Choosing the right chart

The heart of visualisation is picking the right chart for what you want to show. The same data in the wrong chart confuses or misleads; in the right one, the point is instantly obvious. This lesson covers choosing the right chart for a dashboard. Let's match data to visual. Let's choose well.

Matching the chart to what you're showing

THE GOLDEN RULE: pick the chart by WHAT YOU'RE TRYING TO SHOW, not by what looks cool. Each
chart type answers a different kind of question. The main types + when to use them:

COMPARE VALUES ACROSS CATEGORIES -> BAR / COLUMN CHART (the workhorse):
- Sales by product, users by country, votes by option. The clearest way to compare a handful of
  categories. Horizontal BARS for long labels/many categories; vertical COLUMNS for few. SORT by
  value for instant ranking.

SHOW CHANGE OVER TIME (a trend) -> LINE CHART:
- Revenue by month, daily active users. Time on the x-axis. The tool for trends; a few lines
  compare series. (Area charts fill under a line — for cumulative/volume.)

SHOW A SINGLE KEY NUMBER -> A KPI / SCORECARD / STAT CARD:
- One big bold number ("₦2.4M revenue"), ideally with a comparison (vs target, +15%). Don't chart
  a single number — show it big.

SHOW PARTS OF A WHOLE (composition) -> use CAREFULLY:
- A PIE / DONUT shows % share — but only for a FEW slices, + it's hard to compare precisely. Often
  a (sorted) BAR chart is clearer. STACKED BAR shows composition across categories/over time.
- Use pie sparingly; prefer bars for accuracy.

SHOW A DISTRIBUTION (spread of one variable) -> HISTOGRAM (bins of a numeric variable). BOX PLOT for
spread + outliers across groups.

SHOW A RELATIONSHIP between two numeric variables -> SCATTER PLOT (one on each axis — correlation +
patterns).

SHOW DATA ON A MAP (geographic) -> a MAP / filled map. SHOW two categories crossed -> a HEATMAP or
grouped bar. SHOW a table of exact values -> a TABLE (when people need the precise numbers).

The golden rule: pick the chart by what you're trying to show, not by what looks cool — each chart type answers a different kind of question. The main types + when to use them: compare values across categories → bar / column chart (the workhorse — sales by product, users by country — the clearest way to compare a handful of categories — horizontal bars for long labels/many categories; vertical columns for few — sort by value for instant ranking); show change over time (a trend) → line chart (revenue by month, daily active users — time on the x-axis — a few lines compare series; area charts for cumulative/volume); show a single key number → a KPI / scorecard / stat card (one big bold number — "₦2.4M revenue" — ideally with a comparison — don't chart a single number, show it big); show parts of a whole (composition) → use carefully (a pie / donut shows % share — but only for a few slices, and it's hard to compare precisely — often a sorted bar is clearer; stacked bar shows composition across categories/over time — use pie sparingly); show a distribution → histogram (bins of a numeric variable; box plot for spread + outliers across groups); show a relationship between two numeric variables → scatter plot (one on each axis — correlation + patterns); and geographic → a map; two categories crossed → a heatmap or grouped bar; exact values → a table.

The quick picker and choosing well

THE QUICK PICKER (memorise this):
- COMPARE categories        -> BAR / COLUMN (sorted).
- TREND over time           -> LINE.
- A single number           -> KPI card (big number + comparison).
- PARTS of a whole          -> stacked bar; pie/donut (few slices, sparingly).
- DISTRIBUTION              -> histogram (or box plot).
- RELATIONSHIP (2 numerics) -> scatter.
- GEOGRAPHIC                -> map.
- EXACT values              -> table.

MATCH THE CHART TO THE DATA TYPE (from types-of-data thinking):
- CATEGORICAL data -> compare with bars, compose with stacked bars/pie.
- TIME data -> line (trend).
- NUMERIC distribution -> histogram; two numerics -> scatter.

CHOOSING WELL — questions to ask:
1. What's my MESSAGE? (compare? trend? composition? relationship? one number?) -> that dictates the
   chart.
2. How many CATEGORIES/series? Too many bars/lines/slices = clutter — limit or group into "Other".
3. Will the audience READ it instantly? If it needs explaining, pick a simpler chart.

CHARTS TO AVOID / USE SPARINGLY:
- PIE with many slices (unreadable — use a bar). 3D charts (distort perception — never). Dual-axis
  charts (imply false relationships). Overly clever/novel charts (people can't read them —
  familiar beats fancy).

DOING IT WELL:
- DEFAULT TO BAR + LINE — they cover the vast majority of needs + everyone reads them. Reach for
  fancier only when the message truly needs it.
- ONE MESSAGE PER CHART — each visual makes ONE clear comparison/point.
- MATCH, DON'T DECORATE — the chart serves the message, not aesthetics.

