In this article
A story should make analysis understandable, not replace analysis. The four techniques below are designed for workplace updates in which a number is accurate but its meaning remains unclear.
Use data storytelling techniques to answer a decision question
Duarte recommends connecting data with the people and context behind it. Use that as a prompt to investigate the work represented by the measure.
Before choosing a narrative, write the decision the analysis might support. The same data may inform staffing, process design or further investigation, but those decisions require different explanations.
Keep descriptive findings separate from causes and recommendations. The reader should know when you move from what happened to what you think it means.
1. Name the unit behind the number
Explain what is counted: customers, orders, requests, visits or completed tasks. Clarify whether one person or event can appear more than once.
A measure labelled “engagement” is not self-explanatory. State the actual action and how it was recorded.
Hypothetical example: A training report shows more attendance records than employees. Before presenting this as broad participation, check whether employees attended several sessions.
The story changes when the unit becomes clear. You may be describing repeat attendance rather than wider reach.
2. Choose a comparison that answers the question
A number gains meaning through a relevant baseline: an earlier period, a target, a comparable group or a workload-adjusted rate.
Check whether the comparison is fair. Different definitions, seasonal conditions or missing data can make a neat chart misleading.
Hypothetical example: One team closes more requests than another, but also receives a different mix of simple and complex work. A raw count may not answer the performance question.
Explain what can be compared and what cannot. If the comparison is weak, use it to identify a question rather than rank people.
3. Test the explanation against the work
Talk to the people who understand the process, review records and consider alternative explanations. A pattern on a dashboard does not reveal motive by itself.
A decline in repeat requests might reflect better service, fewer customers or a changed recording rule. Investigate before choosing the most flattering story.
Ask what evidence would distinguish those explanations. The answer may point to a small follow-up analysis or a useful interview.
Use language that matches your confidence: observed, reported, suggests or supports. Avoid “proves” when the analysis cannot support it.
4. Connect the finding to a proportionate next step
Choose an action that fits the strength of the evidence. A clear causal finding may support a change; an unexplained pattern may support an investigation.
Hypothetical example: The team finds that incomplete requests cluster around one form field. A proportionate next step is to test clearer wording and monitor the same measure.
It is premature to promise faster service across the organisation. The trial still needs to show whether clearer wording changes submissions and whether other constraints remain.
Name the owner, scope and review condition. The story should end in a decision someone can evaluate.
Build a one-page interpretation note
Use five headings: question, measure, comparison, interpretation and next step. Add a limitation beside the claim it affects.
| Heading | Essential content |
|---|---|
| Question | What decision or uncertainty prompted the analysis |
| Measure | Unit, definition, period and source |
| Comparison | Baseline and reason it is suitable |
| Interpretation | Finding and plausible alternatives |
| Next step | Action, owner and review condition |
This note can become a presentation outline, a written briefing or the basis of a discussion. It is easier to review than a narrative assembled before the reasoning is clear.
Keep a counter-story beside the preferred explanation
Write the strongest plausible alternative interpretation in a separate paragraph. This is not a rhetorical trick; it is a way to test whether your explanation has earned its confidence.
Ask which observation would distinguish the two accounts. Perhaps you need a different time period, a workload breakdown or a check of the recording process.
If the available evidence supports both, say so and choose a proportionate next step. You may need investigation rather than a firm recommendation.
Do not invent an alternative that nobody informed would take seriously. The exercise is useful only when it tests a real uncertainty.
When presenting, explain the alternative briefly and why you favour your current interpretation. If new evidence could reverse that judgement, name it openly.
Avoid a forced hero and villain
People can help make data concrete, but do not reduce a complex problem to one heroic employee and one obstructive department.
A process, incentive or information gap may explain the result better than a personality. Use human examples to clarify the mechanism, not assign unsupported blame.
If a customer example illustrates the pattern, state that it is an example. Do not imply that one interview represents an entire population.
Explore Storify and storytelling in presentations for communicating the finished interpretation. Storytelling in meetings can help you discuss uncertainty openly.
For your next dashboard update, apply the four techniques to one measure. If the explanation changes when you clarify the unit or comparison, revise the story before presenting the chart.
AI-assisted researched guidance from PeakBelt. Hypothetical examples are labelled in the text. See our editorial approach.