Data Visualization Principles

Choose a chart for the question

Choose a chart for the question

A chart maps a question to visual encodings. Bars compare magnitudes, lines show ordered change, and scatterplots examine relationships between numeric observations. Begin with the reader’s task, not the most decorative chart. April and May provide only two observations, so a column comparison is more honest than suggesting a long trend. Use a zero baseline for value bars because length encodes magnitude. A truncated line axis can sometimes be justified, but label it and consider whether it exaggerates a small change. Do not use a three-dimensional chart whose perspective alters perceived lengths.

Follow a concrete example

Write a chart specification with reader, question, population, unit, category order and comparison baseline. The title should describe what is measured, not declare an unsupported cause. Exact labels and an adjacent table let a reader verify a small EUR10 difference without estimating pixels. A target line is omitted because no authoritative target was supplied.

Further reading: Microsoft model relationships, Microsoft CALCULATE, W3C complex images. The case and exercises are authored for DataCareerHub.

Your task

Sketch a two-bar completed-value chart and a matching data table. State the population, period, unit and source.

Use learner_worksheet.txt and the provided source/code files. Save your own artifact and compare the separate solution after attempting the task.

Build visual hierarchy and accessible alternatives

Build visual hierarchy and accessible alternatives

Hierarchy helps a reader locate the question, headline measure, comparison and limitations in that order. Reserve strong colour for a meaningful distinction rather than every decorative element. Use text, position and shape as well as colour, so a red/green distinction is not the only signal. Provide an adjacent table for exact values and a concise text explanation of what the chart shows. Keyboard focus and readable contrast matter for filters and links as much as for chart labels. Alternative text should describe the finding and scope, not merely say “chart.”

Follow a concrete example

Test the report in grayscale or with the colour meaning removed. The reader should still identify April and May from labels and position. For a chart image, provide a nearby description that states the values and direction of change; for an interactive view, label controls and expose updated scope in text. Alternative text is not a substitute for correct numbers.

Inspect the reference implementation

Language: Python 3.12. The complete downloadable file includes the named helpers and the fresh-output command-line runner. Trace the source, transformation and result against this lesson’s contract before running it locally.

def bi_measures(source,out):
 data=sales(source);groups={}
 for month in ['2026-04','2026-05','2026-06']:
  subset=[r for r in data if r['order_date'].startswith(month)]
  n=len({r['order_id'] for r in subset});value=sum(r['value_cents'] for r in subset)
  groups[month]={'completed_orders':n,'line_value_eur':value/100 if subset else None,'average_order_eur':value/n/100 if n else None,'coverage':'available' if subset else 'no source coverage'}
 return {'months':groups,'grand_average_eur':1190/21,'arithmetic_mean_of_monthly_aov':(600/11+590/10)/2,'warning':'Grand AOV must use grand value/grand orders, not mean of monthly averages.'}

Further reading: Microsoft model relationships, Microsoft CALCULATE, W3C complex images. The case and exercises are authored for DataCareerHub.

Your task

Write chart alternative text, a caption and a colour-independent comparison label. Check the table headers and filter labels in the supplied dashboard.

Use learner_worksheet.txt and the provided source/code files. Save your own artifact and compare the separate solution after attempting the task.

Audit misleading comparisons

Audit misleading comparisons

A correct number can support a misleading visual when populations, denominators or coverage change. April has eleven completed orders and May ten. Value falls slightly while average order value rises, so either headline alone omits useful context. Show counts and scope with the comparison. June is absent from the source: unavailable is not a measured zero. If you add a hypothetical month or target, label it as such. A dashboard that hides empty-state assumptions can mislead even when its default totals reconcile.

Follow a concrete example

List the three measures side by side: April11/EUR600/AOV54.545…, May10/EUR590/AOV59. This makes the different movements inspectable. A reader deciding on capacity might care about order count, while a pricing question might require item mix and discounts. The same chart cannot justify every decision without additional evidence.

Further reading: Microsoft model relationships, Microsoft CALCULATE, W3C complex images. The case and exercises are authored for DataCareerHub.

Your task

Review a claim that “performance improved because AOV rose.” Rewrite it with both value and order counts and identify missing evidence.

Use learner_worksheet.txt and the provided source/code files. Save your own artifact and compare the separate solution after attempting the task.