Resources · Guide

Data visualization best practices

The golden rule: design for the lazy brain.¹ Whoever looks at your chart doesn't want to think. If they have to stop and study it to tell whether something is going up or down, the chart has failed. Eleven principles, each with its own "not like this / better like this".

1

First decide what you want to say, then choose the chart

Are you comparing things? Tracking how something evolves? Splitting up a total? That question alone tells you which chart works and which doesn't. There's no perfect chart, just the right one for each case. Bars and lines solve 8 out of every 10.

❌ Not like this

9 brands in a pie: impossible to tell which is biggest.

✓ Better like this

Bars sorted from highest to lowest: the ranking reads itself.

2

Remove everything that isn't data

Everything you draw has to add information.² Drop the 3D, the shadows, the gradients, the busy gridlines and the borders. They decorate, but they get in the way: the cleaner it is, the sooner it's understood.

Bad

3D bars with a gradient and a dense grid.

Good

Flat single-color bars and thin axes.

3

Color is for signaling, not for decorating

Use color to direct the eye to what matters, not to fill space. Two or three colors plus gray for everything else is plenty. Always the same color for the same thing. And avoid red and green together: 1 in 10 people can't tell them apart well.³

❌ Not like this

One color per bar: the rainbow means nothing and distracts.

✓ Better like this

All gray except the one below target, highlighted.

4

The axis starts at zero (or you're lying)

With bars, the baseline goes at zero. If you cut it, a 2 % rise looks like it has doubled.⁴ You only cut it when a tiny variation genuinely matters (a tenth of a degree), and always with a warning. No dual axes with different scales, and no showing only the slice that suits you. Misleading your boss or your client with an axis costs you their trust.

❌ Not like this

Axis from 99 to 103: a 2 % looks like a surge.

✓ Better like this

Axis from 0: that 2 % looks flat, which is what it is.

5

A number on its own says nothing

Every headline figure needs three things: whether it's up or down (▲ % versus the previous period), whether it's above or below target, and which period you're talking about. A 76 % says nothing if you don't know whether the goal was 70 or 80.

Bad

"Margin 18 %".

Good

"Margin 18 % ▲ +3pp vs quarter · target 15 % ✓".

6

What matters most goes top left

Just like when reading, the eye starts at the top left.⁵ Put the figure that matters most there. Separate blocks with white space, not with boxes and lines. The 5-second test: if someone can't find the key figure in that time, it needs reworking.

Bad

The key KPI lost in a bottom corner among boxes.

Good

Main KPI top left; the detail moves down the page.

7

The title tells the takeaway, it doesn't describe the chart

Showing the chart isn't enough: tell its story. A title that says what's happening, axes with their units, a note where needed, and where the data comes from. The title is the sentence you'd take into the meeting, not the axis label.

Bad

"Sales by month".

Good

"Sales have fallen below target since March".

8

First the question, then the metric

The right order is: what do I want to achieve → what do I need to know to decide → what data answers that. A good metric can be measured, genuinely matters and drives you to act, with its goal, its deadline and its owner. Distinguish those that warn you in time from those that merely confirm what already happened. Five to nine metrics per report is enough.⁶

Bad

25 metrics "just in case", with no goal or owner.

Good

7 KPIs that answer specific decision questions.

9

The big picture first, the detail only if asked

Start with the overview and leave the detail for those who want to dig deeper.⁷ Key figures at the top, trends in the middle, detail at the bottom, all on one screen. A report you can filter and navigate is worth more than 20 fixed pages. And most reports fail not for technical reasons, but for not thinking about who is going to use them.

Bad

20 static tabs that no one navigates.

Good

A summary screen with filters to drill down.

10

Tell a story, don't just dump data

A table of numbers is forgotten; a story sticks. A good chart is useful, clear and reveals something that wasn't visible at first glance.⁸ The most common mistake: cramming in so many lines it looks like a plate of spaghetti. Highlight just one and leave the rest in background gray.

❌ Not like this

Eight tangled lines: you don't know which to look at.

✓ Better like this

One highlighted, the rest in gray: the focus is obvious.

11

If the data is dirty, the chart lies

Most of the work is in cleaning the data before plotting it. And be careful about settling for the average alone: two datasets with the same mean can hide opposite realities.⁹ Always look at how the data is distributed, not just the average.

Bad

Deciding by the average without checking outliers or nulls.

Good

Validate the distribution before charting; clean first.

Checklist before signing off on a visualization

  1. 01Does the function justify the form? Bars/lines before anything exotic?
  2. 02Y axis from 0? If I cut it, is it justified and flagged?
  3. 03Every number with context (trend + target + period)?
  4. 04Purposeful color (≤3, no red+green)?
  5. 05High data-ink ratio (no 3D, shadows or gridlines)?
  6. 06Does the title tell the takeaway? Axes with units and source?
  7. 07What matters top left? Does it pass the 5s rule?
  8. 08Does it fit on one screen with ≤5-9 KPIs?
  9. 09Interactivity instead of static pages?
  10. 10Designed for the user and their decision, with reliable data?

Best practices, out of the box

Sapioverse already applies several of these principles by default: honest axes, context against period and target, the right chart for each question, and an explanatory insight. It's the tool that applies these best practices for you.

Request a demo See the chart catalog

References

The body is in plain language on purpose. Here's the source for each idea, in case you want to dig deeper.

  1. ¹ "Lazy brain" = fast, automatic thinking (System 1) — Daniel Kahneman, Thinking, Fast and Slow.
  2. ² Data-ink ratio: maximize the ink devoted to data — Edward Tufte. Decluttering — Cole Nussbaumer Knaflic.
  3. ³ Red-green color blindness: affects ≈8-10 % of men. Best practices for accessible color.
  4. ⁴ "Lie Factor": the visual effect should match the real effect — Edward Tufte.
  5. ⁵ Z/F reading patterns and the grouping principles of Gestalt (proximity).
  6. ⁶ Chain of objective → key question (KPQ) → metric (KPI). Balanced Scorecard — Kaplan & Norton.
  7. ⁷ Visualization mantra: "overview first, zoom and filter, then details on demand" — Ben Shneiderman (1996).
  8. ⁸ Qualities of a good chart: functional, beautiful, insightful, enlightening — Alberto Cairo, The Truthful Art.
  9. ⁹ Anscombe's quartet / Datasaurus: same averages, opposite distributions. Synthesis on dashboards: Wexler et al., Dashboards That Deliver (2025).