Main Decomposition Chart

Chart Overview

The main decomposition chart is the centerpiece of your MMM analysis, visually showing how different marketing and non-marketing factors contribute to your KPI over time.

Purpose: Visualize group-level contributions over time using stacked bars and show model fit with actual vs. predicted lines.

Two-Chart Layout

Chart 1: Actual vs Predicted (Line Chart)

Top chart shows model fit quality:

Black Solid Line - Actual KPI:

  • Your real historical data

  • Ground truth

  • What actually happened

Red Dashed Line - Predicted KPI:

  • Model's prediction

  • Sum of all contributions

  • What the model says should happen

Gap Between Lines:

  • Small gaps: Good model fit

  • Large gaps: Model missing something

  • Systematic patterns: Specification issues

Chart 2: Contribution Breakdown (Stacked Bar Chart)

Bottom chart shows the decomposition:

Stacked Bars:

  • One bar per time period

  • Height = Predicted KPI value

  • Each segment = one group's contribution

Color Coding:

  • Each group has unique color

  • Consistent across all time periods

  • Legend shows group-color mapping

Stacking Order:

  • Base at bottom (usually gray)

  • Other groups stack on top

  • Negative contributions below zero

Reading the Stacked Bar Chart

Bar Height and Composition

Total Bar Height:

  • Equals predicted KPI value for that period

  • Should approximately match actual (black line)

  • Shows total explained value

Segment Heights:

  • Each colored segment = one group's contribution

  • Taller segments = larger contribution

  • Height varies over time based on activity

Example Interpretation:

Week 1 Bar (Height: $100,000):
├── Base: $40,000 (40%)
├── Media: $35,000 (35%)
├── Price: -$5,000 (-5%)
├── Promotions: $20,000 (20%)
└── Seasonality: $10,000 (10%)

Color Segments

Base (Usually Gray):

  • Foundation/baseline

  • Constant term + baseline variables

  • Steady across periods

  • Represents organic demand

Marketing Groups (Blues, Teals):

  • Paid media channels

  • Variable over time

  • Spikes during campaigns

  • Incremental impact

Price (Red):

  • Can be positive or negative

  • Price decreases → positive (boost sales)

  • Price increases → negative (reduce sales)

  • Trade-off visible

Promotions (Orange):

  • Temporary spikes

  • Promotional periods visible

  • Event-driven contributions

  • Clear lift patterns

Seasonality (Green):

  • Regular patterns

  • Cyclical contributions

  • Predictable variation

  • Annual rhythms

Identifying Patterns

Campaign Spikes:

  • Marketing segments grow during campaigns

  • Align bars with known campaign periods

  • Measure incremental lift

Seasonal Cycles:

  • Regular up-and-down patterns

  • Quarterly or monthly rhythms

  • Align with business calendar

Baseline Trends:

  • Gradual changes in Base height

  • Growing or declining baseline

  • Long-term trajectory

Promotional Events:

  • Sharp spikes in promotion segment

  • Align with known sale events

  • Black Friday, holiday periods

Price Changes:

  • Step changes in price segment

  • Positive = price cuts boost sales

  • Negative = price increases reduce sales

Chart Elements and Controls

X-Axis (Time)

Format: Date labels (e.g., "Jan 2024", "Week 1")

Features:

  • Chronological order

  • Adjusts to data frequency (weekly, monthly)

  • Zoomable and pannable

Reading:

  • Left to right = past to recent

  • Identify specific periods

  • Compare across time

Y-Axis (KPI Value)

Format: KPI units (dollars, units, etc.)

Features:

  • Shows contribution magnitude

  • Starts at zero (or goes negative)

  • Auto-scales to data

Reading:

  • Higher = more contribution

  • Below zero = negative impact

  • Absolute values matter

Legend

Location: Below or beside chart

Elements:

  • Group name

  • Color swatch

  • Clickable for interaction

Interaction:

  • Click to hide/show groups

  • Isolate specific contributions

  • Compare subsets

Use Cases:

  • Hide Base to see only marketing

  • Show only Media to focus

  • Compare different scenarios

Toolbar (Top Right)

Available Tools:

🔍 Zoom: Enter zoom mode ➕ Zoom In: Magnify view ➖ Zoom Out: Reduce magnification ✋ Pan: Move view across time 🔄 Reset: Return to original view

Usage:

  • Access via icons

  • Hover for tooltips

  • Click to activate

Interactive Features

Zoom Functionality

Mouse Wheel Zoom:

  1. Hover over chart

  2. Scroll wheel to zoom in/out

  3. Maintains aspect ratio

  4. Focuses on mouse position

Selection Zoom:

  1. Click and drag to select area

  2. Releases to zoom to selection

  3. Detailed view of period

  4. Precise control

Button Zoom:

  1. Click Zoom In/Out buttons

  2. Step-wise magnification

  3. Centered on current view

  4. Controlled increments

Pan Navigation

Click and Drag:

  1. Click on chart

  2. Hold and drag left/right

  3. Navigate across time

  4. Explore different periods

Use Cases:

  • Move through long time series

  • After zooming, navigate to different periods

  • Compare different years

Reset View

When to Use:

  • Lost in zoomed view

  • Want to see full picture

  • Start fresh exploration

How:

