Adjustment Parameters
What Are Adjustment Parameters?
Adjustment parameters modify how variable contributions are calculated in decomposition analysis. They allow you to transform contributions to better highlight incremental effects or focus on specific aspects of performance.
Purpose: Apply optional mathematical adjustments to variable contributions during decomposition to isolate incremental effects.
Available Adjustments
MixModeler offers three adjustment options for each variable:
None
Original contribution
No change, use as calculated
Min
Contribution - Min(Contribution)
Removes minimum, shows only incremental
Max
Contribution - Max(Contribution)
Removes maximum, rare use case
None Adjustment (Default)
What It Does
Formula: Contribution = Coefficient × Value
Uses the contribution exactly as calculated by the model, with no modifications.
When to Use
Default choice for most variables:
Shows actual contribution values
Maintains interpretability
Sum of contributions equals predicted value
Most intuitive for stakeholders
Recommended for:
All variables in initial analysis
When showing total impact
When contributions are already meaningful
For most business reporting
Example
Min Adjustment
What It Does
Formula: Adjusted = Original - Min(Original)
Subtracts the minimum contribution value across all time periods from each period's contribution.
Effect:
Shifts all contributions up
Minimum period becomes zero
Shows only incremental effects above minimum
When to Use
Isolating Incremental Effects:
When variable has a consistent baseline contribution
To show variability rather than absolute levels
When comparing variables with different baselines
Highlighting Changes:
Focuses on peaks and changes
Removes constant baseline
Emphasizes campaign effects
Typical Use Cases:
Always-on channels with baseline spend
Variables with constant effects to show incremental only
Comparative analysis where baselines differ
Example
Important Note
The subtracted amount is added to the Base group:
When Min adjustment removes $50,000 from TV in the example above, that $50,000 is added to the Base group in each time period.
Why: This maintains the total - the sum of all group contributions must still equal the predicted value.
Visual Impact
In decomposition charts:
Adjusted variable shows only the incremental portion
Base group increases by the removed amount
Total contribution (bar height) remains unchanged
Easier to see variability in the adjusted variable
Max Adjustment
What It Does
Formula: Adjusted = Original - Max(Original)
Subtracts the maximum contribution value across all time periods from each period's contribution.
Effect:
Shifts all contributions down
Maximum period becomes zero
Creates negative values for most periods
When to Use
Rarely used in practice:
Special analytical scenarios
When showing distance from peak
Typically not needed for standard MMM
Possible Use Cases:
Gap to peak analysis - showing how far each period is from maximum
Negative framing - emphasizing under-performance
Specific analytical requirements - unusual business questions
Example
Note: Max adjustment is rarely recommended for standard MMM analysis.
Setting Adjustments
In Contribution Groups Page
For individual variables:
Locate variable in the table
Click Adjustment dropdown
Select: None, Min, or Max
Repeat for other variables
For multiple variables (bulk):
Check boxes next to variables
Use bulk adjustment dropdown
Apply same adjustment to all selected
Efficient for related variables
Best Practices
Default to None:
Start with no adjustments
Only apply adjustments with clear rationale
Most analyses don't need adjustments
Document Decisions:
Record which variables are adjusted
Explain why adjustment was applied
Note in reports and presentations
Consistent Application:
Apply same adjustment logic across similar variables
If adjusting TV, consider adjusting Radio similarly
Maintain consistency across models
Common Use Cases
Use Case 1: Baseline Removal
Scenario: Channel has always-on spend with consistent baseline contribution
Example: Email marketing runs every week with steady $20,000 contribution, but campaigns add incremental lift
Solution:
Apply Min adjustment to Email variable
Shows only incremental campaign effects
Baseline moves to Base group
Result: Easier to see campaign performance
Use Case 2: Comparing Channels with Different Baselines
Scenario: Want to compare variability across channels with different spending levels
Example:
TV: $100K-$150K contribution
Radio: $20K-$30K contribution
Without adjustment: TV dominates visually
With Min adjustment on both:
TV: $0-$50K incremental
Radio: $0-$10K incremental
Result: Can compare variability on similar scale
Use Case 3: Highlighting Promotional Spikes
Scenario: Variable has steady contribution except during promotional periods
Example: Base price effect is constant, but promotions create spikes
Solution:
Apply Min adjustment to promotion variable
Shows only the promotional lift
Baseline effect moves to Base
Result: Promotional impact is isolated and visible
Impact on Interpretation
What Changes
With Adjustments:
Variable contribution values shift
Base group contribution increases
Visual emphasis changes in charts
Easier to see variability
What Stays the Same:
Total predicted value (sum of all contributions)
Model coefficients
Statistical significance
Overall model fit
Interpretation Differences
None Adjustment: "TV contributed $100,000 to sales this week"
Min Adjustment: "TV contributed $50,000 above its minimum baseline this week"
Both statements can be true - they're just different ways to frame the contribution.
