Quick Start Guide
🚀 Build Your First MMM Model in 15 Minutes
This quick start guide will walk you through creating your first Marketing Mix Model from data upload to actionable insights. By the end, you'll have a working model showing which marketing channels drive your business results.
What You'll Need
Data Requirements
Excel File Format
.xlsx or .xls format (CSV also supported)
Minimum 26 weeks of data (52+ weeks recommended for reliable results)
Time series structure: One row per time period (weekly or monthly)
First column MUST be named "Observation" containing your time periods
Required Variables
1 KPI variable: Sales, revenue, conversions, brand awareness, etc.
2+ Marketing variables: TV spend, digital spend, search spend, social spend, etc.
Optional control variables: Seasonality indicators, price, weather, competitor activity
Example Dataset Structure
✅ Correct Format:
❌ Incorrect Format:
Why it's wrong: First column must be named "Observation", not "Week"
Step 1: Data Upload (2 minutes)
Navigate to Data Upload Page
Open MixModeler application
Click "Data Upload" in the left sidebar
You'll see the upload interface with two sections
Download Template (Optional but Recommended)
First time users should start with our template:
Click the "Download Excel Template" button
Open the template in Excel
Replace sample data with your actual marketing data
Keep the column structure and "Observation" column name
Save your file
Template includes:
Pre-formatted date column (Observation)
Example marketing channels (TV, Digital, Social, Search, Radio)
Sample KPI column (Sales)
Example control variables (Holiday, Price)
Upload Your Data
Click "Choose File" or drag your Excel file into the upload zone
Wait for the file to be validated (usually 5-10 seconds)
Review the Data Summary that appears:
Number of variables detected
Number of observations (time periods)
Date range of your data
Preview of first 10 rows
🔍 Verify Your Upload:
Check that all columns were imported correctly
Ensure dates are in the right format
Confirm no missing values in critical variables
Review variable names for clarity
⚠️ Common Upload Issues:
"Observation column not found": Rename your date/period column to "Observation"
"Insufficient observations": You need at least 26 weeks of data
"Missing values detected": Fill in gaps or remove incomplete rows
"Non-numeric data": Ensure all data columns contain numbers only
Step 2: Create Your First Model (3 minutes)
Navigate to Model Library
Click "Model Library" in the left sidebar
Click the "+ Create New Model" button
A dialog box will appear
Configure Basic Model
Model Setup:
Model Name: Enter a descriptive name
Good: "Sales_MMM_Q1_2024" or "Revenue_Model_Full_Year"
Avoid: "Model1" or "Test"
Select KPI Variable:
Choose your dependent variable from the dropdown
This is what you want to predict/explain
Examples: Sales, Revenue, Conversions, Brand_Awareness
Click "Create Model"
Your model is now created with just the baseline (intercept). Next, we'll add marketing variables.
Step 3: Add Marketing Variables (5 minutes)
Navigate to Model Builder
Click "Model Builder" in the left sidebar
Select your newly created model from the dropdown at the top
You'll see two tabs: "Current Variables" and "Available Variables"
Add Your First Marketing Channels
Switch to "Available Variables" tab:
Search or browse for your marketing variables
Examples: TV_Spend, Digital_Spend, Search_Spend
Select 3-5 major marketing channels (start simple)
Focus on your largest spend channels
Include channels you actually want to optimize
Configure Adstock for Media Variables:
For each media variable, set adstock rate
TV/Video: Start with 70% (high carryover)
Digital/Search: Start with 40% (moderate carryover)
Radio: Start with 60% (moderate-high carryover)
Social Media: Start with 30% (lower carryover)
Click "Add to Model" button
Initial Model Results
Review the updated model:
R-squared: How much variance your model explains (aim for above 0.60)
Coefficients: Impact size of each variable
T-statistics: Statistical significance (bold values are significant)
What Good Results Look Like:
R-squared above 60% (0.60)
Marketing variables with positive coefficients (green)
T-statistics above 1.96 (shown in bold)
Coefficients that make business sense
If Results Look Poor:
