Choose Chart Scales That Do Not Exaggerate Dashboard Changes
Prepared with AI assistance by Alactic Editorial. Examples identified as hypothetical are illustrative, not customer results.
In a hypothetical dashboard, a rate moves from 48% to 52%. On an axis running from 47% to 53%, that movement fills most of the chart. On an axis running from 0% to 100%, it looks small. Neither range changes the values, but each changes the impression a reader takes away.
Choose the scale according to the comparison the reader needs to make. Start ordinary magnitude bars at zero, use a clearly labelled narrower range when a line chart needs to reveal movement, preserve elapsed time on date axes, and give directly comparable panels the same scale. Add the actual change in words. When exact values matter more than shape, consider a table.
The useful test is whether a reader can correctly describe the size, timing and limits of the change without reconstructing the evidence. All examples, dates, teams and results in this article are hypothetical, not Alactic customer results.
1. Write the comparison before choosing the range
Start a chart review with one sentence: “The reader needs to compare…” Finish it with something observable, such as current levels across teams, movement within one team, or distance from a target.
Those are different tasks. A bar chart comparing current levels relies on length to communicate magnitude. A line chart showing movement relies more on position and direction. Side-by-side panels invite comparisons even when their creator intended readers to inspect each panel separately.
- Compare levels: make equal values occupy equal positions and equal magnitudes use equal lengths.
- Inspect movement: reveal changes while making the displayed range explicit.
- Assess a target: show the target and enough context to reveal the distance from it.
- Check an exact difference: consider a table containing values, counts and changes.
Do not choose bounds to produce an emotional response. Record the comparison purpose beside the proposed scale so another reviewer can challenge whether the design serves that purpose.
2. Work through one dataset before comparing designs
The fictional dashboard tracks the share of eligible enquiries receiving a first response within an internal target. Each reported week has 200 eligible enquiries per team. The reporting definition and target remain unchanged.
- Week starting 5 January: Team A, 96 of 200, or 48%; Team B, 136 of 200, or 68%.
- Week starting 12 January: Team A, 98 of 200, or 49%; Team B, 138 of 200, or 69%.
- Week starting 19 January: Team A, 100 of 200, or 50%; Team B, 140 of 200, or 70%.
- Week starting 26 January: no extract available for either team. Counts and rates are unknown.
- Week starting 2 February: Team A, 102 of 200, or 51%; Team B, 142 of 200, or 71%.
- Week starting 9 February: Team A, 104 of 200, or 52%; Team B, 144 of 200, or 72%.
The review question is: “How did the reported rates change, and how do the teams’ levels compare?” Team A’s change is 52% minus 48%, or 4 percentage points. Team B’s change is 72% minus 68%, also 4 percentage points. The between-team gap is 20 percentage points at every observed date.
For Team A, the relative increase is 4 divided by 48, approximately 8.3%. That is not an 8.3-percentage-point increase. Put percentage-point change first when discussing changes in these rates.
The missing week also limits the conclusion. Rates rose at every successive observed point; the data do not establish that they rose every week or explain why they changed.
3. Keep magnitude bars anchored at zero
For ordinary bars representing rates or counts, use a zero baseline. Readers interpret bar length as quantity. Cropping the baseline changes the ratio between visible lengths without changing the underlying values.
In this hypothetical example, Team A’s endpoint bars should represent 48% and 52%. If the axis starts at 47%, their visible lengths instead represent 1 and 5 percentage points above that baseline. One bar looks five times as long, although the rate has not increased fivefold.
An axis-break symbol does not restore the original length comparison. If zero-based bars make a small difference difficult to see, label the difference or switch to a dot plot, line chart or table.
For a bounded percentage measure, a 0–100% axis provides full-range context. A zero-based axis with a lower upper bound can provide more detail, but its endpoint must be clear and consistent across comparable views. Do not let the top of the plotting area suggest an unstated target.
Bars explicitly showing percentage-point change are different: they can extend from zero change. That changes the measure being plotted; it does not justify cropping bars that represent the original rates.
4. Use line-chart close-ups for a defined task
A line chart does not always need a zero baseline. A full-range percentage axis can make operational movement difficult to inspect. A narrower axis can help if readers can see the bounds, units and actual change.
For Team A alone, a hypothetical 45–55% range makes the sequence from 48% to 52% easy to follow. Label it as a Team A close-up and add: “48% to 52%; up 4 percentage points across reported weeks.” Do not ask the line’s steepness to communicate magnitude by itself.
Chart proportions also matter. A tall, narrow plot makes a line look steeper than a short, wide plot with identical bounds. Keep plotting proportions consistent across comparable views and inspect the chart at its intended reading size.
A tight range becomes harder to defend when it clips earlier observations, presses ordinary variation against the chart edges, or changes automatically with every filter. If historical context matters, widen the range or provide a clearly identified context view.
The close-up answers a movement question about Team A. It cannot answer a question about Team B’s level or the gap between teams. Keep that distinction explicit in both the report layout and its review test.
5. Preserve time gaps and comparable panel scales
Dates are not merely category labels. Equal spacing between irregular observations compresses some periods and stretches others, changing the apparent pace of movement.
