1. Write the question in one sentence
“Show the data” is not a useful brief. Write what the reader should learn: which region is largest, how demand changed, where customers leave a process, or whether two measurements move together.
2. Prepare one clean table
Give every column a descriptive header, keep one observation per row, remove decorative merged cells, and state units. Decide how to represent missing values instead of silently converting them to zero.
3. Choose the encoding
Use bars for categories, lines for ordered trends, scatter plots for paired numeric values, and pie or donut only for a few parts of a meaningful whole. Specialized questions may justify a funnel, gauge, radar, or Gantt chart.
4. Build and simplify
Import the data, add an informative title, choose accessible colors, and remove decoration that does not help interpretation. Preserve labels and context even when the chart feels self-explanatory to its creator.
5. Review the claim
Check the baseline, scale, time interval, denominator, sorting, and omissions. Ask a colleague to describe the conclusion without being told what to look for. If they reach a different conclusion, revise the chart or its explanation.