Can I just ask it to make the chart?

You can, and for a throwaway look it is fine. For anything someone else will act on, the question and the hidden trade-off are the parts that decide whether the chart is honest.

What if my data cannot answer my question?

That is the most valuable output this pattern produces. It is why the last line forbids suggesting charts that need data you do not have — better to learn it now than after the chart is in the deck.

Data & spreadsheets

Question before chart

Charts fail because the question was never stated. Name the question and the data's shape first, then get chart options with the honest trade-off of each.

Use this before making any chart somebody else will look at.

The pattern

Copy this and replace everything in [SQUARE BRACKETS] with your own specifics.

Prompt template

Data: [WHAT EACH COLUMN IS, its type — number, category, date, proportion — and how many rows and categories there are.]

The question this chart must answer, for a reader who sees it for six seconds: [ONE QUESTION.]

The reader is [WHO] and will use it to [WHAT DECISION].

Give me:
1. Two or three chart types that would answer this question, and for each: what it makes obvious, what it hides, and when it stops working (too many categories, too few points, missing data).
2. Your recommendation, with the reason.
3. What the axes, units and sort order should be — and the one thing about this data that could make the chart mislead.

Do not suggest a chart type that needs data I do not have.

Why each part is there

Column types, not just names: Chart choice follows from data type. Two numbers over time, one number across categories and a proportion of a whole are three different problems with three different answers.
One question, six seconds: A chart answering three questions answers none. The six-second frame is what forces the choice.
What it hides: The line that earns its keep. Every chart hides something — a bar chart hides distribution, an average hides spread — and you want that trade-off stated before you commit.
When it stops working: Most bad charts are good charts used past their limit: a pie chart with eleven slices, a line chart over four points.
The one thing that could mislead: Turns the model into a reviewer rather than a generator. A truncated axis or one dominant outlier is what turns a technically correct chart into a false impression.

Before and after

The lazy version

What is the best chart for this data?

What goes wrong: You get “a bar chart”, with no question stated, no mention of what it conceals, and no warning that your 23 categories will be unreadable. The chart gets built, and then rebuilt.

The pattern, filled in

Worked example

Data: 340 rows. Columns: student (text), class (category, 4 values: 9/10/11/12), weekly minutes practised (number, range 0-600, roughly 30 rows are 0), month (date, 6 months).

The question this chart must answer, for a reader who sees it for six seconds: is practice time rising or falling, and is that the same across all four classes?

The reader is a school principal and will use it to decide whether to keep the programme next term.

Give me:
1. Two or three chart types that would answer this question, and for each: what it makes obvious, what it hides, and when it stops working.
2. Your recommendation, with the reason.
3. What the axes, units and sort order should be — and the one thing about this data that could make the chart mislead.

Do not suggest a chart type that needs data I do not have.

What that buys you: A small-multiples recommendation over a single four-line chart, with the reason — and a warning that averaging over the thirty zero-minute students will drag every line down, so you must decide whether a zero means “did not practise” or “was not enrolled yet” before you plot anything.

Now do it on your own work

Take a chart you have already made (or one from a news article) and write down the single question it answers in six seconds. If you cannot, run this pattern on the underlying data and rebuild it.

Check yourself — every box should be true

Your question is one sentence and answerable from the chart alone.
You can state what your chosen chart hides.
You can state the point at which it stops working for this data.
You checked the misleading-factor warning against your real data and either fixed it or labelled it on the chart.

Then explain it back

Your chosen chart hides something — what is it, and why was hiding that the right call for this particular reader and this particular decision?

Write two or three sentences in your own words. If you cannot, you copied a prompt rather than learning a pattern — which is the whole difference this library is trying to make.

Does this depend on which AI tool I use?

Keep the reasoning separate from the rendering. Decide the chart in this pattern, then build it in whatever tool you use — a spreadsheet, a notebook, or a chart feature inside an assistant. Tools that generate a chart image directly are convenient but hard to correct; tools that emit a formula or code give you something you can adjust and re-run.

Why it actually works

Asking what shape the data has before you draw it is exactly the first move in statistics — the distribution decides what an average is even allowed to mean.

See why the shape comes first →

Prompting well is a skill you can pick up in a week. Understanding what is happening underneath is the 18-month curriculum — and it is what stays useful when the tools change.

FAQ

Can I just ask it to make the chart?: You can, and for a throwaway look it is fine. For anything someone else will act on, the question and the hidden trade-off are the parts that decide whether the chart is honest.
What if my data cannot answer my question?: That is the most valuable output this pattern produces. It is why the last line forbids suggesting charts that need data you do not have — better to learn it now than after the chart is in the deck.

Next patterns

The Prompt Patterns Library is a separate, optional track — it does not change the 78-week curriculum or your roadmap. See what the curriculum actually covers →