Verified against ChatGPT · 2026-08-12
Pick the chart type that actually proves your point instead of just displaying your data
Recommends the right chart type for the specific claim a visualization needs to support, rejecting whichever default chart type would technically show the data but bury or misrepresent the actual point.
The prompt
Ready to copy — highlighted parts are example details you can swap.
I need help picking a chart type — not for "this data" in the abstract, but for the specific claim I need this chart to make clear to someone glancing at it. THE DATA Monthly active users for 6 product features over the last 12 months. THE CLAIM THIS CHART NEEDS TO SUPPORT Two of the six features have been steadily losing usage for 4 straight months while the other four are flat or growing. AUDIENCE AND VIEWING CONTEXT Product leadership, glancing at a slide for maybe 10 seconds during a roadmap review, not exploring it interactively. CHART TYPES ALREADY BEING CONSIDERED A stacked bar chart and a pie chart showing feature share of total usage were the first two ideas. Do this: 1. State what a viewer needs to be able to see at a glance for the claim to land — a comparison between a small number of categories, a trend over time, a part-to-whole relationship, a correlation between two variables, an outlier standing apart from a distribution. Different claims need genuinely different chart types, and naming this first prevents defaulting to whatever chart type is most familiar regardless of fit. 2. For each chart type already being considered, say plainly whether it would actually make the claim visible or would technically display the data while burying the point (e.g., a pie chart with nine slices of similar size showing a part-to-whole relationship no one can visually rank; a line chart connecting categorical, non-sequential data as if it were a trend). 3. Recommend the best chart type for this specific claim, even if it's not among the ones being considered, and justify it against the claim stated in step 1, not against general chart-type popularity. 4. Flag one specific way this chart type could still be built badly (wrong axis start point, too many categories crammed in, a misleading dual axis) that would undercut the claim even with the right chart type chosen, and how to avoid it. WHAT NOT TO DO Do not recommend a chart type based on how much data it can technically hold — recommend based on what the eye can extract from it in the few seconds a real viewer will actually spend looking. OUTPUT FORMAT 1. What the viewer needs to see at a glance (one sentence) 2. Verdict on each considered chart type: makes the claim visible / buries it, and why 3. Recommended chart type, with justification tied to the claim 4. One specific execution risk for that chart type and how to avoid it
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
Anchoring the recommendation to a specific stated claim rather than the data in the abstract matters because the same dataset genuinely supports different correct chart choices depending on what point is being made — six features' usage over time could be shown as a stacked bar, a small-multiples set of line charts, or a pie chart of current share, and only one of those actually makes "two features are declining while four are stable" visible at a glance; a model asked only "what chart should I use for this data" has no way to know which of those equally valid-looking options actually serves the point, so it defaults to whichever chart type is statistically most common for that data shape, not the one that proves the claim. Explicitly requiring a verdict on chart types already under consideration, rather than only proposing an alternative from scratch, forces confrontation with the actual failure mode at hand — a pie chart with six unevenly-changing slices over time doesn't even show change, it shows a single snapshot, and if this gap isn't named explicitly the recommendation reads as a preference rather than a correction of a real mismatch between the chart chosen and the point being made. Requiring one specific execution risk for the recommended chart type (rather than presenting the recommendation as risk-free) matters because chart-type selection is necessary but not sufficient — a line chart is the right structural choice for a declining-trend claim, but a truncated y-axis or an unlabeled inflection point can still make that correct chart type fail to land the claim in the ten seconds a real executive audience will spend on it, and naming the specific risk up front is what turns a correct chart type recommendation into one that will actually survive being put on a slide.
What you get back
What the viewer needs to see: two lines trending down against four that are flat or rising, over 12 months. Verdict: the pie chart doesn't show change over time at all, it's a single-moment snapshot, so it can't support this claim regardless of how it's built; the stacked bar chart could work but risks burying the two declining features' trend inside stacked segments that shift relative size for reasons unrelated to their own trend. Recommendation: a small-multiples line chart, one small panel per feature, all sharing the same y-axis scale, so the two declining lines are visually obvious without requiring the viewer to mentally subtract stacked segments. Execution risk: if each panel gets its own independent y-axis scale instead of a shared one, a small feature's modest decline could look visually identical to a large feature's steep one — fix by locking all six panels to the same y-axis range.
Verified against
ChatGPT GPT-5.1 · 2026-08-12
Changelog
- 2026-08-12 — Initial publish, verified against ChatGPT GPT-5.1.
Need this built into your business?
If a prompt isn't enough — what Scult builds, built and maintained for you — that's Scult's day job.
EXPLORE WHAT SCULT BUILDS
