Skip to main content

Chart Type Decision Framework

Question Type → Chart Type Mapping​

This is deterministic. Given a question type and field characteristics, follow the branching logic below to arrive at exactly one chart type.


1. Temporal Questions​

Signal: "over time", "trend", "by month/week/year", "changed", "growing"

ConditionChart Type
Single measure, continuous time (>8 periods)Line chart
Single measure, ≤8 discrete periodsColumn chart (vertical bar)
Multiple measures (2-3), same scaleMulti-line chart
Multiple measures, very different scalesSeparate charts (never dual axis by default)
4+ measures over timeSmall multiples (sparkline grid)
Cumulative/running total emphasisArea chart
Micro-trend (space-constrained, shape > values)Sparkline

2. Categorical Comparison​

Signal: "by [category]", "compare", "across", "between", "which is bigger"

ConditionChart Type
3-20 categories, one measureHorizontal bar (sorted high to low)
>20 categoriesHorizontal bar (TOP 10 + filter for rest)
2 measures across categoriesGrouped bar (side-by-side)
With target/benchmarkBar + reference line
Precise value emphasis over magnitudeDot plot
Time periods as categories (Q1-Q4)Vertical column (left-to-right time order)

3. Composition​

Signal: "breakdown", "parts of", "share", "proportion", "mix", "what makes up"

ConditionChart Type
Parts across categoriesStacked bar
Parts changing over timeStacked area (caution: only baseline is accurate)
Hierarchical/nested categoriesTreemap
Sequential buildup to totalWaterfall
Simple 2-3 parts, whole is the pointPie (rare exception; prefer stacked bar)
Proportions more important than absolutes100% stacked bar
Only one dimension available (no sub-breakdown field)Stacked bar with a single bar: measure on the axis, the one dimension straight on Color, nothing on the other axis. Never a pie by default — see the Pie chart gate below before choosing pie.

Pie chart gate (ALL must be true):

  1. Exactly 2-5 segments (ideally 2-3)
  2. Segments sum to a meaningful whole
  3. Part-to-whole IS the primary message
  4. Values labeled clearly
  5. No 3D effects
  6. Not comparing multiple pies

4. Distribution​

Signal: "spread", "distribution", "range", "histogram", "variance", "outliers"

ConditionChart Type
Single variable frequencyHistogram
Compare distributions across categoriesBox plot
Full distribution shape neededViolin plot
Small dataset (<50 points)Dot plot (strip)

5. Correlation​

Signal: "relationship", "correlation", "vs.", "impact of X on Y", "associated with"

ConditionChart Type
Two continuous variablesScatter plot
+ third variable (size)Bubble chart (size variable must never be negative; a signed third variable goes on diverging Color)
+ categorical groupingScatter with color (max 5-6 colors)
Many variable pairsHeatmap (correlation matrix)

6. Ranking​

Signal: "top", "bottom", "best", "worst", "leaders", "laggards", "rank"

ConditionChart Type
Simple top/bottom NSorted horizontal bar
Rank change between two periodsSlope chart
Emphasis on endpoint valuesLollipop chart

7. Sequential/Flow​

Signal: "funnel", "flow", "stages", "conversion", "bridge", "waterfall", "timeline", "Gantt", "duration", "schedule"

ConditionChart Type
Step-by-step additions/subtractionsWaterfall
Flow between categories/stagesSankey (extension required)
Conversion drop-off at stagesFunnel (sorted bar with stage labels)
Duration/elapsed time per entity (start + end dates)Gantt chart
Project timeline or schedule visualizationGantt chart

8. Geographic​

Signal: "by state", "by country", "where", "map", "location", "region" (geographic)

ConditionChart Type
Value by polygon region (state, country)Filled map (choropleth)
Discrete locations + magnitudePoint map (bubble; size measure must never be negative)
Density/concentrationHeat map (geographic)
<5 regions totalBar chart (more precise than map)

9. Status/KPI​

Signal: "current", "are we on track", "status", "KPI", "how are we doing", "metric"

ConditionChart Type
Single metric with contextKPI tile (BAN + comparison + sparkline)
Actual vs. targetBullet chart
In naturally bounded range (0-100%)Gauge (sparingly)

Audience Overrides​

When audience is known, apply these modifications to chart selection:

Executive (5-30 second time budget)​

  • Suppress: Scatter, box plot, histogram, heatmap, violin, complex stacked
  • Prefer: KPI tiles, simple line (1-2 series), sorted bars (top 5), status indicators
  • Max charts per worksheet context: Keep it scannable

Analyst (3-15 minute time budget)​

  • Suppress: Over-simplified gauges, single-number-only views
  • Prefer: Scatter, distributions, small multiples, heatmaps
  • Allow: Higher complexity, more marks, deeper drill paths

Operations (5-30 seconds, alert-driven)​

  • Suppress: Complex analytics, historical deep dives
  • Prefer: Real-time KPIs, status dashboards (green/red), simple trends, gauges (if naturally bounded)

External/Public (30-90 seconds)​

  • Suppress: Statistical charts, jargon-heavy labels, box plots
  • Prefer: Simple bars, clean lines, heavy annotation, minimal chart types

Manager (1-3 minutes)​

  • Suppress: High-density analytical charts
  • Prefer: Trends with comparisons, category breakdowns, light filtering

Ambiguity Resolution​

When the question could map to two types:

AmbiguityResolution
Temporal vs. Categorical ("sales by quarter")If ≤8 periods and comparison is the point → Column. If >8 or trend shape matters → Line.
Ranking vs. Comparison ("sales by region")If user says "top" or "best" → Ranking (sorted). If neutral → Comparison (still sort by value).
Composition vs. Comparison ("breakdown by category")If "parts of whole" language → Composition (stacked). If "which is bigger" → Comparison (bar).
KPI vs. Temporal ("how are sales doing")If "right now" / current state → KPI. If "lately" / over a period → Temporal.

Always state your interpretation: "I'm reading this as a [type] question. If you meant [alternative], let me know."