/data-narrative-builder

Data Writing Guide

Writing about numbers

Contextualise every number. A number without context is noise. Add:

• Comparison (vs last period, vs target, vs benchmark)
• Direction and magnitude ("up 12%", not "12%")
• Significance ("material" or "within normal variation")

Round appropriately. Report to the precision that is meaningful:

• Revenue to the nearest thousand (or million at scale)
• Rates to 1–2 decimal places
• Counts to the nearest whole number
• Very small percentages (e.g., error rates) to 2–3 significant figures

Avoid false precision. "Revenue grew by 7.43821%" signals you don't understand the uncertainty in your data.

Sentence-level guidance

Active voice for findings, passive for process.

• Finding: "Conversion rate declined 3 points in EMEA" (active — clear subject)
• Process: "Users were segmented by plan tier" (passive — subject is the analyst, irrelevant)

Numbers at the start of a sentence: Spell them out. "Twelve percent of users…" not "12% of users…" at the start of a sentence.

Quantify comparisons. "Revenue increased significantly" is vague. "Revenue increased 18% — twice the prior-quarter rate" is precise.

One idea per sentence. Long compound sentences that combine a finding, its context, and a recommendation confuse readers. Split them.

Describing trends

Framing uncertainty

Analytical conclusions always carry uncertainty. Be explicit:

• "This is consistent with…" — correlation, not causation
• "The data suggests…" — directional, not confirmed
• "We cannot rule out…" — acknowledge alternative explanations
• "We are confident that…" — reserve for validated findings
• "Subject to [assumption X]" — flag load-bearing assumptions

Never use false certainty to make a finding sound more compelling.

Chart titles and annotations

Chart title = the finding, not the description.

• Description: "Monthly Revenue by Region, 2024"
• Finding: "EMEA revenue declined every month in Q4 2024"

Annotation placement:

• Annotate the anomaly, not the obvious pattern
• Use a single callout box; multiple annotations create visual noise
• Keep annotation text under 10 words

Executive vs analytical writing style