The dashboard has another colourful chart. Nobody is sure whether it shows more work completed, more work started or simply more rows entered into the system.
Ask AI to suggest a chart only after defining the comparison you want the reader to understand. Specify the measure, population, period and source. A chart should communicate an evidenced relationship, not create the impression of analysis through colour and shape.
Photo by Berke Citak on Unsplash.
Write the message in plain language
Describe what the chart should help somebody see. If the proposed message is not supported by the data, revise it before choosing a chart type.
For an illustrative delivery review, the question might be whether completed cases changed across several weeks. That differs from comparing waiting time between stages or showing the mix of enquiry types.
Government Analysis Function guidance recommends starting with the message and statistical relationship when choosing a chart. The business worksheet here turns that principle into a brief for AI-assisted preparation.
Give the chart a specification
Record:
Reader and decision supported.
Definition of each measure.
Included population and exclusions.
Time period and comparison point.
Units, labels and source reference.
Missing values and qualifications that must be shown.
Ask the assistant to explain why its proposed chart suits the relationship. Do not accept a format solely because the tool can produce it quickly.
An illustrative volume comparison
Suppose one week contains four working days and another contains five. A comparison of total completed cases may be valid for one question, while an average per working day answers another.
Choose the measure deliberately and label it accurately. Do not let the assistant switch between the two to produce a stronger-looking trend.
If incomplete cases are excluded, say so. A neat line does not explain whether the result represents all work or only the cases that reached an agreed finish.
Keep the data connection inspectable
Retain the values, filters and calculations supporting the chart. If the AI tool returns only an image, ensure somebody can still verify the numbers in an appropriate working file.
Check the axis, category order and legend. A correct number can be misleading when its label or comparison is wrong. Where a visual exaggerates a minor difference, choose a clearer presentation.
Do not describe an observed change as an improvement caused by an intervention unless the evidence supports that conclusion.
Make the explanation accessible
Provide a short text description of what the chart shows and the important limitation. Avoid using colour alone to identify the crucial difference.
Ask a colleague unfamiliar with the analysis what they understand from the result. If their interpretation differs from the intended message, revise the chart or the question.
A useful chart leaves the reader better able to decide what to investigate or change. It should not require a separate performance of explanation every time it appears in a report.
For related practical guidance, see agreeing the meaning of metrics and measuring process improvement honestly.
Hili Consulting can help specify a truthful chart comparison before choosing a visual style. Arrange a free 30-minute consultation and bring one example of the work you want to improve.
