Presenting Data: Which Chart When, and When No Chart At All
The chart was beautiful and nobody could tell what it meant
Nine colours. A legend down the right. Four years, six categories, two axes. Genuinely a lot of work.
And when it went up on screen, the room did what rooms always do: went quiet, squinted, and started reading the legend — which meant they stopped listening to you. By the time anyone had worked out which line was which, you were two sentences past the point the chart was supposed to support.
The mistake wasn't the design. It was earlier: the chart was built to display the data rather than to make a point. Those are different jobs, and only one of them belongs in a talk.
An exploration chart is for you, at your desk, finding out what's in there. A presentation chart is for one claim, already decided, made visible. Most bad slides are exploration charts that escaped.
The structure: Claim → Comparison → Form → Strip
1. Claim — say what the chart is for
Same discipline as PS-01. Before choosing a chart type, finish:
"This chart exists to show that ______"
- "Revenue grew but margin didn't."
- "Three of our nine products make all the profit."
- "The drop happened in March, not gradually."
If you can't finish the sentence, you don't need a chart — you need to look at the data yourself until you have a claim, or drop it from the talk.
2. Comparison — what is being compared to what
Every claim is a comparison, and naming which one determines the chart. There are only about five:
| Comparison | Sounds like |
|---|---|
| Over time | "grew", "fell", "spiked in March" |
| Between categories | "A is bigger than B" |
| Part of a whole | "three products are most of it" |
| Relationship | "as X rises, Y falls" |
| Distribution | "most are clustered, a few are extreme" |
3. Form — the chart follows from the comparison
Not from taste. From what the eye can actually judge.
| Comparison | Use | Not |
|---|---|---|
| Over time | Line | Bars — they imply discrete buckets and hide the shape |
| Between categories | Horizontal bar, sorted by value | Pie. Column charts when labels are long |
| Part of a whole | Bar with a total, or stacked bar | Pie beyond ~3 slices. Never a donut |
| Relationship | Scatter | Two lines on twin axes — see below |
| Distribution | Histogram or dot plot | A single average |
Why the eye, not taste. Cleveland and McGill tested how accurately people judge encoded quantities. The ranking is consistent: position along a common scale is judged most accurately, then length, then angle, then area, then colour intensity. Bars and dots on a shared axis use the top of that list. Pies use angle and area — near the bottom. That's the entire argument against pie charts, and it's an argument about perception rather than fashion.
Sort your bars. Alphabetical order is almost never the comparison you mean. Sorting by value does half the interpretive work before anyone reads a label.
Two lines on two different axes: don't. You can produce any apparent relationship by choosing the two scales, and audiences read the crossing point as meaningful when it's an artefact. If you must compare two differently-scaled series, index both to 100 at the start and use one axis.
4. Strip — remove everything that isn't the claim
Now delete, in this order:
- Gridlines — lighten heavily or remove
- The legend — label the lines directly at their ends. A legend forces a lookup for every glance
- Decimal places — 3.7% not 3.72941%
- Every colour but one — grey for context, one colour for the thing your claim is about
- Borders, shadows, 3D, gradients — 3D actively distorts; it's not a style choice
- The chart's default title — replace it with your claim
One colour is the highest-leverage move on this list. Nine categories in nine colours forces the audience to decode. Eight in grey and one in red is the sentence you were going to say, drawn.
5. The headline — the most important text on the slide
Replace the title with the claim, as a sentence:
| ❌ Label | ✅ Headline |
|---|---|
| "Monthly Revenue 2024–2026" | "Revenue grew 40%, and all of it came after March" |
| "Survey Results" | "Most people didn't know the feature existed" |
| "Q3 Performance by Region" | "Two regions carried the quarter" |
A label makes people work out the point. A headline tells them, and then the chart becomes evidence for a claim rather than a puzzle. This single change improves more slides than any redesign.
When no chart at all
Genuinely common and almost never done.
- One number → say the number. "Revenue was £2.4m" needs no chart. Put it on the slide in large type and say it.
- Two numbers → a sentence. "We grew from 1.8 to 2.4." A two-bar chart is a decoration of a sentence.
- Three or four values → a small table, which is more precise and read faster.
- The claim is about a mechanism, not a quantity → a diagram, or nothing.
- You want them listening rather than reading → no slide at all. A blank moment focuses attention on you, and it is startlingly effective.
Rule of thumb: under five numbers rarely needs a chart. Charts earn their space when there's a shape — a trend, a cluster, an outlier — that prose can't carry efficiently.
The rest of this walkthrough is for members
Behind this: the full step-by-step, the exercise with a verifiable output, and the downloadable cheatsheet. Everything you've read above stays free, always.