Concept: Chart Selection Matrix

flowchart TD
    Start{What is your primary analytical goal?}
    
    Start -->|Compare Categories| Comp{Label length or category count?}
    Comp -->|Short labels, 3-7 items| Col["Clustered Column Chart"]
    Comp -->|Long labels, >7 items| Bar["Horizontal Bar Chart"]
    Comp -->|Total + Subgroups| SBar["Stacked Bar / Column Chart"]
    
    Start -->|Analyze Trend over Time| Trend{Measure type?}
    Trend -->|Continuous line| Line["Line Chart with Markers"]
    Trend -->|Cumulative volume / mass| Area["Area Chart"]
    
    Start -->|Part-to-Whole Composition| CompPart{Number of categories?}
    CompPart -->|2 to 4 categories| Donut["Donut Chart (Center KPI)"]
    CompPart -->|5 to 30 categories| Tree["Treemap (Nested Rectangles)"]
    
    Start -->|Examine Distribution & Spread| Dist{Goal of distribution?}
    Dist -->|Median, IQR, Outliers| Box["Box Plot (Box & Whisker)"]
    Dist -->|Binned frequency shape| Hist["Histogram"]
    Dist -->|Smooth probability density| Dens["Density Curve (KDE)"]
    
    Start -->|Correlation / Relationship| Rel["Scatter Plot (XY Coordinates)"]
    Start -->|Pipeline Stage Drop-off| Funnel["Funnel Chart"]
    Start -->|High-Density 2D Matrix| Heat["Heat Map (Conditional Formatting)"]
    Start -->|Geographic Territories| Map["Filled Map (Choropleth)"]

The 10 Essential Architectural Questions

1. What is it?

The Chart Selection Matrix is a standardized decision-support tool that prescribes the mathematically and perceptually correct chart type based on the analytical objective (Comparison, Trend, Composition, Distribution, Relationship, Process, Intensity, Geographic) and the data types present.

2. Why is it used?

Choosing charts based on aesthetic novelty rather than analytical intent creates cognitive distortions. For example, using a line chart for nominal categories implies a false continuity, while using a 14-slice pie chart makes angle comparison impossible. The matrix ensures accuracy, cognitive speed, and visual integrity.

3. How does it work?

The matrix evaluates three inputs:

1. The Core Analytical Question: Are we comparing entities, tracking temporal trends, analyzing part-to-whole share, evaluating statistical spread, or testing correlation?
2. Primary Data Type: Qualitative (Nominal vs. Ordinal) or Quantitative (Continuous vs. Discrete).
3. Data Cardinality: Number of categories (2–4 vs. >7 items) or time horizons (discrete quarters vs. continuous daily timestamps).

4. Syntax & Decision Architecture

IF Goal = "Comparison" AND Labels = "Long"        --> Horizontal Bar Chart
IF Goal = "Trend"      AND Time = "Continuous"    --> Line Chart
IF Goal = "Part/Whole" AND Categories <= 4        --> Donut Chart
IF Goal = "Part/Whole" AND Categories > 5         --> Treemap
IF Goal = "Spread"     AND Multi-Group Comparison --> Box Plot
IF Goal = "Bivariate"  AND Variables = 2 Numeric  --> Scatter Plot (XY)
IF Goal = "Conversion" AND Stages = Sequential    --> Funnel Chart
IF Goal = "Spatial"    AND Key = Country/State    --> Filled Map

5. Practical Example in Excel

In the course project Hotel Reservation Analysis:

• Channel Performance: market_segment_type is Nominal with long text names ("Online Travel Agency", "Offline Tour Operator"). $\rightarrow$ Horizontal Bar Chart sorted descending.
• Seasonality: arrival_month is continuous chronological time. $\rightarrow$ Line Chart with Markers.
• Booking Status: booking_status has 2 discrete categories summing to 100%. $\rightarrow$ Donut Chart with 32.8% Cancellation Rate in the hollow center.
• Lead Time Outliers: lead_time is continuous numeric days. $\rightarrow$ Box Plot exposing that cancellations cluster above 90-day lead times.

6. Common Mistakes to Avoid

• Line Charts on Nominal Data: Plotting non-temporal categories on a line chart.
• Truncating the Column Baseline: Starting a vertical column chart at non-zero, exaggerating trivial variances.
• The 3D Pie Chart Trap: Distorting slice proportions via 3D perspective angles.
• Overcrowded Legends: Forcing viewers to decode 10 floating legend swatches instead of direct labeling.

7. When to Use

• Whenever designing a new executive dashboard or report tab.
• During exploratory data analysis (EDA) to expose distributions and correlations.
• In client or stakeholder presentations to ensure instant visual clarity.

8. When NOT to Use

• When raw tabular records or structured accounting schedules are legally or operationally mandated (e.g. general ledger auditing).
• When a single summary KPI number is sufficient (use a direct KPI scorecard instead of a full chart).

9. Real-World Analytics Use Case

In financial risk analysis, analysts frequently default to plotting portfolio losses as a simple column chart of averages. By consulting the Chart Selection Matrix, the team switches to a Box Plot, exposing extreme tail-risk outliers and negative skewness that were completely invisible when only evaluating the mean.

10. Related Concepts

• Dashboard Design Principles: The macro layout system housing selected charts.
• Pivot Tables: The aggregation engine that structures data into the exact format required for charting.
• Conditional Formatting: The mechanism for building in-cell heatmaps and data bars.