In this article
For a manager, the practical starting point is a repeatable review rhythm. Make it normal to ask what a number represents, what could explain it and what evidence would justify a response.
Build data culture around shared questions
Duarte's account of Freeman describes an organisation developing strategic data storytelling. It is a practitioner case account, not proof that a particular training approach will produce the same outcomes elsewhere.
For your team, begin with a smaller objective: make one recurring review easier to understand and challenge.
Use a consistent set of questions: What changed? How do we know? What might explain it? What should happen next? Who will check?
The questions should apply to favourable findings as well as disappointing ones. A culture that scrutinises only bad news can reward optimistic interpretations.
Create a minimum definition sheet
Choose the measures your team actually uses to decide. Record the unit, inclusion criteria, source, period and owner for each.
Terms such as active customer, completed task or overdue request often hide differences in interpretation. Resolve those differences before comparing teams.
Keep the definition sheet easy to find and update. A document nobody owns can become another source of confusion.
When a definition changes, record the date and effect on comparisons. Do not present a discontinuity as a sudden performance improvement.
Use a four-part review contribution
Ask each presenter to prepare a short contribution with finding, evidence, interpretation and next step.
The finding should be descriptive. The evidence should identify the source and scope. The interpretation should include relevant uncertainty. The next step should name an owner.
Hypothetical example: A support team notices an increase in repeat contacts.
A useful contribution explains how repeat contact is defined, which period is affected, what sample of cases was examined and which explanations remain plausible. It might recommend reviewing a particular handover rather than immediately blaming staff performance.
This structure creates a common language without requiring identical slides.
Reward a corrected interpretation
Make it acceptable to say, “Our first explanation was wrong.” If every revision is treated as embarrassment, people have an incentive to defend weak stories.
A manager can model the behaviour by explaining what new evidence changed their view. Keep the emphasis on reasoning and correction, not personal humiliation.
Distinguish a good-faith analytical error from deliberate manipulation or repeated negligence. Psychological permission to question does not eliminate accountability.
Also protect time for the work. Asking for more rigorous interpretation without reducing competing demands can create superficial compliance.
Give different roles a voice
Analysts may understand the measure, frontline colleagues may understand the process and managers may understand the decision constraints. A useful review needs those perspectives to meet.
Invite people to challenge the interpretation with observations or records. Do not make job seniority the deciding factor in whether an explanation is accepted.
Ask quieter participants for input without forcing public disclosure. Written questions before a meeting can help reveal issues people may hesitate to raise aloud.
If the review concerns employee performance, use appropriate privacy and management processes. A public data discussion should not become an improvised disciplinary forum.
Track decisions, not just dashboards
Keep a lightweight decision log.
| Field | Purpose |
|---|---|
| Question | Records why the analysis was requested |
| Decision | States what was agreed |
| Evidence | Identifies the basis at the time |
| Owner | Makes follow-through clear |
| Review condition | States when to revisit the choice |
This helps the team distinguish an insight that changed work from a chart that was merely interesting.
Review the log occasionally for repeated unresolved questions. They may indicate a missing definition, inaccessible data or an ownership problem.
Prevent the review from becoming a performance theatre
A recurring data meeting can reward polished slides while leaving difficult questions untouched. Reduce that risk by asking presenters to include one uncertainty and one decision they need help with.
Allow the group to inspect the definition and source rather than judge the confidence of the speaker. A less fluent colleague may have a stronger analysis.
Keep the review focused on a manageable number of measures. Adding more charts can dilute attention and increase preparation work without improving decisions.
Notice whether the same person always interprets the data. Rotate contributions where appropriate and provide support for people unfamiliar with the format.
End by recording what changed in the team's understanding. If nothing changed and no decision followed, ask whether the meeting needs a different purpose or less frequent cadence.
Start with one recurring meeting
Choose a meeting where numbers are presented regularly. Add the definition sheet and four-part contribution for a limited period.
Ask participants whether the discussion became clearer and whether decisions were easier to trace. Do not declare a cultural transformation after one session.
Explore Storify and storytelling in meetings for improving the discussion. Storytelling in presentations can help when findings need wider communication.
The next step is simple: select one measure and ask the team to define it independently. Differences in their answers will show where your data culture needs practical work before it needs better storytelling.
AI-assisted researched guidance from PeakBelt. Hypothetical examples are labelled in the text. See our editorial approach.