Exporting Decomposition Data

Overview

Decomposition data can be exported to Excel for further analysis, custom visualizations, and stakeholder reporting. MixModeler includes decomposition sheets automatically in model exports.

Purpose: Access decomposition data outside MixModeler for advanced analysis, custom reporting, and integration with other tools.

How to Export

Step 1: Navigate to Model Library

• Click "Model Library" in the left sidebar

• View all your saved models

Step 2: Locate Your Model

• Find the model you want to export in the table

• Ensure it's the model you've run decomposition on

Step 3: Click Export Button

• Click the 📁 Export icon for your model

• Excel file downloads automatically to your browser's download folder

Step 4: Open Excel File

• Locate downloaded file (usually named: ModelName_export.xlsx)

• Open in Microsoft Excel or Google Sheets

Excel File Structure

The export includes multiple sheets:

Group Decomposition Sheet

Purpose: Time-series of group-level contributions

Columns:

• Date: Time period (week, month, etc.)

• Actual: Real KPI values from your data

• Predicted: Model's predicted KPI values

• Base: Base group contribution

• Media: Media group contribution (if exists)

• Price: Price group contribution (if exists)

• Promotions: Promotions group contribution (if exists)

• Seasonality: Seasonality group contribution (if exists)

• [Other Groups]: Additional groups you've defined

Structure:

Date       | Actual  | Predicted | Base   | Media  | Price  | Promotions | Seasonality
2024-01-01 | 50000   | 48500     | 20000  | 18000  | -2000  | 8000       | 4500
2024-01-08 | 52000   | 51000     | 20000  | 20000  | -2000  | 9000       | 4000
2024-01-15 | 48000   | 47500     | 20000  | 17000  | -2000  | 8500       | 4000
...

Each row: One time period Each column: One contribution group

Variable Decomposition Sheet

Purpose: Time-series of individual variable contributions

Columns:

• Date: Time period

• Actual: Real KPI values

• Predicted: Model predictions

• const: Intercept/constant term

• TV_Spend: TV variable contribution

• Digital_Spend: Digital variable contribution

• Price_Index: Price variable contribution

• [All Variables]: Each variable in your model

Structure:

Date       | Actual  | Predicted | const  | TV_Spend | Digital_Spend | Price_Index | Holiday
2024-01-01 | 50000   | 48500     | 20000  | 12000    | 6000          | -2000       | 4500
2024-01-08 | 52000   | 51000     | 20000  | 13000    | 7000          | -2000       | 4000
...

Use: Most granular view, variable-level ROI calculation

Other Sheets Included

The export also contains:

• Model Info: Model name, KPI, features, date range

• Coefficients: OLS or Bayesian coefficient estimates

• Model Statistics: R², Adjusted R², diagnostics

• Residuals: Actual vs. predicted with residuals

• Variable Details: Transformations, groups, adjustments

Using Exported Data

ROI Calculation in Excel

Step 1: Sum Contributions

In Excel:
=SUM(D2:D53)  // Sum TV_Spend column for 52 weeks
Result: $600,000

Step 2: Get Total Spend

From your marketing data:
TV total spend: $300,000

Step 3: Calculate ROI

In Excel:
=(Contribution - Spend) / Spend
=($600,000 - $300,000) / $300,000
=100% or 1.0

Creating Pivot Tables

Summarize by Time Period:

Step 1: Select Data

• Highlight decomposition data

• Insert > PivotTable

Step 2: Configure

• Rows: Date (grouped by month/quarter)

• Values: Sum of each group contribution

• Show contribution by period

Step 3: Analyze

• Monthly trends

• Quarterly summaries

• YoY comparisons

Example Output:

Quarter | Base    | Media   | Total
Q1      | 60000   | 50000   | 110000
Q2      | 60000   | 65000   | 125000
Q3      | 60000   | 55000   | 115000
Q4      | 60000   | 80000   | 140000

Custom Visualizations

Beyond MixModeler Charts:

Waterfall Charts:

• Show contribution build-up

• Period-over-period changes

• Group contributions

Area Charts:

• Stacked area showing trends

• Proportion changes over time

• Smoother than stacked bars

Combo Charts:

• Bars for contributions

• Line for actual/predicted

• Multiple Y-axes

Heatmaps:

• Contribution intensity by period

• Identify patterns

• Visual correlation

Time Series Analysis

Moving Averages:

In Excel:
=AVERAGE(B2:B14)  // 13-week moving average

Growth Calculations:

Week-over-week growth:
=(This Week - Last Week) / Last Week

Month-over-month:
=(This Month - Last Month) / Last Month

Trend Analysis:

Use Excel's TREND() function
Or add trendline to chart
Forecast future performance

Combining with Other Data

Join with Marketing Data:

Match by Date:

VLOOKUP or INDEX/MATCH to join
Combine decomposition with:
- Detailed spend data
- Campaign calendars
- External factors (weather, events)

