Choosing management charts without misleading comparisons

Write the management question first

“Show sales data” is not specific enough to choose a chart. “Which department is behind target?” and “Have sales declined continuously?” require different encodings. The first compares actual and target values; the second examines time. Putting both into a decorative ring can leave readers searching for the original numbers before they can decide anything.

State the objects, period, unit and intended action, then decide whether a chart is necessary. A current balance or the state of one request may be clearer as a number with context. Empty space in a design is not evidence that another chart is needed.

Inspect the baseline when comparing size

Bar length represents quantity, so a truncated baseline can magnify apparent differences. The UK Office for National Statistics guidance on axes recommends zero baselines for bar and area charts, while allowing narrower ranges for line and scatter charts where useful. It also warns that dual axes can mislead. These are design references, not evidence of results achieved in a company project.

In a hypothetical comparison of ninety-eight and one hundred, starting bars at ninety-seven makes the gap look much larger. If small changes are the point, a clearly labeled dot or line chart can expose them while retaining the actual difference. Do not obtain that effect by quietly changing a bar baseline. The numbers here exist only to explain the visual decision.

Challenge dual axes and percentages

Two lines with independently scaled axes can appear to move together. First determine whether the reader needs absolute values, relative change or timing. Separate charts sharing a time axis may be clearer. Indexed change can also help if its baseline is explained. Similar-looking curves alone do not establish causation.

A percentage needs a stated denominator. Categories may overlap, one object may carry several labels, and unclassified objects may still belong to the total. Those facts determine whether a composition chart is appropriate. Do not force values to add to one hundred by removing genuine overlaps or inventing a balancing category merely to complete a pie.

Do not give missing data the appearance of zero

• Zero is a confirmed numerical result and should retain that meaning.
• Missing means no usable observation is available; it should not automatically become a smooth connecting line.
• Not yet due means the period is incomplete, which differs from an abnormal final result.
• Revised values need an explanation when historical results change.

Partially updated data needs a clear time and coverage statement in the title or supporting text. Legends, axis labels and units must be readable on the intended display. Hover should not be the only way to learn what a value means, especially when the same information is used on a touch device.

Review alternative charts with identical data

1. Ask a business representative to state the apparent conclusion before hearing the designer’s explanation, then check it against the source values.
2. Compare two designs using the same data and look for contradictory impressions caused by scales or area.
3. Add negative values, outliers, missing periods and long category names. Check that the chosen encoding still makes sense.
4. Change a filter and verify that the title, denominator, unit and legend follow it.
5. Retain a way to inspect values or relevant detail instead of making the picture the only evidence.

When the metric itself is unresolved, settle its definition before repeatedly changing charts. UI design communicates confirmed data; statistical interpretation, operating thresholds and decision responsibility require business participation. A polished visual style cannot replace those decisions, and a different chart should not silently change the question the page claims to answer.

If reviewers disagree about a chart’s conclusion, settle its denominator, period and comparison group using the dashboard metrics guide. Then establish text size and density from the display conditions. Supply the confirmed sample to UI design services (Chinese) and compare whether alternative charts answer the same question, rather than selecting whichever image attracts the most attention.