How a chart works
Every chart does the same thing underneath. It takes a dimension — dates, campaigns, countries, landing pages — and gives each value a position or a colour. It then takes a metric and turns it into a length, a height, an angle or an area. The reader’s eye does the comparison, which is the whole point: a table asks you to read, a chart lets you see.
Chart types differ in what the eye is good at. People judge the lengths of bars sharply and compare heights along a shared baseline well. They judge angles, areas and colour intensity poorly. That single fact explains most of the standard advice: bars for comparing categories, a line for change over time, a table when the exact figures matter, a map when geography is genuinely the question, and pie charts only when there are very few slices and the point is that one dominates.
Why charts matter
A chart is usually where the decision happens. Nobody reads the underlying rows. They look at the shape, form a view in a second or two, and act on it — so a badly chosen chart does not merely look untidy, it produces a different decision from the one the data supports.
The choice also sets what gets noticed. A metric plotted on its own draws attention to its movement; the same metric plotted beside its volume shows whether the movement was meaningful at all. A conversion rate rising while sessions collapse is a bad month, and a chart showing only the rate hides that completely.
Common mistakes with charts
Starting a bar chart’s axis somewhere other than zero exaggerates small differences dramatically, which is fine on a line chart tracking change but misleading on bars the reader is comparing by length. A pie with many thin slices is unreadable, and a stacked bar with many series is close behind.
Two axes on one chart is the persuasive one: putting spend and revenue on separate scales makes them appear to move together whether or not they do. Then there is decoration — charts added because a page looked empty. Every chart a reader cannot act on makes the ones that matter harder to find.
How to act on it
Write the question above the chart before choosing the type, then pick whatever answers that question most directly. One question per chart. If the answer is a single number, it belongs in a scorecard rather than a chart with one bar.
Label the units and the date range on the chart itself, because charts are screenshotted and forwarded without their page. Sort bars by value rather than alphabetically so the ranking is visible, limit series to the number a person can hold in their head, and pair every rate with the volume behind it. A reporting dashboard that follows those rules gets read; one that does not gets ignored, whatever it cost to build.
