line

When to use

• The axis is ordered and the order means something. Quarters, months, sprints, release numbers, stages of a rollout. A line connects its points, and that connection is a claim that moving from one to the next is meaningful. Plot unordered categories — regions, products, teams — with bar instead, where nothing is joined.
• Two to six series, named at the end of their own lines. Past six the categorical palette repeats and the lines stop being distinguishable (Wong 2011). Each series is labeled at its right-hand end in its own color, so the chart carries no key to look away to — consolidate a long tail into ‘Other’ rather than adding a seventh line.
• Four to about twenty-four points. Below four there is no shape to read — say it with big-number or stats. The kernel prints a dot per measurement while the points stay far enough apart to read as measurements, and drops the dots once the series is dense enough that the shape is the story.
• Values in one unit, on one scale. Every series shares the value axis, so the comparison the chart invites is a comparison of magnitudes. Two series in different units (dollars and percent) on one plot is a chart that lies quietly; split them across two slides, or plot the one that matters and state the other.

When not to use

• Two points. Two points are a slope, and slope draws it better. Note it is authored transposed: one bullet per entity, its two points nested.
• Categories that are not ordered. Joining 'Legal, Finance, Sales, Ops' asserts a progression that does not exist. Use bar, or stacked-bar if each one decomposes.
• A filled area for several series. Two opaque regions hide each other. Use stacked-area when the series sum to a total; the kernel drops the fill rather than lie.
• A trend where the story is attainment against a target. If the question is 'did we hit the number', the target is not on the line. bullet puts actual, target and a band on one row.

Common mistakes

• Adding area to a chart that has several series and expecting several filled regions. area fills only when there is exactly one series; with more the kernel renders plain lines, because opaque overlapping regions hide each other. Use stacked-area if the series sum to a total you mean.
• Assuming the value axis always starts at zero, as it does for a bar. A plain line or step axis floats when the data sits well above zero (it starts at zero once the low point falls inside the bottom sixth of the range), because forcing zero flattens a real trend into a scratch at the top of the box. The tick column always prints where the axis actually starts. area and stacked-area always reach zero — the filled region is read as quantity.
• Reordering the series under each category to ‘tidy’ a stacked area. In stacked-area the authored order IS the stacking order, bottom-first, and it is never sorted for you. Author the largest and steadiest series first so the noisy one rides on a stable base; a series stacked on a jagged neighbor inherits the neighbor's wobble.
• Leaving a series out of one category and expecting the line to bridge the gap. A missing point BREAKS the line — a gap is drawn, because a straight bridge would invent a measurement. Author every series at every category, or accept the gap. The one exception is stacked-area, where a missing point has to contribute zero to keep the stack's baseline, which is a claim you may not mean.
• Writing a nested bullet with a non-numeric pill under a category and expecting it to plot. A nested item is a data point only when its trailing pill parses as a number; - Best quarter in \EMEA\`` is detail, not a series called ‘Best quarter in’. That rule is what keeps a note ending in inline code from being silently plotted at zero.
• Reaching for area on a series whose movement is small relative to its own size. An area chart's axis always reaches zero, so a four-percent move flattens into a line across the top of a filled block. Plot it as a plain line, whose axis floats, and let the tick labels say where it starts.
• Mixing depths — giving one category its own value while the others carry a nested series list. A flat category among nested ones belongs to no series, so every line breaks across it at once. Nest it like the rest (- Q3 then - Product \4.2\``), or make the whole chart flat and plot one series.

Authoring

<!-- _class: line -->

## The number moved, and here is the shape of it.

- Q1 `12`
- Q2 `18`
- Q3 `17`
- Q4 `26`

Slots

Data shape

• Two points is the hard floor — below it the kernel leaves the list alone and renders no chart, because a single reading has no movement to draw. Four or more is where a shape starts to read.
• The authored order of the categories is the axis order — the kernel never sorts. Write them in the order they happened.
• Two to six series. Past six, parseSeries reports the overflow rather than truncating; consolidate the tail into one ‘Other’ series before it gets there.
• Series names are printed at the end of their own lines, so keep them under about twenty-five characters. A name too long for the right-hand column makes the chart fall back to a key, which is the weaker read.
• Use one unit across every series, and write it into the values ($4.2M, 12%, 840kg) — an affix every value agrees on is carried onto the axis, and a mixed series gets none.
• A magnitude suffix is scale, not decoration: 1.2M is 1200000 and 800k is 800000, so the two can sit on one axis. % is not a magnitude — 12% is 12.
• A category that omits a series leaves a HOLE, and the line breaks across it. In stacked-area a hole is forced to contribute zero — the stack needs a baseline — so author every series at every category before stacking.
• stacked-area assumes non-negative values; a negative contribution makes the bands cross and the total stop reading as a total. Use plain line for a series that can go below zero.
• Four to twenty-four points. Dots appear per measurement while the points are far enough apart to read as measurements and are dropped once they would bead the line.
• Keep every category at the SAME depth. A category written flat (- Q3 \4.2\``) among nested ones carries no series, so it renders as a hole in every line at once — the parser reports the mix, and the fix is to nest that category like its neighbors.

Anatomy

┌─────────────────────────────────────────┐
│  header                                 │
│       Services grew into the gap.       │
│                                         │
│     6M |        .-*-.      Renewals     │
│              |   .-*-`    `*            │
│       3M | *-`     .-*--*    New        │
│                 |.--*--*-`              │
│          0 +-------------------         │
│             Q1  Q2  Q3  Q4  Q1          │
│                                         │
│  footer                           1/19  │
└─────────────────────────────────────────┘

Variant decision rule

• default (no modifier) — The value changes continuously between the points you plotted — revenue, headcount, latency. The slope between two points is a real quantity, so drawing it as a slope is honest.
• step — The value HOLDS and then JUMPS — a list price, a rate, a band, a seat count. A straight interpolation between two step values draws a change that did not happen.
• area — There is exactly ONE series and its magnitude matters as much as its shape. With more than one series the fill stops being honest and the kernel drops it back to plain lines.
• stacked-area — The series ADD UP to a total the audience cares about, and the story is how the mix shifted inside it. If they do not sum to a meaningful total, stacking invents one.

Variants

<!-- _class: line area -->

## Cash on hand never dipped below one quarter of runway.

- Jan `8.2`
- Feb `7.6`
- Mar `6.9`
- Apr `7.4`
- May `8.8`
- Jun `9.6`

<!-- _class: line stacked-area -->

## The mix shifted; the total barely moved.

- FY23
  - License `6.2`
  - Support `2.8`
  - Services `1.4`
- FY24
  - License `5.6`
  - Support `3.4`
  - Services `2.0`
- FY25
  - License `4.9`
  - Support `3.9`
  - Services `2.7`
- FY26
  - License `4.1`
  - Support `4.4`
  - Services `3.4`

<!-- _class: line step -->

## The list price held for three quarters, then moved twice.

- Q1 `1,200`
- Q2 `1,200`
- Q3 `1,200`
- Q4 `1,450`
- Q1 `1,450`
- Q2 `1,690`

Related

• barthe categories are unordered and the claim is magnitude, not movement
• slopethere are exactly two points and the story is which entities changed rank
• stacked-bareach period decomposes into parts and the periods are compared, not connected
• bulletthe number is read against a target and a qualitative band, not against its own past
• ganttthe horizontal axis is real dates carrying durations rather than an ordered set of measurements