Technical Details
Mathematical Properties
Additive Property:
Adjustments are additive transformations
Don't change relative ordering
Preserve differences between periods
Total Preservation:
The total is maintained because adjustments shift values between groups, not change the total.
Calculation Order
Calculate base contributions: Coefficient × Value for each variable
Apply adjustments: Subtract min or max if specified
Sum to groups: Add adjusted contributions within each group
Add to Base: Subtracted amounts added to Base group
Verify total: Sum across all groups equals predicted
When NOT to Use Adjustments
Avoid adjustments when:
First-time analysis: Start with None to understand actual contributions
Executive reporting: Simpler to explain unadjusted values
ROI calculation: Need actual contributions for accurate ROI
Unclear rationale: Don't adjust "just because"
Most variables: Typically only 0-2 variables need adjustment
Keep it simple: Most successful MMM analyses use no adjustments at all.
Communicating Adjusted Results
For Technical Audiences
Be explicit about adjustments:
"Min-adjusted TV contribution shows incremental impact above $50K baseline"
"Values reflect contribution above minimum observed level"
Include footnote explaining adjustment
For Business Stakeholders
Use plain language:
"This shows the extra sales from TV beyond its typical baseline"
"We've isolated the campaign lift from always-on effects"
Focus on business meaning, not technical details
In Reports
Label clearly:
Chart title: "Incremental Contribution (Min-Adjusted)"
Axis label: "Contribution Above Baseline"
Legend note: "TV (Incremental)"
Provide context:
Explain what was adjusted and why
Note that totals still match predicted value
Clarify interpretation
Adjustment Strategy Checklist
Before applying adjustments, ask:
[ ] Is there a clear business reason for this adjustment?
[ ] Will this make the analysis more or less interpretable?
[ ] Can I explain this adjustment to stakeholders?
[ ] Does this serve the analytical goal?
[ ] Have I documented the adjustment rationale?
[ ] Is this consistent with how I've adjusted similar variables?
If unsure → Use None adjustment
Examples
Example 1: E-Commerce Always-On Email
Situation:
Email runs every week
Steady baseline of 5,000 opens
Campaigns create spikes
Original Contribution:
With Min Adjustment:
Benefit: Campaign lift is isolated and visible
Example 2: Seasonal Baseline
Situation:
Store traffic has seasonal baseline
Marketing amplifies seasonal effect
Original Contribution:
With Min Adjustment:
Benefit: Shows incremental marketing impact above seasonal baseline
Example 3: Price Elasticity
Situation:
Base price effect is constant
Price changes create deviations
Decision: Keep as None
Actual price impact is meaningful
Don't want to remove baseline
Total effect is what matters for pricing decisions
Rationale: Not all situations benefit from adjustments
Saving Adjustments
How to Save:
Set adjustments for all desired variables
Click "Save Groups" button
Adjustments save with group configuration
Applied automatically in decomposition
Persistence:
Saved with model
Used in all future decomposition runs
Can be changed anytime
Update applies to future analyses only
Next Steps
After setting adjustment parameters:
Save group configuration
Navigate to Decomposition page
Run decomposition analysis
Review charts to verify adjustments work as intended
Iterate if needed
Last updated