R-squared below 50%: Add more relevant variables
Negative coefficients on media: Check data quality or try different adstock rates
Low t-statistics: Variable might not be impactful, consider removing
Step 4: Run Model Diagnostics (2 minutes)
Navigate to Model Diagnostics
Click "Model Diagnostics" in the left sidebar
Select your model from the dropdown
Click "Continue to Diagnostics"
Run Basic Tests
Select Tests to Run:
✅ Residual Normality (check if errors are normally distributed)
✅ Autocorrelation (check for patterns in residuals)
✅ Heteroscedasticity (check if error variance is constant)
✅ Multicollinearity (check for correlated variables)
Click "Run Selected Tests"
Interpret Results
Green Checkmarks (✓) = Good
Test passed, assumption satisfied
Model is statistically sound
Red Warnings (⚠) = Issues Detected
Click "View Details" for more information
Some violations are acceptable in MMM
Focus on fixing critical issues first
Critical Issues to Address:
VIF above 10: Remove one of the correlated variables
Severe autocorrelation: Add lagged variables or time trends
Many influential points: Check data quality
Acceptable Issues (Often Normal in MMM):
Mild autocorrelation (Durbin-Watson 1.5-2.5)
Slight non-normality of residuals
Minor heteroscedasticity
Step 5: Set Up Contribution Groups (2 minutes)
Navigate to Contribution Groups
Click "Contribution Groups" in the left sidebar
Select your model from the dropdown
You'll see all variables in your model
Organize Variables into Business Groups
Assign each variable to a logical group:
Example Groupings:
Base: Intercept/constant (baseline performance)
Media: TV_Spend, Digital_Spend, Radio_Spend, Social_Spend
Seasonality: Holiday, Month_Indicators
Price: Price_Variable, Promotion_Indicator
External: Competitor_Activity, Economic_Indicator
How to Assign:
Click the "Group" dropdown for each variable
Select or type a group name
Use consistent names (e.g., always "Media" not sometimes "Marketing")
Choose Colors for Each Group:
Once groups are assigned, a color picker appears
Choose distinct, professional colors
Recommended: Blue for Media, Red for Price, Green for Seasonality
Click "Save Groups"
Step 6: Run Decomposition Analysis (2 minutes)
Navigate to Decomposition
Click "Decomposition" in the left sidebar
Select your model from the dropdown
Click "Run Decomposition"
Understand Your Results
Main Decomposition Chart:
Stacked bars: Each color shows a group's contribution over time
Black line: Actual KPI values from your data
Bar height: Total predicted value
What to Look For:
Largest Contributors:
Which groups have the tallest bars?
These are your primary growth drivers
Consistency vs. Variability:
Steady bars (Base): Reliable baseline performance
Variable bars (Media): Campaign-driven spikes
Timing Patterns:
When did Media contributions spike?
Do spikes align with known campaign periods?
Drill Down into Specific Groups
Analyze Media Channels in Detail:
Select "Media" from the group dropdown
Click "Run Group Decomposition"
See individual channel contributions (TV, Digital, Radio, etc.)
Identify Top Performers:
Which channel has the largest contribution?
Which channels show most consistent performance?
Where are the spikes and why?
Step 7: Calculate Channel ROI (1 minute)
Quick ROI Calculation
For each marketing channel, calculate:
Example:
TV Total Contribution from decomposition: $500,000
TV Total Spend: $200,000
TV ROI = ($500,000 / $200,000) - 1 = 1.5 or 150%
Interpretation:
For every $1 spent on TV, you generated $2.50 in incremental sales
Net return: $1.50 per dollar invested
Compare Across Channels:
Which channels have highest ROI?
Which are underperforming?
Where should you increase/decrease investment?
Step 8: Export Your Results (1 minute)
Save Your Model
Go back to Model Library
Find your model in the table
Click the "Export" (📁) button
Your model will be saved as an Excel file containing:
Model coefficients and statistics
Diagnostic test results
Decomposition data by group
Variable transformation details
Excel Export Includes:
Model Info sheet: Basic model details
Coefficients sheet: All variable coefficients and t-stats
Diagnostics sheet: Test results summary
Decomposition sheet: Contribution data by time period
Transformations sheet: Applied transformations (adstock, curves, etc.)
