In this article
- Start data storytelling with the decision context
- Separate finding, explanation and recommendation
- Show what the decision depends on
- Compare action, investigation and waiting
- Use one hypothetical decision memo
- Prepare for the strongest reasonable objection
- Agree on a stop or revision condition
- Turn the decision into a reviewable commitment
For analytical audiences, evidence and limitations are part of the story. A clear account of what remains unknown can help people decide whether to act, investigate or wait.
Start data storytelling with the decision context
Duarte discusses balancing narrative and evidence for analytical audiences. A useful workplace application is a short decision memo that keeps the evidence beside the recommendation.
Name the decision-maker, the available choices and the timing. “Sales changed” is a finding; “whether to trial a different replenishment rule” is a decision.
Do not assume every insight requires immediate action. Some findings are too uncertain, too small or too disconnected from current priorities to justify a change.
Explain why this finding deserves attention now.
Separate finding, explanation and recommendation
Use three distinct statements.
The finding describes what the data shows. The explanation describes why you think it happened. The recommendation describes what you suggest doing.
Hypothetical example: A team observes more incomplete submissions after a form revision. That is the finding. Ambiguous wording may be an explanation. Testing alternative wording is a recommendation.
The timing alone does not establish causation. Other changes may have occurred, including a different user group or revised recording practices.
Keeping the three statements separate prevents a plausible explanation from becoming an unquestioned fact.
Show what the decision depends on
Identify the assumptions that would change the recommendation. Perhaps the pattern needs to persist across another period, or the proposed change depends on technical capacity.
Make these dependencies visible in the main discussion. Do not hide them in a small footnote while presenting the recommendation as certain.
A useful phrase is: “This option is preferable if…” followed by the condition. Then show what you would do if the condition is not met.
This helps an audience choose responsibly without requiring a false sense of certainty.
Compare action, investigation and waiting
A data-led recommendation often has more than two options. Include the possibility of further investigation or a smaller reversible trial.
| Option | When it may fit |
|---|---|
| Implement a change | Evidence is strong and the downside is acceptable |
| Run a limited trial | The mechanism is plausible but important uncertainty remains |
| Investigate further | The pattern is real but explanations are unresolved |
| Wait and monitor | Immediate action is disproportionate to the evidence |
These are decision prompts, not universal rules. Safety, contractual or regulatory requirements may override a general preference for testing.
State the consequence of waiting as carefully as the benefit of acting. Avoid inventing a dramatic cost of inaction.
Use one hypothetical decision memo
Hypothetical example: An internal service team is reviewing a request form.
Question: Should the team revise the field used to describe urgency?
Finding: Reviewers record inconsistent interpretations of the current field.
Interpretation: The wording may combine priority and deadline in a way users understand differently.
Options: Keep the field with clearer guidance, split it into two fields or test a revised definition with one group.
Recommendation: Trial a revised definition before changing the form for everyone.
Review: Check whether users and reviewers interpret the field consistently, while watching for extra completion effort.
This memo makes the next step visible without claiming a measured outcome that does not exist.
Prepare for the strongest reasonable objection
Ask what evidence a sceptical but informed colleague would request. They may question the baseline, representativeness, mechanism or implementation cost.
Answer the objection directly. If you cannot, narrow the recommendation or identify the missing investigation.
Do not equate disagreement with a failure of storytelling. The audience may understand the narrative perfectly and still reject the proposal for sound reasons.
A useful presentation allows the reasoning to improve under challenge.
Agree on a stop or revision condition
Before starting a trial, decide what evidence would lead you to stop, revise or continue it. Without that agreement, a team can reinterpret every outcome as a reason to keep going.
The condition should relate to the original problem and relevant downside. A clearer form that creates an unreasonable completion burden may need revision even if it improves one field.
Choose a review period that gives the trial a fair chance without allowing indefinite drift. The appropriate period depends on workload and context, not a universal calendar rule.
Record any change to the criteria during the trial and explain why. Otherwise the final story may quietly move the goalposts.
A responsible data narrative includes the possibility that the recommended action will not work as expected.
Turn the decision into a reviewable commitment
Record what was agreed, who owns it and what would trigger reconsideration. Keep the measure consistent enough to assess the change.
If the intervention changes how the data is collected, explain that the before-and-after comparison may need adjustment.
Explore Storify, storytelling in presentations and storytelling in meetings for communication support.
For your next analytical update, draft the decision memo before making slides. If the action cannot be justified in plain language with its limitations visible, a more polished story will not make it ready.
AI-assisted researched guidance from PeakBelt. Hypothetical examples are labelled in the text. See our editorial approach.