There are four charts worth using and most people reach for the wrong one. Here is how to pick, what to strip out, and how to stop spending an afternoon on something that should take four minutes.
Four chart types cover nearly everything anybody asks you for.
Bar, for comparing things. Line, for showing change over time. Pie, for parts of a whole — and only when there are very few parts. Combo, when two measures on different scales belong on the same picture, like revenue against margin.
Almost every chart that confuses its audience is one of these four applied to the wrong question, or one of the exotic types applied to anything at all.
The question decides the chart, not your preference.
Comparing separate things — regions, products, months as categories, people — use bars. Horizontal bars if the labels are words, because words read horizontally and nobody should have to tilt their head. Sort them by size unless there is a natural order; an unsorted bar chart makes the reader do the ranking you should have done.
Showing something changing over continuous time — daily, monthly, across years — use a line. Lines imply continuity, which is why using one for unrelated categories is misleading: it suggests the space between two points means something.
Showing parts of a whole — use a pie, with fewer than five slices, and only when the parts genuinely sum to something meaningful. Beyond five slices nobody can compare the wedges, and a sorted bar chart says it better. Never two pies side by side for comparison; the eye cannot compare angles across circles.
Two different measures at once — revenue in money and margin in per cent — use a combo, bars for one and a line for the other, with the second axis labelled clearly.
What to avoid entirely: three-dimensional anything, because perspective distorts the values it is meant to show. Doughnut charts, which are pies with the middle removed and are harder to read for it. And any chart chosen because it looks impressive rather than because it answers the question.
Most bad charts are not badly chosen. They are cluttered.
The default output of any spreadsheet contains things that add nothing: heavy gridlines, a border, a legend for a single series, a title that repeats the axis labels, drop shadows, a background fill. Removing all of it takes thirty seconds and improves every chart.
The rules that matter.
Start bar charts at zero. Truncating the axis exaggerates differences, and whether or not you meant to mislead, that is what it does.
Label directly where you can. A line labelled at its end beats a legend that makes the reader look back and forth matching colours.
Use few colours. One colour for one series. If you must distinguish several, use shades of the same colour rather than a rainbow — and remember that a meaningful share of readers cannot reliably tell red from green, so never let colour alone carry the meaning.
Write the headline as the finding, not the topic. "Revenue up 18% in Q3" is a headline. "Q3 Revenue" is a label. If somebody reads only the title, they should still have learned something.
Twelve months of sales for four regions.
The default attempt is a three-dimensional stacked column chart with a legend at the bottom, twelve grouped clusters and four colours. It is technically complete and nobody can extract anything from it — comparing the height of a floating segment against another one in a different cluster is not something eyes do.
The version that works is a line chart: four lines, one per region, labelled at their right-hand ends, no legend, light gridlines, and a title saying what happened. Same data. One is a picture of the answer; the other is a picture of the spreadsheet.
Do I need to learn the chart tools properly?
Not to produce good charts. The choices above are about which chart and what to remove, and both are decisions rather than skills. The software part is a few clicks.
What if my boss wants something impressive?
Impressive and clear are usually the same thing, in the sense that a chart somebody understands instantly reflects better on you than one they have to squint at. Where they genuinely want decoration, give the decoration and keep the axis honest.
How many series on one chart?
Four or five lines is the practical limit before it becomes a bowl of spaghetti. Above that, split into several charts or highlight one series and grey the rest.
Should numbers be on the chart itself?
When there are few of them, yes — it removes the need to estimate against the axis. When there are many, no; the labels become the clutter.
The KISS Chart Generator is a spreadsheet with the four chart types already built and formatted — consistent colours, sensible axes, nothing decorative. Paste your data in and the chart is ready to present. No learning the chart tools, no fighting the axis settings. Seven euro, one file.
See the templates →