Key takeaway
Hand off the data before the styling: source data, units, time range, definitions, source, and calculation. Annotate key points and explain conclusions and limits in text. Label demo data as hypothetical and never present schematic numbers as real performance. Color must not be the only way to distinguish series.

Take this illustrative scenario. A business report page holds six charts: a growth curve, regional distribution, channel share, customer structure, monthly comparison, and trend forecast. The colors look good. In the meeting someone asks: is this growth by order count or revenue, and does it include refunds? The chart maker opens three versions of the spreadsheet and cannot say which one was used.
Chart problems are rarely about "looking bad." More often the source is unclear, definitions conflict, units and time are missing, and key points are unannotated. Readers fail not because of taste but because the evidence is incomplete.
A chart is evidence, not decoration
A chart should let readers answer: what this number is, where it comes from, how it is defined, what period it covers, what it shows, and what it does not.
If those answers live only on the maker's computer, the chart is merely attractive. On an external page or formal report, it invites misreading and wrong business judgment.
Hand off six things before design
Source data: a spreadsheet or database export, not a screenshot. Record-level data that can be recalculated.
Unit: yuan or ten-thousands, pieces or units, percent or percentage points. A wrong unit changes magnitude by ten or a hundred times.
Time range: calendar month, rolling 30 days, through a specific date. Cross-time-zone business states the zone.
Definition: by order or by customer, refunds included or not, deduplication, status scope. Different definitions produce different numbers.
Source: business system, general ledger, research report, or manual count. External data states publisher and publication date.
Calculation: how growth is computed, what the denominator of a share is, what assumptions a forecast uses. Write the formula so others can check.
Not all six need to be printed on the chart, but the maker must have them and explain the key ones in the caption or body.
Choose the right chart
Lines fit trends over time, bars fit comparisons across categories, pies fit a few categories' shares, scatter fits relationships between two variables. Ask what the reader needs to compare before choosing.
Do not use dual axes to manufacture a dramatic crossing; do not put a dozen nearly equal slices in a pie; do not hide data behind 3D effects. A chart exists to make comparison clear, not to make the picture complex.
Color and readability
Under WCAG 2.2 Success Criterion 1.4.1, color must not be the only visual means of conveying information, indicating an action, prompting a response, or distinguishing a visual element. For charts specifically:
- Distinguish series with color plus shape, line style, or pattern.
- Annotate key points with text, not only a color highlight.
- Mark positive and negative values with signs and words, not only red and green.
- Use color-blind-friendly palettes; avoid red-green as the only contrast.
The chart should remain readable in black-and-white print or a forwarded screenshot.
Text explanation and annotation
The chart needs accompanying text: what question it answers, the key conclusion, the limitations. Do not leave the conclusion only in the title, and do not say "as shown" and let readers guess.
Annotate key points directly: peaks, turning points, outliers, missing data. State time and value, not just an arrow.
Demo data and real data
Schematic numbers on a page or in a demo must be labeled "hypothetical" or "sample for illustration." Do not use realistic-looking numbers as performance, and do not forget to replace them after the demo.
Real data needs a source and time. External research states publisher, sample, and definitions; do not mix years and samples. Without data, explain how to measure rather than invent a number.
A chart handoff table
Item | Example content | Provided by | Verification |
|---|---|---|---|
Source data | Order detail export | Data owner | Re-aggregate to match the chart |
Unit | Ten-thousand yuan, tax included | Finance | Match report definitions |
Time range | January–October 2026 | Data owner | Match filters |
Definition | Paid orders, excluding refunds | Business owner | Check one order |
Source | Order table in the business system | Data owner | Source traceable |
Calculation | YoY = (current − prior) / prior | Analyst | Recalculate from source data |
Fill the table before design. Reverse the order and the design gets dragged by data until nobody knows the version.
When a chart is updated, handle the old image and old data together. After replacing it, check the caption, body references, downloadable files, and share image. If another page still uses the old chart, update it too or mark it as historical. Alt text must change with the chart; updating only the image leaves screen-reader users with the old description. Keep the previous data file with the previous chart version so a published figure can still be explained later. Name the file with the period it covers.
The file name should match the chart version. Keep them together.
Trace the Data Before Handing Off a Chart
[Page chart evidence checklist]
1. Data
- Is there a source data file rather than a screenshot:
- Are units, time range, and definitions stated:
- Are source and calculation checkable:
2. Chart
- Does the chart type match what readers compare:
- Do axes start at zero or clearly show truncation:
- Are dual axes, 3D, and pie slices used appropriately:
- Are key points annotated in text:
3. Readability
- Are series distinguished by color and shape/line style:
- Is the chart readable in black-and-white or a screenshot:
- Are positive, negative, and outlier values explained in words:
4. Explanation and labels
- Does nearby text state conclusions and limits:
- Do external numbers name publisher, date, and sample:
- Is demo data explicitly labeled hypothetical:Before publishing, have someone who did not make the chart read only the chart, captions, and body, then answer: what is this number, where is it from, and does it support the conclusion? If they cannot, the chart lacks evidence, not decoration.
How to compare external research definitions is in real numbers on enterprise AI adoption; how content becomes easier to cite is in what content AI is more likely to cite. The scope of charts and content production can be discussed within content and interactive support.
Use one fact source and version across the web, infographic and slide deck, then verify the accompanying copy.