Data visualisation begins with the comparison a reader needs to make. Learn to choose a chart from the question, build it from checked values, and explain what the display does and does not establish. A finished chart needs a meaningful title, readable labels, an appropriate scale, and an accessible text explanation. This roadmap develops three small views of a fictional repair service instead of asking you to build a large dashboard immediately.

Translate questions into visual tasks

List three questions: which repair category received the most requests, how requests changed over four weeks, and how long individual jobs took. These ask for category comparison, change over time, and distribution. They should not automatically receive the same chart. Before opening software, sketch what a reader should compare and write the unit beside each number.

Use a small spreadsheet to verify the inputs. Microsoft’s table guidance supports maintaining consistent records Source 1. Record whether the data counts requests, completed jobs, or customers. Those quantities can differ. A person with three repairs should not silently become three unique customers merely because the chart looks cleaner with a single headline number.

Build three views of the same fictional service

For category comparison, use 12 bicycle repairs, 8 lamp repairs, and 5 fan repairs. A simple bar chart should show 25 total requests and make bicycle repairs the largest category. Begin the quantitative bar scale at zero so bar length represents quantity honestly. Label the values directly if that improves readability, and avoid a decorative three-dimensional perspective.

For time, use weekly counts of 4, 7, 6, and 8. They also total 25. A line chart can show their ordered sequence, but four points do not establish a seasonal trend. Keep week labels explicit and say that week four had the highest count in this four-week sample. Do not turn that observation into a growth forecast.

For completed-job durations, use 1, 1, 2, 2, 3, and 9 days. A dot plot or a small frequency display makes the long job visible. The median is 2 days, while the mean is 3 days. Describe both if they serve the reader’s question. Statistics Canada’s explanation of spread provides background for examining variation Source 2; the invented example here is designed to show why a single average can conceal a long tail.

Make the explanation part of the chart

Write a sentence below each view naming the main observation and one limit. For example: “Bicycle repairs account for 12 of 25 requests in this fictional month; the data does not record time spent or profit.” This prevents request volume from being mistaken for workload or value. Keep the table of values available for readers who need exact numbers.

Use a small palette and redundant cues such as labels or line styles. Do not require a reader to distinguish red from green to understand a result. W3C’s guidance for complex images explains the need for text alternatives that convey essential chart information Source 3. A generic description such as “a graph” fails to communicate the actual finding.

Check the chart at the size it will be read. Axis labels that look fine on a large monitor may become illegible in an article card or phone screenshot. Shorten wording without removing units, increase spacing, or choose a simpler chart. Exporting a larger image alone does not solve a crowded design.

Review the claim before sharing

Ask another learner to answer the three original questions using your charts. Record where they hesitate or infer more than the data supports. If they call bicycle repairs the most profitable category, revise the title or explanation because your display may be inviting an unsupported conclusion.

Submit the source table, three charts, text summaries, and a short design note explaining one rejected chart choice. Then create a deliberate misleading version with a truncated bar axis and explain why it exaggerates the difference. Keep that version clearly labelled as a critique exercise. Being able to identify a misleading visual is part of learning to produce a trustworthy one.

Practice sequence
  1. QuestionName the comparison the reader needs.
  2. ValuesCheck units, totals, and missing records.
  3. EncodingChoose position, length, or another clear visual cue.
  4. InterpretationTest what a reader concludes.
Practice decisions and evidence
Checkpoint What to produce How to review it
Category comparison Bars for 12, 8, and 5 requests Zero baseline and total of 25
Time sequence Four weekly counts Dates ordered; no unsupported forecast
Distribution Six individual durations Long 9-day job remains visible
Text alternative Observation plus limitation Reader can recover the essential meaning
Evidence before advancement
Observed
Week four has 8 requests in the sample.
Not observed
The next month will keep growing.
Needed next
More periods and relevant context before forecasting.

Common questions

Is a pie chart always a bad choice?

No, but it is often harder to compare similar slices than aligned bars. For this exercise the aim is comparing three category counts, so bars make the task straightforward. If you use a part-to-whole chart elsewhere, make sure the categories form a meaningful whole and avoid excessive slices or decorative perspective.

Must every chart axis start at zero?

Bars use length to represent magnitude, so a zero baseline is important for honest comparisons. Other chart types may use a narrower range when the purpose is examining small changes, but the scale must remain clear. Make the choice deliberately and explain it when a reader could otherwise misunderstand the apparent difference.

How many colours should I use?

Use only as many as the meaning requires. A single colour can show all repair categories, with one darker shade highlighting a relevant comparison. Add direct labels and sufficient contrast. Colour should reinforce the information rather than serve as the only way to identify a group, status, or important result.

Should I show the mean or the median?

Choose according to the question and the distribution. In the six-duration example, the median describes the middle observation while the mean reflects the long nine-day job. Reporting both with the underlying values is useful. Do not choose whichever statistic makes the service appear faster without explaining the difference.

Can I claim a trend from four weekly points?

You can describe the observed sequence, including the final week’s higher count. Four points alone provide weak support for a broader trend or forecast. Mention the period and sample size, then identify the additional history and context you would need. A line joining points does not create evidence about future demand.

What belongs in a chart title?

A title should identify the subject and, where helpful, the finding. Include the period or population nearby and keep the unit visible. “Requests by repair category, fictional month” is more informative than “Overview.” If the title states a conclusion, ensure that the values directly support that conclusion without hidden assumptions.

Do accessible charts require long descriptions?

The description should be long enough to convey the chart’s purpose and important information, not longer by default. A simple three-bar chart may need a short summary and its values. A more complex display may need a structured explanation and data table. Test whether the essential meaning survives without seeing the image.

What should I include in a visualisation portfolio?

Include the question, source data, finished chart, text alternative, and a brief reason for the design choices. Show a revision prompted by reader feedback. This demonstrates judgement rather than software decoration. Label fictional data clearly so a viewer does not mistake your practice service for an actual organisation or verified business result.

Continue your learning

Spreadsheet Skills Roadmap: Clean, Analyse and Present a Small Dataset. Statistics for Beginners: A Learning Roadmap for Interpreting Data.

Sources and further reading