bar

When to use

• The claim is a difference in magnitude. A bar earns its place when the audience should leave knowing which category is biggest and roughly by how much. Length from a shared zero is read more accurately than angle, area or color, which is why this beats a pie for anything that is not parts of one whole.
• Three to eight categories. Two categories is a ratio — say it as a big-number or a two-tile stats. Past eight the names crowd even in the row form and the reader stops comparing and starts scanning; consolidate the tail into 'Other'.
• Long names mean row. Column labels wrap into the width of one band, so a name past about fifteen characters is shortened or dropped. The row form puts the name in a wide left gutter where it reads on one line. When the names are business units, product lines or people, start from row.
• One value per category, or two to four series. The flat list is the single-series chart. Nesting numeric children turns it into the grouped form for a plan-vs-actual or year-over-year comparison. Past four series the bars inside a group get too thin to compare across groups — split the slide or switch to a small multiple.

When not to use

• Parts of one whole. If the categories add up to a total and the split is the story, use piechart or stacked-bar. A plain bar says nothing about it.
• A continuous series over time. Twelve monthly bars ask a reader to compare twelve lengths when the claim is a trend. Use line, whose job is the movement.
• A percentage against a target. '68% of goal' is attainment, not magnitude across categories. progress shows attainment; bullet adds the target and the band.
• A rainbow single series. One series is one hue by design: the length carries the comparison. Color earns its place in grouped, where it names the series.

Common mistakes

• Adding a grouped class and leaving the list flat. The grouped form is selected by the DATA, not the class: nest the series under each category with numeric pills. A grouped token on a flat list draws the ordinary single-series chart rather than an empty one.
• Expecting the value axis to appear on a small chart. The chart prints each value on its own bar and drops the axis and gridlines whenever those labels fit — the number is right there, so a second way to read it is redundant chrome. The axis returns only when a label would no longer fit the room it has: its whole band in a single-series column chart, or just its own bar's slot when the bars are grouped. A row chart never needs it, because its values print into a gutter sized to hold them.
• Writing values in mixed units, like $4.2M next to 3100000. Both parse to a number, but the axis affix is only adopted when EVERY value agrees on it — a mixed series gets a bare axis and loses the currency. Author one vocabulary for the whole list.
• Reaching for diverging just because one value happens to be negative. The ordinary form already hangs a negative bar below the zero rule with the axis extended to cover it. diverging is a stronger claim — it centers zero, gives both sides the same reach, and colors by sign — so use it when direction IS the story, not merely when a minus sign appears.

Authoring

<!-- _class: bar -->

## Which category is biggest.

- First `120`
- Second `86`
- Third `54`
- Fourth `31`

Slots

Data shape

• The value pill is a number with optional currency prefix, thousands commas, magnitude suffix and unit: $4.2M, 12,400, 68%, -3.1. The magnitude suffix is SCALE, not decoration — 1.2M is 1 200 000 — so 800k and 1.2M sit correctly on one axis.
• A negative value is legitimate in every form and hangs from the zero rule; the axis is widened to include it. Zero is legitimate too and draws no bar, with its value printed on the baseline.
• Bars are drawn in AUTHORED order, never re-sorted. Sort the list yourself when ranking is the story; leave it in natural order when the categories have one (quarters, stages, sizes).
• In the grouped form, a category that is missing one series simply draws fewer bars in that group rather than a zero — an absent measurement and a measured zero are different claims.
• A category name that will not fit its column band on two lines is not shortened: the chart switches itself to the row form, where the name reads in the gutter. How long is too long depends on how many categories share the width — about forty characters at four categories, about twenty at eight — so row is worth asking for outright whenever the names are business units, product lines or people.

Anatomy

┌─────────────────────────────────────────┐
│  header                                 │
│       Growth sits in two regions.       │
│                                         │
│              $4.2M  $3.1M               │
│           [####] [###]  $1.8M           │
│       [####] [###]  [##]   $0.6M        │
│        [####] [###]  [##]   [#]         │
│      -----------------------------      │
│       N.Am   EMEA   APAC   LATAM        │
│                                         │
│  footer                           1/19  │
└─────────────────────────────────────────┘

Variant decision rule

• default (no modifier) — Short category names — a quarter, a product, a region code. Columns read left-to-right as a sequence, which suits anything with a natural order.
• row — Category names are long, or there are more than about seven of them. A row label reads on one line in the gutter instead of wrapping into a narrow band, so nothing gets shortened. Reach for it the moment a name is longer than about fifteen characters.
• grouped — Two to four series measured on the same scale across the same categories — this year vs last, plan vs actual. Authored as a nested list; the nesting is what selects it.
• diverging — Values carry a sign and the direction is the point: variance to plan, sentiment, a tornado / sensitivity chart. Zero sits at the middle of the plot and the two sides share one scale.

Variants

<!-- _class: bar row -->

## Support load sits with two teams.

- Platform Engineering `1,240`
- Customer Success `980`
- Data & Analytics `410`
- Design Systems `260`
- Developer Relations `95`

<!-- _class: bar grouped -->

## We beat plan everywhere but APAC.

- Americas
  - Plan `3.2`
  - Actual `3.9`
- EMEA
  - Plan `2.6`
  - Actual `3.1`
- APAC
  - Plan `1.9`
  - Actual `1.4`

<!-- _class: bar diverging row -->

## What moves the forecast, and which way.

- Renewal rate `+2.4`
- Enterprise pipeline `+1.6`
- Price realization `+0.7`
- Services attach `-0.5`
- Churn in mid-market `-1.8`
- FX exposure `-2.9`

Related

• progressattainment against 100% rather than magnitude against each other
• piechartparts of a single whole, where the shares sum to one total
• funneleach stage is a subset of the one before and the drop-off is the story
• statsa row of headline figures with no like-for-like comparison between them
• big-numberone figure is the whole slide