The one idea
A chart is an argument, not decoration.
Before you make one, write the sentence it exists to prove:
"Revenue fell in the North region, and the whole fall is in one product."
Then choose the chart that makes that sentence obvious. If a reader can look at your chart for five seconds and say the sentence back to you, it works. If they have to study it, it does not — regardless of how much data it contains.
The corollary: if you cannot write the sentence, you do not need a chart. You need to go back to your analysis.
Decide what you want to show, then pick the chart that shows it most clearly.
Match the chart to the claim: comparison, change over time, relationship, or exact values. Put the finding in the title.
Encode the quantity you want compared in the visual channel people judge most accurately — position along a common axis — and reserve colour for categories, not for magnitude.
The mental model: claim first, chart second
| Your claim | Chart | Why |
|---|---|---|
| "A is bigger than B" | Horizontal bar | Length along a shared baseline is the easiest comparison humans make |
| "It changed over time" | Line | Slope reads as change; time belongs on the horizontal axis |
| "These two things move together" | Scatter | The only chart that shows a relationship between two numbers |
| "Here are the exact figures" | Table | A chart of eight numbers people need precisely is a worse table |
| "This part of the whole is dominant" | Bar, usually | A pie only works for two or three slices, and rarely then |
The most common mistake is skipping straight to a chart type because it is the default, or because it looks impressive. Start from the claim and this table mostly decides for you.
Why pie charts usually fail
People compare lengths well and angles badly. In a pie with seven slices, nobody can tell whether the 14% slice or the 16% one is bigger — which is exactly the comparison the chart was meant to enable. The same data as a horizontal bar chart sorted largest to smallest reads instantly.
A pie is defensible with two or three slices where one clearly dominates — "72% of orders came from mobile" — because the point is the proportion, not the ranking. It fails on anything with a long tail, and a 3D pie is worse, because perspective makes the front slices look larger than they are.
Recommendation If you are unsure, use a horizontal bar chart sorted by value. It is the right
answer more often than any other chart, and it is never badly wrong.
Truncated axes and the honest chart
A bar chart encodes value as length from zero. Start the axis at 95 instead of 0 and a 2% difference becomes a bar three times taller than its neighbour. The numbers are correct and the picture lies. Bar charts start at zero. There is no exception worth the credibility risk.
Line charts are different. They encode change, not magnitude, and forcing a line chart of body temperature to start at zero flattens every meaningful variation into a straight line. A non-zero baseline on a line chart is legitimate — but choose the range around the actual data, label the axis clearly, and do not zoom in so far that ordinary noise looks like a trend.
Two more ways charts mislead without any number being wrong:
- A chosen start date. Beginning in 2023 when the data starts in 2015 is a decision. Ask what the earlier years showed.
- Uneven intervals on the horizontal axis. Evenly spaced points labelled 2019, 2020, 2021, 2024 compress four years into one step and produce whatever slope was wanted.
Two versions of the same quarterly data.
The flattering one: a bar chart with the y-axis starting at 1.15 crore, so the 1.2 crore quarter towers over the 1.18 crore one. It says "strong growth".
The honest one: the same bars from zero, nearly identical in height, titled "Revenue flat across the quarter, up 1.7%".
You may be asked for the first. The argument that works is about risk, not ethics: a truncated axis is the first thing an experienced reader checks, and being caught costs more than a flat quarter ever would.
You have marks for 40 students and want to show the class did well. A bar chart with 40 bars is unreadable.
The claim decides it. "Most students scored above 70" is a histogram — bars of count per mark band. "Attendance relates to marks" is a scatter, one dot per student. "Here are the top five and bottom five" is a table. Same data, three different claims, three different right answers.
Labelling so the chart reads alone
Your chart will be screenshotted into a message with no context. Assume that.
- Put the finding in the title. Not "Q1 Revenue by Region" but "North fell 18% while every other region grew". The title is where your argument goes; the chart is the evidence.
- Label the axes with units. "Revenue (₹ lakh)". Numbers without units cause meetings.
- Say what the data covers. "Completed retail orders, Jan–Mar 2026, n = 8,412" in small text under the chart.
- Label directly where possible. Text at the end of each line beats a legend the reader must map back and forth.
- Sort bars by value, not alphabetically, unless the categories have a natural order like months.
- Use one colour. Add a second only to highlight the thing you are arguing about. Six colours for six bars adds nothing and makes the chart look busier than the finding is.
From a finished analysis to something a manager can use
1 of 6Write your conclusion as one sentence, before making anything. "North revenue fell 18% and it is entirely the mid-range product line." If you cannot write it, the analysis is not finished.
