Data & decisions · Field guide
A bad chart is often accurate. The chart fails because it answers a different question from the one the audience is asking. Start from the question, and the question usually picks the chart. Pick a chart for each of the three questions below first.
Guess first
You need to show how a $4.2 million budget splits across nine departments. Which chart?
You have a year of weekly support tickets, and a new onboarding flow launched in week 30. Which chart shows whether tickets fell?
You have training hours and sales for each of 120 reps. Which chart shows whether more training goes with more sales?
The chooser
Choose what you are trying to show, and the chooser narrows the list.
Use it whenYou are comparing one value across categories, such as spend by department or tickets by product. Sort the bars from largest to smallest unless the categories have a natural order.
How it misleadsThe length of a bar is the message, so the axis must start at zero. A truncated axis can make a 3% difference look like a doubling.
A question it answersWhich five suppliers make up most of our spend?
Use it whenYou are showing how a value changes over continuous time, such as weekly sales or monthly churn. Mark the date of any change you want people to judge.
How it misleadsA short time window can make normal variation look like a trend. Show enough history for the reader to see what an ordinary month looks like.
A question it answersDid ticket volume fall after the new onboarding launched in week 30?
Practise it: The Process Behaviour Chart →Use it whenYou have exactly two points in time, such as last year and this year, for several categories. Each line shows which categories rose, which fell and which overtook others.
How it misleadsTwo points hide everything that happened in between. Use a line chart if the path matters.
A question it answersWhich regions improved customer satisfaction between the 2025 and 2026 surveys?
Use it whenYou have more than four or five series. Instead of one chart with tangled lines, draw one small chart per series on the same scale.
How it misleadsEvery panel must share the same axis range. Otherwise the reader compares shapes drawn at different scales.
A question it answersHow has headcount changed in each of our twelve offices?
Use it whenYou are comparing the make-up of several groups, such as the channel mix in each region.
How it misleadsOnly the first segment and the last segment share a baseline, so the middle segments are hard to compare. Put the segment that matters most first.
A question it answersDoes each region get its revenue from the same channels?
Practise it: Simpson’s Paradox Splitter →Use it whenYou are showing one whole with two or three parts, and the message is roughly “about a quarter” or “more than half”.
How it misleadsPeople judge angles and areas less accurately than lengths (Cleveland and McGill, 1984), so a pie with more than a few slices is hard to read. Use a sorted bar chart instead.
A question it answersWhat share of the IT budget runs the business, and what share changes it?
Use it whenYou are explaining how one total became another, such as last year’s operating cost becoming this year’s, with each driver shown as a step up or down.
How it misleadsThe order of the steps shapes the story. Agree the order before the meeting, and show the start and end totals as full bars.
A question it answersWhat moved our operating cost from $12 million to $13.5 million?
Use it whenYou want to see how values are spread, such as delivery times or deal sizes. The shape shows the typical value, the spread and any long tail.
How it misleadsAn average hides the shape. Two teams with the same average resolution time can have very different worst cases. The width of the bins also changes the picture, so try more than one width.
A question it answersHow long do our customers wait for a refund, including the worst cases?
Use it whenYou are comparing the spread of a value across several groups, such as call handling time by team.
How it misleadsSome readers have not learned to read a box plot. Label the median, and say in the title what the box covers.
A question it answersWhich team’s handling times vary the most?
Use it whenYou want to know whether two measures move together, such as training hours and sales per rep. Each dot is one case.
How it misleadsA clear pattern shows that the measures move together. The pattern does not show that one measure causes the other, so look for a third factor before you act.
A question it answersDo reps with more training hours sell more?
Practise it: Correlation ≠ Causation →Use it whenYou are looking for a pattern across two categories, such as support tickets by day of the week and hour of the day.
How it misleadsPeople read shade less precisely than position (Cleveland and McGill, 1984), so a heat map is for spotting patterns rather than reading exact values. Use one hue from light to dark so colour-blind readers can follow it.
A question it answersWhen during the week do our systems carry the most load?
Use it whenThe reader needs to look up exact values, or will compare specific numbers line by line.
How it misleadsA table does not show a pattern. If there is a message, chart it and put the table beside the chart or in an appendix.
A question it answersWhat is each department’s approved budget for next year?
These charts can be drawn accurately and still mislead, because the design does the persuading.
Why it misleadsTwo measures share one plot with a different scale on each side. Whoever sets the two scales decides where the lines cross, so the chart can show almost any relationship.
Use insteadIndex both series to 100 at a common starting point, or draw two charts one above the other.
Why it misleadsPerspective makes the nearest slice or bar look larger than it is, and the depth adds nothing the data contains.
Use insteadUse the flat version of the same chart.
Use it this week
Before you build or approve the next chart in a report, write the question it answers as its title, such as “Did tickets fall after the onboarding change?”. If the chart does not answer that question at a glance, change the chart and keep the title.
Reading and questioning data is one part of thinking well in a digital workplace. The Digital Intelligence course builds the rest of that skill one day at a time.