A chart prompt should begin with the decision the viewer must make, not a request for a fashionable graphic. State the measured quantity, unit, observation grain, population, and relevant threshold. A shipment count, a temperature reading, and time spent above a limit are different measures. The model can propose a display, but the application should supply verified data and a typed metric contract. If a requested conclusion cannot be supported by the available grain, the prompt must refuse that claim and offer a narrower one. A title should name what the marks encode rather than imply a causal finding.
Make this comfortable
Chart prompts: state the decision and measured quantity
Operational case
A fictional cold-chain team has six hourly readings for container C-47: 4.1, 4.4, missing, 5.8, 5.5, and 4.9 degrees Celsius. The operational question is which observed hours crossed a 5.0-degree limit and whether a missing hour prevents an all-clear. A line chart can show the observed sequence and a threshold, but the readings cannot prove how long the container stayed above the limit between samples. The prompt asks for observed exceedances, not continuous exposure time or product safety.
Decision: identify observed readings above 5.0 °C.
Entity: container C-47; grain: one scheduled reading per hour.
Series: 4.1, 4.4, missing, 5.8, 5.5, 4.9 °C.
Allowed claim: two observed readings exceed the limit.
Unsupported claim: hours continuously above the limit.Performance and operating cost
Defining M measures takes O(M) review effort before rendering, while inspecting N records takes O(N) for a direct scan. That small up-front cost avoids expensive revision after a chart communicates the wrong quantity. Do not ask the model to infer a missing unit or population from a column name alone. A prompt with a precise decision contract is shorter than a broad creative brief followed by repeated corrections.
Common Mistakes
- Do not equate two high readings with two hours of continuous exposure.
- Do not omit the unit or observation grain.
- Do not write a safety conclusion that the readings cannot establish.
Connected lessons
- Prompt engineering applications
- Prompt Engineering
- Dataset intake prompts: define a column before analyzing it
- Chart prompts: inspect axes, units, and missing series first
- Document prompts: define the reader's decision before drafting
- Chart prompts: choose marks and scales by data type
- Chart prompts: expose missing data and checked denominators
- Chart prompts: produce a reproducible spec and data lineage
- Chart prompts: review rendered and nonvisual outputs
- Project: review a cold-chain temperature chart
- Chart-creation prompt decisions
prompt engineering
chart creation