What to put in a summary for a manager
Most people send the analysis. Managers need the decision. The shape that works, in this order:
- The answer, in one sentence. "Yes, keep the discount — it added roughly ₹6 lakh in net revenue with no fall in margin."
- The two or three numbers that support it. Not twelve.
- What you recommend. Even if not asked. An analysis without a recommendation puts the thinking back on the reader.
- What would change your mind, or what you could not check. "This assumes returns stay at 4%; I could not get February's return data."
The chart goes after this, or attached. The detail goes in an appendix nobody opens, which is fine — its job is to exist when someone asks.
Try this
Which chart for each claim?
- "Our response time is slowest on Mondays."
- "Bigger orders take longer to ship."
- "Region A did 34% of total revenue, the other six split the rest."
- "Here are the exact monthly figures the finance team asked for."
Your challenge
Level 3 · IndependentTake a result from earlier in this module. Produce two charts of the same data: one designed to flatter, one designed to be honest.
Then write a short note explaining precisely which choices differ — axis, sorting, time range, colour, title — and how much each one changes the impression.
Success criteria: you can hand both to someone and they can identify every technique used in the flattering version. That is the skill: you will be shown charts like this, and you need to see it in under five seconds.
What people usually get wrong
- Making the chart before deciding the claim. You get a picture of your data instead of an argument.
- Truncating a bar chart axis. The fastest way to lose a reader's trust.
- Pie charts with more than three slices. Nobody can rank the slices, which was the point.
- Titles that name the data instead of the finding. "Q1 Revenue by Region" wastes the most-read line on the page.
- Two claims in one chart. Split it. Dual axes especially, where the relationship between the lines is an accident of the scales you chose.
- Sending charts with no summary. The reader has four minutes and will not do your analysis for you.
How someone experienced does it
Experienced people sketch the chart on paper before opening any tool. Thirty seconds of pencil settles the claim, the shape and the axis while changing your mind is still free. Once it is built in Excel, sunk effort makes you defend the chart you have rather than the one the argument needs.
They also design for the medium. A chart for a slide seen from the back of a room needs four bars and enormous labels. The same finding in a document read at a desk can carry twenty categories and a footnote. One chart reused for both is wrong for at least one.
And they put the number in the text rather than making the reader extract it from the picture. "North fell 18%, from 4.2 crore to 3.4 crore", with the chart underneath as evidence. Decision-makers read the sentence; the chart is for whoever wants to check it.
When not to use this
Do not chart three numbers. Write them in a sentence.
Do not chart when the reader needs exact values — use a table. And do not chart when the finding is genuinely "nothing changed". A chart of a flat line invites people to read meaning into noise, which is the opposite of what you found. Say "no meaningful change, within normal month-to-month variation" and move on.
Prove it
Produce one chart from your own data, with the finding as its title, labelled so it reads with no explanation from you.
Send it to someone with the three-line summary above it and no other context. If they reply with a decision or a follow-up question about the finding — rather than a question about what they are looking at — it worked.
Keep learning this
Paste this into any AI assistant. It turns the assistant into a tutor that tests you instead of just answering you.
Act as an experienced practitioner who is good at teaching. I have just learned choosing the right chart for a claim and labelling it honestly. Assume I am intelligent but relatively new to this — treat me as beginner level. Work through this in order, and wait for my reply at each step: 1. Ask me 5 questions that test whether I actually understood choosing the right chart for a claim and labelling it honestly. Do not reveal the answers yet. 2. After I answer, tell me which parts I got right, which I got wrong, and which I only half-understand. Explain only what I misunderstood — do not re-teach what I already know. 3. Give me one practical challenge based on something I could genuinely encounter at work or in daily life. Do not solve it for me. 4. Evaluate my solution the way an experienced person would judge it, including what a professional would have done differently. 5. Tell me what to learn next, and why that comes next. 6. Give me trustworthy sources for deeper study — prefer official documentation, primary research or standards bodies over blogs and videos. Rules for you: no buzzwords. No motivational filler. Say "I'm not certain" when you are not certain, and tell me which parts of your answer I should verify myself. Clearly separate facts from your recommendations and your opinions.
Become independent at this
Use this when you want a path from where you are to actually good, with checkpoints you can test yourself against.
I want to become independently capable at turning an analysis into a chart and summary someone will act on — not permanently dependent on AI, tutorials or step-by-step guides. Design a progression for me with five stages: Beginner, Guided practice, Independent practice, Real-world application, Professional level. For each stage tell me: - what I must know - what I must be able to do without help - the mistakes people make at this stage - one practical challenge - one real project that would prove I reached this stage - one way I can test myself honestly Then tell me the signals that I am ready to move to the next stage, and the signals that I have skipped ahead too early. Keep the theory to the minimum I actually need. Focus on ability I can transfer to situations you and I have not discussed.