows you to reload previously exported models back into MixModeler, recreating them exactly as they were. This feature enables archiving, sharing models across teams, backing up work, and continuing analysis after extended breaks.

Why Reimport Models

Archive and Restore: Save models for long-term storage, reload when needed

Collaboration: Share models with colleagues who can import and explore

Backup: Protect against data loss by regularly exporting models

Versioning: Maintain multiple versions of a model, compare iterations

Migration: Move models between accounts or environments

Reproducibility: Ensure exact model recreation for audit or compliance purposes

What Gets Reimported

Complete Model Recreation

When you reimport, MixModeler recreates:

✓ Model Configuration: Name, KPI, features, date range ✓ Variable Transformations: Adstock, splits, multiplications, all engineering ✓ Model Type: OLS or Bayesian specification ✓ Bayesian Settings: Priors, MCMC settings (if Bayesian) ✓ Fixed Coefficients: Any coefficients that were fixed ✓ Model Results: Coefficients, statistics, fit metrics ✓ Decomposition Groups: Attribution categories and assignments

What's Required

Essential Components in Excel export:

  • Model Info sheet (model metadata)

  • Data sheet (complete dataset)

  • Coefficients sheet (model results)

  • Variable Transformations sheet (if transformations exist)

Optional but Recommended:

  • Model Statistics sheet

  • Decomposition sheets

  • All other sheets (preserved for reference)

Reimport Process

Step-by-Step Instructions

1. Ensure Original Data Loaded

Before importing a model, you must have the original dataset uploaded:

  • Go to Data Upload page

  • Upload the same Excel/CSV file used for original model

  • Or upload the Data sheet from the exported model

  • Verify data loads successfully

Why Required: Model reimport needs the complete dataset to recreate transformations and refit the model

2. Navigate to Model Library

  • Click Model Library in navigation

  • Look for Import Model button at top of page

3. Select Export File

  • Click Import Model

  • Browse and select your exported Excel file (.xlsx)

  • File uploads and validation begins

4. Validation and Confirmation

MixModeler validates the file:

✓ Required sheets present ✓ Data structure intact ✓ Variable names match loaded dataset ✓ Transformations can be recreated

Validation Messages:

  • "✓ Model Info validated"

  • "✓ Data matches loaded dataset"

  • "✓ Coefficients sheet verified"

  • "✓ Transformations validated"

5. Model Recreation

If validation passes:

  • Transformations are recreated

  • Model is refit using original data

  • Results are validated against exported coefficients

  • Model added to library

Success Message: "Model 'ModelName' successfully imported!"

Duration: 3-10 seconds depending on model complexity

Viewing Reimported Model

After successful import:

  1. Model appears in Model Library

  2. Name may have "(Imported)" suffix if duplicate name exists

  3. All features accessible (diagnostics, decomposition, export, etc.)

  4. Results match original export (within numerical precision)

Requirements and Compatibility

File Format

Supported: Excel files (.xlsx) exported from MixModeler

Not Supported:

  • Modified Excel files (with structural changes)

  • Manual Excel files (not exported from MixModeler)

  • CSV files (use Data Upload instead)

  • Other formats (.xls, .ods, etc.)

Data Requirement

Critical: Original dataset must be loaded before importing model

Why: Model export contains model specification and results, but reimport recreates the model from scratch using your data

Workflow:

  1. Upload data (Data Upload page)

  2. Import model (Model Library)

  3. Model recreated using uploaded data

Common Mistake: Trying to import model without loading data first → Fails with error "Dataset not found"

MixModeler Version

Same Version: Best compatibility - export and import on same MixModeler version

Newer Version: Usually compatible - newer versions can import older exports

Older Version: May fail - older versions cannot import exports from newer versions with new features

Recommendation: Keep MixModeler updated for best import compatibility

Validation and Error Handling

Automatic Validation Checks

File Structure:

  • Required sheets exist (Model Info, Data, Coefficients)

  • Column headers match expected format

  • No critical data corruption

Data Consistency:

  • KPI variable exists in loaded dataset

  • Feature variables exist in loaded dataset

  • Variable names match across sheets

  • Transformation source variables exist

Model Specification:

  • Model type valid (OLS or Bayesian)

  • Coefficients sheet has valid data

  • Transformation parameters are complete

  • No circular transformation dependencies

Common Import Errors

Error: "Dataset not loaded"

Cause: Original data not uploaded before import attempt

Solution:

  1. Go to Data Upload

  2. Upload the original dataset

  3. Return to Model Library and retry import

Error: "Variable [X] not found in dataset"

Cause: Dataset loaded doesn't contain all variables from model

Solution:

  1. Verify correct dataset is loaded

  2. Check variable names match exactly (case-sensitive)

  3. Ensure dataset includes all features used in model

Error: "Invalid model specification"

Cause: Model Info sheet corrupted or modified

Solution:

  1. Use unmodified export file

  2. Check Model Info sheet has all required fields

  3. Verify field values are valid

Error: "Transformation failed: [details]"

Cause: Variable transformation cannot be recreated

Solution:

  1. Check Variable Transformations sheet is intact

  2. Verify source variables exist in dataset

  3. Ensure transformation parameters are valid

Error: "Sheet [X] not found"

Cause: Required sheet deleted or renamed

Solution:

  1. Use original, unmodified export file

  2. Ensure no sheets were deleted

  3. Check sheet names weren't changed

Warning Messages

Warning: "Coefficient mismatch detected"

Meaning: Reimported model coefficients differ slightly from exported values

Typical Cause: Numerical precision differences, expected variation

Action: If difference is very small (<0.1%), this is normal. If large, investigate data differences.