Here, 19 January to 2 February spans twice as much elapsed time as 5 January to 12 January. The horizontal distance should reflect that. Keep a position for the missing week rather than sliding February left to close the gap.
For this weekly report, break the line across the unavailable week and annotate: “26 January extract unavailable; no value inferred.” A straight connection could suggest uninterrupted evidence. Do not plot the missing rate as zero or silently carry the previous rate forward.
Side-by-side panels create another risk. If Team A uses 45–55% and Team B uses 65–75%, their lines occupy matching positions despite a 20-percentage-point difference in levels.
For the stated comparison, use shared 0–100% axes and direct value labels in the main view. A shared narrower range containing both series could also reveal their separation, provided the bounds are conspicuous. Independently scaled panels should be reserved for clearly labelled movement close-ups, not used as the main evidence for comparing levels.
6. Reuse this filled chart-review worksheet
This completed artifact compares treatments of the same hypothetical data. Copy its fields into a review and replace the evidence and decisions for the chart under consideration.
Purpose and evidence
- Reader task: assess reported movement and compare team levels.
- Measure: share of eligible enquiries receiving a first response within the internal target.
- Window: weeks starting 5 January through 9 February; 26 January unavailable.
- Required context: counts, percentage-point changes, missing-week disclosure and consistent comparisons.
- Inference boundary: no claim about cause or movement during the unobserved week.
A. Endpoint bars
- Candidate: Team A’s 48% and 52% bars on a 47–53% axis.
- Failure: visible lengths exaggerate the proportional difference.
- Decision: reject cropped magnitude bars; use a 0–100% axis, endpoint labels and “+4 percentage points.”
- Check: bars show comparable lengths; text supplies the exact change.
B. Team A trend close-up
- Candidate: a 0–100% line chart.
- Trade-off: full-range context, but little detail for inspecting movement.
- Decision: accept 45–55% for a dedicated, labelled close-up with endpoint values and a +4-percentage-point annotation.
- Check: readers identify 48%, 52% and +4 percentage points without estimating from slope. They recognise that this view does not compare teams.
C. Date spacing
- Candidate: five equally spaced observations without a missing-week marker.
- Failure: a two-week interval appears to be an ordinary weekly interval.
- Decision: use elapsed-time spacing, retain the missing-week position and break the line across the gap.
- Check: readers identify 26 January as unavailable rather than describing complete weekly evidence.
D. Team comparison panels
- Candidate: Team A at 45–55%; Team B at 65–75%.
- Failure: matching positions obscure different levels.
- Decision: use shared 0–100% axes, matching plot dimensions, identical dates and endpoint labels for the main comparison.
- Check: readers identify the 20-percentage-point gap and both teams’ 4-percentage-point increases.
E. Exact-value summary
- Candidate: another chart for endpoint arithmetic.
- Decision: use the following table when readers need to check or quote figures.
- Check: rates, underlying counts and changes can be read together; the missing-week note stays attached.
| Team | First reported rate | Last reported rate | Change |
|---|---|---|---|
| A | 48% — 96/200 | 52% — 104/200 | +4 percentage points |
| B | 68% — 136/200 | 72% — 144/200 | +4 percentage points |
7. Annotate the limits, not an unsupported explanation
An annotation should resolve a likely misunderstanding. Useful annotations identify cropped ranges, missing observations, changed definitions or the exact difference under discussion.
For the hypothetical report, “Both teams increased by 4 percentage points between the first and last reported weeks” is supported. “The new process improved performance” is not supported by these values alone. An event marker can show timing, but proximity does not establish causation.
Choose a table when exact values, denominators or exceptions drive the decision, especially when a chart needs several qualifications before its pattern can be trusted. Preserve the same evidence window and missing-data disclosure. Unknown values should not become unexplained blanks.
8. Test the complete comparison and the close-up separately
Ask a colleague who has not seen the design discussion to answer these questions using the complete proposed report or main comparison view:
- What were each team’s starting and ending rates? State each change in percentage points. If a relative percentage change is shown, explain how it differs.
- Which week is missing, and what claim about weekly movement does that prevent?
- Can the teams’ levels be compared directly? State their gap at the observed dates.
For this hypothetical review, passing means all three answers are correct: A moves from 48% to 52%, B from 68% to 72%, both gain 4 percentage points, 26 January is unavailable, and B is 20 percentage points higher at observed dates. Team A’s approximately 8.3% relative increase is distinct from its 4-percentage-point change.
Test the Team A-only close-up against its narrower task: can the reader state 48%, 52% and +4 percentage points, identify the missing week, and recognise that the view cannot establish the between-team gap? Do not require a single-team view to answer the team-comparison question.
Record answers rather than asking whether the chart looks clear. If movement is exaggerated, repair the baseline or labels. If the gap is missed, standardise panel ranges. If readers infer uninterrupted improvement, strengthen the missing-week treatment. Retest the specific failure before publishing.
Put the next conversation to work.
Explore Alactic Communications for WhatsApp conversations, campaigns and AI-assisted workflows. Choose the product and plan that fit your team.
Explore Communications