Writing a Data Insight Narrative
Practice structuring written narratives around data insights for engineering and product audiences.
Data insight narrative structure
- Opening: "The data shows that..." — key finding first, quantified and scoped
- Evidence: numbered data points with metric names, values, and timestamps
- Transition: "This suggests that..." — interpretation, not assertion
- Limitations: "however, this data does not account for..." — named, quantified, directional
- Recommendation: "Based on the above analysis, we recommend..." — specific, with a next step
Question 0 of 5
Which sentence best opens a data insight narrative for an engineering audience?
"The data shows that..." is the standard opening for a data insight narrative — lead with the key finding immediately.
- Key finding first — "API error rates increased 3×" — the reader knows the headline before they have read one more word.
- Quantified — "3×", "82%" — numbers anchor the claim. "Some errors" is meaningless.
- Scoped — "mobile clients", "v2.3 deployment", "Tuesday" — the reader knows where to look and when it started.
- Context — "82% of all errors this week" — gives the insight a sense of scale.
Which passage correctly provides supporting evidence after a key data finding?
Supporting evidence must enumerate specific, numbered data points — each with a metric name, values, and a time dimension.
- Numbered list — "(1), (2), (3)" — makes the evidence easy to reference in a thread or document review ("disagree with point 2").
- Specific metric —
/api/v2/checkout, p95 latency, error log volume — not generic "performance" or "errors". - Before/after values — "420ms to 1,840ms" — the contrast makes the magnitude tangible.
- Time stamps — "14:00 to 15:30 UTC", "14:03 UTC" — enables correlation with other events (deployments, traffic spikes).
- Causal link — "coincides precisely with canary deployment of build #4821" — this is evidence, not assertion.
Which sentence correctly uses "This suggests that..." as a transition in a data narrative?
"This suggests that..." introduces an interpretation of the data — a hypothesis, not a conclusion. The word "suggests" is deliberate:
- Epistemic honesty — "suggests" signals that this is the most plausible interpretation, not proven fact. Use "confirms" or "proves" only when you have controlled evidence.
- Scoped interpretation — "isolated to mobile clients on the v2.3 SDK" narrows the problem space, directing investigation.
- Ruled-out alternative — "rather than a server-side issue" — showing what the data rules out is as valuable as showing what it supports.
- Evidence cited — "given that desktop error rates remain normal" — the suggestion is backed by a specific data point, not intuition.
Which sentence correctly acknowledges a data limitation in a written narrative?
Limitation acknowledgements must name the specific gap, quantify it if possible, and state the direction of bias it introduces.
- "however" — the transitional word signals that what follows qualifies the preceding claim.
- Specific gap named — "users who disabled analytics tracking" — not just "some missing data".
- Quantified — "12–15% of our mobile user base" — turns a vague caveat into an assessable uncertainty.
- Direction of bias stated — "the true error rate may be higher than reported" — tells stakeholders which way the estimate is skewed, which affects decision confidence.
Which sentence best writes an actionable recommendation at the end of a data narrative?
A recommendation closing a data narrative must specify the action, the actor, the target, and the next step.
- "Based on the above analysis" — explicitly ties the recommendation to the evidence. Not "I think" or "we feel".
- Specific action — "rolling back build #4821 to #4820" — not "fix this" or "revert something". The exact build numbers mean the engineer can act without asking for clarification.
- Time qualifier — "immediately" — conveys urgency when warranted.
- Next step scoped — "root-cause investigation with the platform team before re-deploying" — a recommendation without a next step is a complaint, not a proposal.