THE PRINCIPLE: choose the chart by the MESSAGE — BAR to compare, LINE for trends, KPI CARD for one
number, stacked-bar/pie (sparingly) for composition, HISTOGRAM for distribution, SCATTER for
relationships, MAP for geography, TABLE for exact values. Default to bar + line (clear + universal);
avoid pie-overload + 3D. The right chart makes the point instantly obvious. Match the visual to the
message.

The quick picker (memorise this): compare categories → bar / column (sorted); trend over time → line; a single number → KPI card (big number + comparison); parts of a whole → stacked bar; pie/donut (few slices, sparingly); distribution → histogram (or box plot); relationship (2 numerics) → scatter; geographic → map; exact values → table. Match the chart to the data type: categorical → compare with bars, compose with stacked bars/pie; time → line (trend); numeric distribution → histogram; two numerics → scatter. Choosing well — questions to ask: (1) what's my message? (compare? trend? composition? relationship? one number?) → that dictates the chart; (2) how many categories/series? too many bars/lines/slices = clutter — limit or group into "Other"; and (3) will the audience read it instantly? if it needs explaining, pick a simpler chart. Charts to avoid / use sparingly: pie with many slices (unreadable — use a bar), 3D charts (distort perception — never), dual-axis charts (imply false relationships), and overly clever/novel charts (people can't read them — familiar beats fancy). Doing it well: default to bar + line (they cover the vast majority of needs and everyone reads them — reach for fancier only when the message truly needs it); one message per chart (each visual makes one clear comparison/point); and match, don't decorate (the chart serves the message, not aesthetics). The principle: choose the chart by the message — bar to compare, line for trends, KPI card for one number, stacked-bar/pie (sparingly) for composition, histogram for distribution, scatter for relationships, map for geography, table for exact values; default to bar + line (clear + universal); avoid pie-overload + 3D — the right chart makes the point instantly obvious.

The mistake beginners make

The first mistake is the wrong chart for the message — a pie for a trend, a line for categories; match the chart to what you're showing. The second mistake is pie-chart overload — pies with many slices, impossible to compare; use a sorted bar (pies only for a few slices). The third mistake is 3D / fancy / novel charts — 3D effects that distort, or clever charts nobody reads; familiar (bar/line) beats fancy. And too many series/categories — a spaghetti of lines or forest of bars; limit or group into "Other". And charting a single number — a tiny chart for one KPI; show it as a big number (stat card). Match chart to message, avoid pie overload, skip 3D/fancy, limit series, and show single numbers big.

Your turn

---

Your turn

1. Use the quick picker: COMPARE categories -> bar/column (sorted); TREND over time -> line; a single number -> KPI card; PARTS of a whole -> stacked bar / pie (sparingly); DISTRIBUTION -> histogram; RELATIONSHIP -> scatter; GEOGRAPHIC -> map; exact values -> table.
2. Start from the message: for each visual, ask 'what am I trying to show?' (compare/trend/composition/relationship/one number) - and let that dictate the chart, not aesthetics.
3. Default to bar + line: use bars (comparisons) and lines (trends) for the vast majority of needs since everyone reads them instantly - reaching for fancier charts only when the message truly needs it.
4. Avoid the traps: don't use pies with many slices (use a sorted bar), never use 3D charts (they distort), limit the number of lines/bars/slices (group small ones into 'Other'), and show single KPIs as a big number not a chart.
5. Test readability: check that the audience can grasp each chart's point instantly - if it needs explaining, switch to a simpler, more familiar chart.

Key points

• GOLDEN RULE: choose the chart by the MESSAGE (what you're showing), not by what looks cool. QUICK PICKER: COMPARE -> BAR/COLUMN (sorted), TREND over time -> LINE, single number -> KPI/STAT CARD (big number + comparison), PARTS of a whole -> stacked bar; pie/donut (few slices, SPARINGLY), DISTRIBUTION -> HISTOGRAM (box plot for groups), RELATIONSHIP (2 numerics) -> SCATTER, GEOGRAPHIC -> MAP, exact values -> TABLE.
• Match chart to DATA TYPE: categorical -> bars/stacked; time -> line; numeric distribution -> histogram; two numerics -> scatter.
• Choose well: 1) what's my MESSAGE? -> the chart; 2) how many categories/series? (too many = clutter — limit/group into 'Other'); 3) will they READ it instantly? (if it needs explaining, go simpler).
• AVOID: pies with many slices (use a bar), 3D charts (distort — never), dual-axis (false relationships), and clever/novel charts (familiar beats fancy). DEFAULT to BAR + LINE (clear + universal). ONE message per chart. MATCH, don't decorate.
• The mistakes: wrong chart for the message, pie-chart overload (sorted bar instead), 3D/fancy/novel charts, too many series/categories (limit/group), and charting a single number (show it big).