Enhanced Analysis:

Spend by subchannel vs. contribution
Campaign-specific performance
Attribution by campaign type

Statistical Analysis

Correlation Analysis:

Correlation between groups:
=CORREL(MediaRange, PriceRange)

Identify relationships

Contribution Distribution:

Descriptive statistics:
- Mean, median contribution
- Standard deviation (variability)
- Min, max values

Sharing with Stakeholders

Executive Summary

Create One-Page Dashboard:

Key Metrics:

• Total KPI for period

• Marketing contribution ($, %)

• Top 3 contributors

• ROI by channel

Visuals:

• Main decomposition chart (screenshot or recreate)

• ROI bar chart

• Trend line

Recommendations:

• Top 2-3 actions

• Expected impact

Detailed Reports

For Marketing Teams:

Include:

• Variable-level decomposition

• Channel performance details

• Period-over-period changes

• Optimization opportunities

Format:

• Multi-tab Excel workbook

• Charts and tables

• Commentary and insights

Presentation Decks

PowerPoint/Google Slides:

Slide 1: Overview

• Decomposition chart

• Key findings

Slide 2: Channel Performance

• ROI comparison

• Recommendations

Slide 3: Trends

• Time series analysis

• Seasonal patterns

Slide 4: Next Steps

• Action items

• Timeline

Advanced Excel Techniques

Dynamic Charts with Slicers

Create Interactive Reports:

Step 1: Convert to Table

• Format data as Excel Table

• Enables filtering

Step 2: Add Slicers

• Insert > Slicer

• Add Date, Group filters

Step 3: Link to Charts

• Charts update automatically

• Interactive exploration

Scenario Analysis

Model "What If" Scenarios:

Create Scenario Table:

Scenario | TV Budget | Digital Budget | Expected Contribution
Current  | $300K     | $200K          | $1,000K
Shift 1  | $250K     | $250K          | $1,050K
Shift 2  | $200K     | $300K          | $1,100K

Use Historical ROI:

• Apply average ROI to new spend levels

• Estimate contribution changes

• Compare scenarios

Dashboard Creation

Build Excel Dashboard:

Components:

• Summary KPIs (total, YoY growth)

• Contribution breakdown chart

• ROI table

• Trend sparklines

• Top performers

Features:

• Update with new data

• Filter by period

• Drill-down capability

File Management Best Practices

Naming Convention:

ModelName_Export_YYYYMMDD.xlsx

Example:
Q4_2024_Model_Export_20241215.xlsx

Organization:

• Folder by model or time period

• Keep historical exports

• Version control

Documentation:

• Note export date

• Model version

• Any special considerations

Data Validation

Check Exported Data:

Verify Totals:

Sum of group contributions = Predicted
Predicted ≈ Actual (within model fit)

Check for Missing Data:

• All expected columns present?

• No unexpected blank rows?

• Date ranges complete?

Validate Calculations:

Spot-check a few periods:
Manual calculation = Excel values?

Integration with Other Tools

Google Sheets

Import Excel File:

• Upload to Google Drive

• Open with Google Sheets

• Same analysis capabilities

Collaboration:

• Share with team

• Real-time editing

• Comments and suggestions

BI Tools (Tableau, Power BI)

Import Decomposition Data:

• Load Excel file as data source

• Create custom dashboards

• Advanced visualizations

Use Cases:

• Executive dashboards

• Interactive exploration

• Automated reporting

Python/R Analysis

Load Excel Data:

Python:

import pandas as pd
df = pd.read_excel('model_export.xlsx', sheet_name='Group Decomposition')

R:

library(readxl)
df <- read_excel('model_export.xlsx', sheet='Group Decomposition')

Advanced Analysis:

• Statistical modeling

• Machine learning

• Custom algorithms

Troubleshooting Export Issues

File Won't Download:

• Check browser settings

• Disable pop-up blockers

• Try different browser

File Won't Open:

• Ensure Excel/compatible software installed

• Check file isn't corrupted

• Re-download

Missing Data:

• Verify decomposition was run

• Check groups are configured

• Ensure model is fitted

Columns Look Wrong:

• Column names = your group names

• Variable names = your variable names

• Verify grouping is saved

Summary

Exporting Enables:

Further Analysis:

• Custom calculations

• Advanced visualizations

• Statistical tests

Reporting:

• Stakeholder presentations

• Executive summaries

• Detailed documentation

Integration:

• Combine with other data

• Feed into BI tools

• Support decision-making

Best Practices:

Always Export:

• After finalizing model

• For important analyses

• For documentation

Organize Files:

• Clear naming

• Version control

• Archive historical exports

Validate Data:

• Check totals

• Verify calculations

• Spot-check accuracy

Share Appropriately:

• Right level of detail for audience

• Clear documentation

• Actionable insights

Next Steps:

• Export your model

• Explore the data

• Create custom analysis

• Share insights with team