  • Click Reset button

  • Returns to original zoom level

  • Shows all data

Hover Tooltips

On Stacked Bars:

Shows:

  • Date/time period

  • Group name

  • Contribution value

  • Percentage of total (if configured)

Example:

Date: March 15, 2024
Group: Media
Contribution: $45,230
Percentage: 35% of total

On Line Chart:

Shows:

  • Date

  • Actual value

  • Predicted value

  • Difference (gap)

Analyzing Contributions

Absolute vs. Relative Contributions

Absolute (Bar Height):

  • Dollar amount or KPI units

  • Direct impact measure

  • Useful for ROI calculation

Relative (Percentage):

  • Proportion of total

  • Useful for comparing across periods

  • Shows mix changes

Both Matter:

  • Absolute for business value

  • Relative for strategic shifts

Comparing Time Periods

Week-over-Week:

  • How did contributions change?

  • Campaign starts/stops visible

  • Tactical insights

Month-over-Month:

  • Smooths weekly noise

  • Campaign-level patterns

  • Strategic trends

Year-over-Year:

  • Long-term changes

  • Growth patterns

  • Strategic evolution

Identifying Outliers

Unusually Large Contributions:

  • Special events

  • Data quality issues?

  • One-time occurrences

Investigate:

  • Check actual data for that period

  • Verify model input values

  • Confirm business context

Unusually Small Contributions:

  • Campaign pauses

  • Budget cuts

  • Off-season periods

Common Chart Patterns

Pattern 1: Steady Base with Variable Marketing

Appearance:

  • Large consistent gray base

  • Variable colored marketing segments on top

  • Marketing creates peaks

Interpretation:

  • Strong organic baseline

  • Marketing drives incremental growth

  • Clear campaign effectiveness

Business Implications:

  • Marketing is working

  • Incremental ROI calculable

  • Base protects from market downturns

Pattern 2: Growing Baseline

Appearance:

  • Base segment increasing over time

  • Upward trend in total bar height

  • Marketing contributions also grow

Interpretation:

  • Business is growing organically

  • Compound effects building

  • Positive momentum

Business Implications:

  • Strong market position

  • Word-of-mouth/brand building working

  • Marketing amplifies growth

Pattern 3: Seasonal Cycles

Appearance:

  • Regular up-and-down patterns

  • Seasonality segment varies predictably

  • Annual rhythms clear

Interpretation:

  • Strong seasonal business

  • Predictable demand cycles

  • Marketing timing matters

Business Implications:

  • Plan campaigns around seasons

  • Budget allocation by quarter

  • Inventory planning insights

Pattern 4: Campaign-Driven Spikes

Appearance:

  • Large temporary increases in marketing segments

  • Aligned with known campaign periods

  • Clear before-during-after pattern

Interpretation:

  • Campaigns create significant lift

  • Measurable impact

  • Timing is critical

Business Implications:

  • Campaigns are effective

  • Can calculate campaign ROI

  • Optimize frequency and timing

Pattern 5: Price Sensitivity

Appearance:

  • Large price segment (positive or negative)

  • Changes correlate with pricing actions

  • Clear trade-offs visible

Interpretation:

  • Price elastic market

  • Pricing strategy matters

  • Revenue vs. volume trade-off

Business Implications:

  • Pricing power assessment

  • Promotion vs. everyday pricing

  • Margin optimization opportunities

Tips for Effective Analysis

Start with the Big Picture:

  • View full time series first

  • Identify major patterns

  • Note anomalies

Then Zoom for Details:

  • Focus on specific campaigns

  • Examine transitions

  • Investigate outliers

Compare Across Periods:

  • Similar time periods year-over-year

  • Campaign vs. non-campaign periods

  • Growth phases

Use Legend Strategically:

  • Hide Base to see marketing clearly

  • Show only controllable factors

  • Isolate specific groups

Cross-Reference with Actual:

  • Check model fit constantly

  • Large gaps = investigate

  • Good fit = trust decomposition

Document Insights:

  • Screenshot key patterns

  • Note business explanations

  • Build narrative

Troubleshooting Visual Issues

Bars Don't Match Predicted Line:

  • Should be impossible (calculation error)

  • Refresh and re-run

  • Check console for errors

Colors Are Unclear:

  • Too many similar shades

  • Reassign colors in Contribution Groups

  • Use more distinct colors

Chart Is Cluttered:

  • Too many groups

  • Consolidate small groups

  • Use drill-down instead

Can't See Small Contributions:

  • Zoom in on specific periods

  • Hide large Base group

  • Use variable-level drill-down

Negative Values Confusing:

  • Normal for price increases

  • Represents KPI reduction

  • Business context clarifies

Best Practices

Regular Review:

  • Update decomposition weekly/monthly

  • Track changes over time

  • Build historical understanding

Multiple Views:

  • Zoom levels for different insights

  • Hide/show groups for focus

  • Time period comparisons

Combine with Other Analysis:

  • Cross-reference with diagnostics

  • Check against business events

  • Validate with stakeholders

Tell the Story:

  • Use charts to build narrative

  • Explain peaks and valleys

  • Connect to business strategy

Next Steps

After analyzing the main chart:

  • Use Group Decomposition to drill into specific groups

  • Calculate Channel ROI

  • Identify Seasonal Patterns

  • Export charts for presentations

Last updated