Warning: "Date range adjusted"

Meaning: Loaded dataset date range differs from original

Action: Verify correct dataset loaded. Model will use available dates.

Warning: "Decomposition groups not restored"

Meaning: Decomposition group assignments not found in export

Action: Manually recreate decomposition groups if needed

Best Practices for Successful Reimport

Before Export

Document Model:

  • Add notes to Model Info about data source

  • Record data file name and version

  • Document any special considerations

Export with Decomposition:

  • Check "Include Decomposition" when exporting

  • Preserves attribution group assignments

  • Saves time on reimport

Test Export Immediately:

  • After exporting, immediately test reimport

  • Catch any issues while model is still fresh in memory

  • Verify everything imports correctly

Organizing Exports

File Naming Convention:

ModelName_Version_Date.xlsx

Examples:

  • Q4_Campaign_v1_20250104.xlsx

  • Annual_Model_Final_20250115.xlsx

  • Test_Model_Baseline_20241220.xlsx

Storage Structure:

Project_Folder/
  ├── Data/
  │   └── marketing_data_2024.xlsx
  ├── Models/
  │   ├── Exports/
  │   │   ├── baseline_model_v1.xlsx
  │   │   ├── baseline_model_v2.xlsx
  │   │   └── final_model.xlsx
  │   └── Documentation/
  │       ├── model_notes.md
  │       └── diagnostic_reports/

When Sharing Models

Include with Export:

  • Original data file (or export Data sheet)

  • README with import instructions

  • Any documentation about model purpose

  • Note about required MixModeler version

Share Package Structure:

Model_Package.zip
  ├── README.txt (import instructions)
  ├── data_source.xlsx (original data)
  ├── model_export.xlsx (exported model)
  └── documentation.pdf (model notes)

README Template:

Model Import Instructions
=========================

1. Upload data file: data_source.xlsx
2. Go to Model Library
3. Click "Import Model"
4. Select: model_export.xlsx
5. Wait for successful import message

Model Details:
- Name: Q4 Campaign Analysis
- KPI: Revenue
- Time Period: 2024-01-01 to 2024-12-31
- Features: 15 marketing variables
- Model Type: Bayesian

Contact: [your email] with questions

Advanced Reimport Scenarios

Reimporting Bayesian Models

Additional Considerations:

  • Prior specifications preserved and restored

  • MCMC settings restored

  • Can rerun sampling with different settings after import

  • Convergence diagnostics from original export preserved

Workflow:

  1. Import model (recreates structure and priors)

  2. Original MCMC results loaded from export

  3. Optionally rerun Bayesian inference with new settings

  4. Compare original vs new MCMC results

Reimporting with Different Data

Use Case: Apply same model structure to new time period

Not Recommended: Import designed for exact recreation

Better Approach:

  1. Create new model using Model Builder

  2. Copy specifications from exported Model Info

  3. Manually recreate transformations

  4. Fit new model to new data

Workaround:

  1. Import model with original data

  2. Use Model Builder to modify date range

  3. Refit model with new date range

  4. Save as new model version

Partial Imports

Scenario: Want only transformations or configuration, not full model

Workaround:

  1. Import full model

  2. Use as template

  3. Modify as needed in Model Builder

  4. Delete original imported model if not needed

Alternative: Manually recreate transformations using Variable Transformations sheet as reference

Troubleshooting Import Issues

Model Imports But Results Differ

Possible Causes:

  • Different data loaded than original

  • Data file has been updated since export

  • Variable names changed

  • Numerical precision differences

Diagnosis:

  1. Compare loaded data to exported Data sheet

  2. Check variable names match exactly

  3. Verify date range matches

  4. Compare coefficient differences (should be <0.1%)

Solution: Reload original data exactly as it was when model exported

Import Hangs or Times Out

Possible Causes:

  • Very large file (>10 MB)

  • Complex transformations

  • Network issues

  • Browser limitations

Solutions:

  1. Wait 30-60 seconds (large models take time)

  2. Refresh page and retry

  3. Try different browser

  4. Check network connection

  5. Simplify model (if possible) before export

Import Succeeds But Model Errors

Possible Causes:

  • Partial import completed

  • Data integrity issues

  • Model metadata corrupted

Solutions:

  1. Delete partially imported model

  2. Verify export file is complete and valid

  3. Try reimporting with fresh data upload

  4. Contact support if issue persists

Limitations

Cannot Import:

  • Models from other MMM tools

  • Manually created Excel files (not exported from MixModeler)

  • Heavily modified export files

  • Corrupt or incomplete exports

Not Preserved Exactly:

  • Custom notes added after model creation (unless in export)

  • Temporary session state

  • Diagnostic test history (only final results in export)

  • Chart customizations

Requires:

  • Original data file

  • Unmodified export structure

  • Compatible MixModeler version

Next Steps: Review Excel Export Features to understand what gets exported, or explore Sharing Results with Stakeholders for collaboration best practices.

PreviousSharing Results with StakeholdersNextSubscription Plans & Limits

Last updated