Resources · Guide

Which chart to use for what

First think about what you want to show —comparison, trend, proportion…— and that almost picks the chart for you. And when in doubt: bars or lines, which are understood at a glance. Pies, worse.¹

Compare categories

Bars / columns

The workhorse: together with lines, they solve 8 out of every 10 charts you'll make. For comparing how much each thing sells, nothing reads faster.

  • Yesfew-to-moderate categories; horizontal if the labels are long.
  • Notime trends; 3D; truncated Y axis.

"Which channel brings in the most revenue?"

Time trends

Lines

The eye follows the line on its own and sees the trend effortlessly.² It's ideal for watching how something evolves over time and for comparing one year against another: one line per year.

  • Yesany metric over time, YoY.
  • Nomany series = "spaghetti" → highlight one or use small multiples.

"Are sales growing in 2025 vs 2024?"

Part-to-whole proportion

Pie / donut

Shows how a total that adds up to 100 % is split. It's the most misleading chart: the eye is terrible at measuring the slices of a pie.¹ Use it with care.

  • Yesonly ≤5-6 categories with one clearly dominant.
  • No>6 categories or 3D → a horizontal bar is almost always better.

"What share does each channel have of the total?"

Correlation · Segmentation

Scatter

Crosses two figures (for example price and volume) and lays bare who falls outside the norm. If you color each group, it's the natural way to split your customers into profiles.

  • Yesprice vs volume, grouping customers by behavior.
  • Nothings that aren't numbers. And beware: the average alone is misleading, always look at the cloud.³

"Into which groups do my customers segment?"

Intensity

Heatmap

Encodes intensity by color over a matrix (category × period) or a map. Perfect for spotting hot zones at a glance.

  • Yeswhich product × month sells most; routes with incidents.
  • Nowhen you need to read the exact figure (color is imprecise).

"Which category and month concentrate the sales?"

Hierarchy

Treemap / sunburst

Shows a hierarchy with proportion by nested area: category → brand → product. Area is less precise than length, so use it for the big picture, not for fine figures.

  • Yeshierarchical composition of the catalog or of revenue.
  • Notoo many levels at once; exact size comparisons.

"How is my revenue split by category and brand?"

The rest of the catalog

Less common types, but useful when the objective calls for them. Remember: the more exotic, the easier to misread.

Area

Trends emphasizing accumulated volume or temporal composition (stacked).

Beware: several stacked areas read poorly.

Bubbles

Multivariate: X, Y, size and color (3-5 dimensions). Useful in marketing.

Beware: area is imprecise; >6 variables overloads it.

Histogram

Distribution of a quantitative variable: frequency by range.

Beware: adjacent bars (quantitative) ≠ separated bars (categorical).

Boxplot

Compact distribution: median, quartiles and outliers. Good for comparing groups.

Beware: not very intuitive for a non-technical audience.

Scatterplot matrix

Correlations across many variables at once. An exploration tool (EDA).

Beware: too dense for executives.

Choropleth map

Geographic distribution: color by region. Always in rates or ratios, never absolutes.

Beware: large, sparsely populated regions dominate the map.

Flow map

Movements between locations: routes, logistics, origin-destination.

Beware: too many flows = an unreadable blob.

Network graph / Sankey

Relationships and flows: fraud (force-directed), org chart (tree), supply chain (Sankey).

Beware: unreadable without grouping or filtering.

ventas marca precio stock canal

Word cloud

Term frequency in text (reviews). Basic sentiment once stopwords are removed.

Beware: indicative only, not suitable for rigorous analysis.

+18%

BAN / headline KPI

A single key figure read at a glance, in the top-left of the report. Always with trend + target + period — a number on its own says nothing.

Beware: "Margin 18 %" tells you nothing; "Margin 18 % ▲ +3pp vs quarter · target 15 % ✓" does.

Summary: goal → chart

Start on the left (what you want to show) and let the middle column decide.

Goal Chart Watch out for
Compare categoriesBars (horizontal if labels are long)Pie >6; 3D; truncated axis
Time trendsLines (area if volume matters)Spaghetti → small multiples
Part-to-whole proportionPie ≤5-6; otherwise, barImprecise angle
Correlation, 2 varsScatterThe average misleads; look at the scatter³
Multivariate, 3-5BubblesImprecise area
DistributionHistogram / boxplotConfusing it with bars
Geography (region)Choropleth (rates)Absolutes / large regions
Geography (movement)Flow mapBlob
IntensityHeatmapImprecise color
Relationships / networkGraph (Sankey / tree)Unfiltered = unreadable
HierarchyTreemap / sunburstImprecise area
TextWord cloudIndicative only
Key KPIBAN (+ context)A lone number

Choosing the right chart is half the job. The other half is not lying with it. Continue with the best-practices guide.

References

For anyone who wants to dig deeper. The main text is in plain language on purpose; here are the sources behind each idea.

  1. ¹ Hierarchy of perceptual accuracy (position and length > angle and area) — Cleveland & McGill (1984).
  2. ² Gestalt principles of visual grouping (Gestalt, continuity): the eye follows the line as a path.
  3. ³ Anscombe's quartet / Datasaurus: identical averages, radically different distributions. The average alone lies.