Scatter Plots
What Are Scatter Plots?
Scatter plots display the relationship between two variables, with each point representing one time period. They help you see if two variables move together, move in opposite directions, or are unrelated.
Purpose: Explore correlations between two variables, validate expected relationships, and identify outliers.
When to Use
Best For:
Checking if two variables are related
Validating expected correlations (e.g., Spend vs. Sales)
Detecting non-linear relationships
Identifying outliers and unusual observations
Exploring diminishing returns or saturation effects
Examples:
TV Spend vs. KPI
Price vs. Sales Volume
Digital Spend vs. Conversion Rate
Competitor Activity vs. Market Share
How to Create
Requirements:
Select exactly 2 variables
Click "Scatter Plot" button (📍)
Click "Generate Chart"
Chart Layout:
X-axis: First selected variable
Y-axis: Second selected variable
Each dot: One time period
Can zoom in both X and Y directions
Reading Scatter Plots
Positive Correlation:
Points trend upward from left to right
As X increases, Y increases
Strong relationship if points are tight
Example: Higher spend → Higher sales
Negative Correlation:
Points trend downward
As X increases, Y decreases
Example: Higher price → Lower volume
No Correlation:
Random scatter, no clear pattern
Variables are independent
No predictable relationship
Non-Linear Pattern:
Curved relationship
Example: Diminishing returns (curve flattens at high spend)
May need saturation curves in model
What Variables to Visualize
Marketing Variables:
Channel Spend vs. KPI
Impressions vs. Conversions
Reach vs. Sales
Validation Checks:
Expected positive: Spend vs. Sales
Expected negative: Price vs. Volume
No expected relationship: Random variables
Before Modeling:
Test assumptions about relationships
Verify correlations exist
Identify transformation needs
Common Patterns
Linear Relationship:
Straight line pattern
Constant rate of change
Good for linear regression
Diminishing Returns:
Curve that flattens
Each additional dollar has less impact
Needs saturation curve transformation
Threshold Effect:
Flat then steep increase
Minimum spend needed for impact
Consider S-curve transformation
Clustered with Outliers:
Most points together
Few points far away
Investigate outlier time periods
Use Cases
Relationship Validation:
Question: Does TV spend drive sales?
Variables: TV_Spend (X) vs. Sales (Y)
Pattern: Points trend upward
Conclusion: Positive relationship confirmedSaturation Detection:
Variables: Total_Marketing_Spend vs. KPI
Pattern: Curve flattens at high spend
Insight: Diminishing returns at $500K+ spendOutlier Investigation:
Variables: Digital_Spend vs. Conversions
Pattern: Most clustered, 2 points far away
Action: Check those time periods for data issuesInteractive Features
Zoom:
Mouse wheel or selection box
Zoom both X and Y axes
Focus on specific value ranges
Tooltip:
Hover over any point
Shows X value, Y value, and date
Identify specific time periods
Pan:
Click and drag when zoomed
Navigate to different regions
Tips
Variable Selection:
Choose variables you expect to be related
Test one relationship at a time
Use multiple scatter plots for multiple pairs
Interpretation:
Tight clustering = strong relationship
Wide scatter = weak relationship
Look for overall trend, not individual points
Next Steps:
If strong correlation → Include in model
If no correlation → Reconsider variable
If non-linear → Apply transformation
When to Use Other Charts:
Time trends → Use Line Charts
Multiple variables → Use Correlation Matrix
Multicollinearity → Use Correlation Heatmap
Summary
Scatter Plots Show:
Relationship between 2 variables
Correlation strength and direction
Non-linear patterns
Outliers
Use scatter plots to validate assumptions before modeling and identify variables that truly drive your KPI.
Last updated