Next Steps: Improving Your Model
Quick Wins (Do These Next)
1. Add More Variables (if needed)
Include additional marketing channels you're spending on
Add seasonality indicators (monthly dummies, holiday flags)
Consider competitor activity or economic indicators
2. Apply Saturation Curves
Navigate to Variable Workshop
Create saturation curves for media variables
Use Curve Testing to find optimal parameters
Models diminishing returns more realistically
3. Test Different Adstock Rates
Use Variable Testing page
Test multiple adstock rates (30%, 50%, 70%)
Choose rates with highest t-statistics
More accurate carryover modeling
4. Upgrade to Bayesian
In Model Builder, click "Convert to Bayesian"
Get credible intervals and uncertainty quantification
More robust for limited data
Industry standard for production models
Advanced Features to Explore
Variable Workshop:
Create lead/lag variables
Build weighted variable combinations
Apply AVO (Average Value Optimization) transformations
Generate interaction terms
Curve Testing:
Test S-shape vs Concave curves
Optimize curve parameters (alpha, power)
Visualize transformation effects
Create variables with optimal saturation
Variable Testing:
Pre-screen variables before adding to model
Run Granger Causality tests
Identify multicollinearity issues
Optimize variable selection
Model Comparison:
Build multiple model versions
Compare performance side-by-side
Test different variable combinations
Find optimal model specification
Common Quick Start Pitfalls
Data Issues
❌ Wrong Column Name
Problem: First column not named "Observation"
Fix: Rename your date/period column to exactly "Observation"
❌ Insufficient Data
Problem: Less than 26 weeks of data
Fix: Gather more historical data (52+ weeks ideal)
❌ Missing Time Periods
Problem: Gaps in weekly/monthly data
Fix: Fill gaps or indicate missing periods appropriately
❌ Inconsistent Units
Problem: Mixing weekly and monthly data
Fix: Ensure consistent time periods throughout
Model Building Issues
❌ Too Few Variables
Problem: Model only includes 1-2 marketing channels
Fix: Include all major marketing channels
❌ No Adstock Applied
Problem: Media variables used without carryover effects
Fix: Apply appropriate adstock rates (40-70% typically)
❌ Ignoring Diagnostics
Problem: Not checking if model assumptions are met
Fix: Always run diagnostics and address critical issues
❌ Over-Interpretation
Problem: Treating estimates as exact truth
Fix: Remember MMM provides estimates with uncertainty
Success Checklist
Before considering your first model complete, ensure:
✅ R-squared above 60% (preferably above 70%)
✅ Marketing variables have positive coefficients (make business sense)
✅ Most variables statistically significant (t-stat above 1.96)
✅ Adstock applied to all media variables (carryover effects captured)
✅ Diagnostic tests mostly passing (especially multicollinearity check)
✅ Decomposition results align with business knowledge (reality check)
✅ Contribution groups logically organized (business-friendly)
✅ Results exported and saved (documentation)
Pro Tips for Quick Start Success
1. Start Simple, Add Complexity
Begin with 3-5 major channels
Get a working model first
Add sophistication incrementally
2. Focus on Major Spend
Include channels where you spend the most
Don't worry about small channels initially
Cover 80% of your marketing budget
3. Trust Your Business Knowledge
If results don't make sense, investigate why
Data quality issues are common
Model should align with experience
4. Document Everything
Note your data sources and assumptions
Keep track of model versions
Record why you made specific choices
5. Iterate Quickly
Build multiple versions
Test different approaches
Learn from each iteration
Getting Help
Need Assistance?
Check our detailed documentation sections
Review specific feature guides
Contact support: support@mixmodeler.com
Watch tutorial videos (coming soon)
Common Questions:
Data Format Requirements
Understanding Adstock
Interpreting Diagnostics
Troubleshooting Guide
Congratulations! You've built your first Marketing Mix Model. This foundation enables you to understand which marketing channels drive results and make data-informed budget allocation decisions. Explore the advanced features to refine your models and extract even deeper insights.
Last updated