Design around the reader’s task
A regional manager does not need to admire your reporting software. They need to understand what happened, how much confidence to place in the numbers and what to inspect next. Begin with a written task: compare known net sales between North and South in the illustrative January 2026 export, then identify the data check needed before making a broader recommendation.
The same structure works in Microsoft Power BI, Tableau or Looker Studio. Those tools offer different data models and publishing controls, but the design decisions precede the software. Sketch the question, headline metric, comparison and qualification on paper before connecting a data source.
Use a small, explicit hierarchy
Our dashboard begins with a period and scope statement. Three indicators follow: known net sales, complete orders and order-count coverage. A horizontal bar chart compares regions. A short note states that one North order has an unknown sales amount. This hierarchy allows a reader to notice both the result and its limits without opening a separate methodology document.
January 2026 illustrative export · Currency units
Common zero baseline; bar lengths are proportional to known net sales. One North order is incomplete.
The bars share a zero baseline, so the visible length ratio matches the values. Their direct labels remove the need to alternate between a legend and the plot. The text caption provides the same information without relying on color, bar length or visual inspection.
Choose the chart for the comparison
For two categorical totals, a bar chart is usually sufficient. Use a line chart when the question concerns change over a meaningful ordered time axis. Avoid turning these few January records into an apparent monthly trend: the export does not establish complete daily coverage, and there is no comparison period.
A pie chart could show regional shares of known net sales, but it would make the magnitude comparison harder and could suggest a complete total. A map is unnecessary merely because the labels are regions. Choose a visual because it reduces the reader’s effort, not because the software makes it available.
Keep definitions within reach
Write a small metric dictionary. Specify that net sales equals sales minus refunds for complete orders; the reporting period is January 2026; the export is illustrative; and coverage is complete orders divided by unique orders. Name the source table and document the duplicate decision from the spreadsheet lesson.
In a production dashboard, also identify the refresh schedule, latest successful refresh time, time zone and person responsible for questions. Do not display a “live” badge unless the data actually updates live. A filter should make its active state visible, and the denominator should change consistently with the filtered population.
Explain what the numbers permit
North’s known net sales are twice South’s in this export. That statement describes the displayed records. It does not establish that North is more profitable, that its marketing caused higher sales or that South should lose investment. Costs, customer mix, exposure and historical patterns are absent.
The four complete order values are 100, 70, 120 and 40. Their mean is 82.5 and their median is 85. These summaries describe the observed complete orders only. With so little information and no defensible sampling design, attaching a confidence interval would not automatically create a reliable population estimate.
Put the limitation close enough to the finding that it cannot be mistaken for an afterthought.
Missingness deserves particular attention. A105 belongs to North, so North’s coverage is two of three orders while South’s is two of two. You cannot assume the missing amount resembles the observed ones. Recover the amount if possible, then rerun the summary. If it cannot be recovered, report the known total and the missingness clearly rather than producing an unsupported estimate.
Turn the display into a responsible recommendation
A defensible recommendation is modest: verify the incomplete North order and confirm that both regions have comparable extraction coverage before reviewing operational causes. If the source owner confirms full coverage later, add a longer comparison period and relevant exposure measures. A lower total might reflect fewer orders rather than weaker performance per opportunity.
Replace “North performs twice as well” with a sentence that names the metric, scope and qualification. Then add one action that would reduce uncertainty before a business decision.
Self-check: what should a reader learn in ten seconds?
Known net sales are 330; North accounts for 220 and South for 110; one of five unique orders lacks a sales amount. The next action is a source-data check, not a causal claim.
Before publishing, test the dashboard on a narrow screen, with keyboard navigation where controls exist, and in a text-only reading of the captions. Then package the specification, queries and interpretation as evidence of your working practice.
Next studyProjects: Show Your Reasoning →