Reporting and Communication

Reporting and Communication

Taxonomy context: Data Analysis > Reporting and Communication

Reporting and Communication is the phase that converts analytical work into decisions. Every upstream phase — problem framing, data preparation, modeling, evaluation — is wasted if the audience cannot understand the finding, cannot trust its rigor, and cannot act on the recommendation. Reporting is not a write-up of what the analyst did; it is an argument constructed for a specific decision-maker.

The full skill body (12 sections + worked example + full sources) lives at the canonical install path: ~/.claude/skills/da-9-reporting-communication/SKILL.md with references/worked-example.md and references/sources.md.

Section index

  1. The bottom-line-first inversion — BLUF, Minto Pyramid, SCQA (Situation/Complication/Question/Answer).
  2. Data storytelling (Knaflic) — context-aware framing, chart selection, clutter elimination, preattentive attributes, narrative arc, action titles.
  3. Audience adaptation — same finding three ways: executive (60-90s decision), technical (verify and reproduce), operational (Monday-morning rule).
  4. The executive summary one-pager — six fixed slots, the “so what” test, headline metric framing.
  5. Technical report structure — IMRaD adaptation; canonical sections; reproducibility appendix minimum contents; Limitations sections done well.
  6. Recommendation framing — evidence/opinion separation, confidence levels (high/moderate/low), alternatives considered, ownership.
  7. Notebook-as-report — Jupyter, Quarto, Observable, R Markdown; six conversion patterns; what notebooks should not be used for.
  8. Dashboards and live reports — when a dashboard is the right answer, structural rules, dashboards vs exploratory tools.
  9. Honest framing of uncertainty — what the data does and does not say, uncertainty taxonomy (sampling / measurement / model / data quality / generalization), the “what we did not find” paragraph.
  10. Anti-patterns — burying the lede, decoration over information, topic titles, false precision, single-source confirmation, reverse-engineering the conclusion, AI-generated tells, method-section autobiography, confidence-by-omission, no-owner recommendations.
  11. Worked example — same analysis rendered as executive one-pager, technical write-up excerpt, and operational runbook.
  12. Quick reference — pre-send checklist, format-by-audience cheat sheet, five-minute report-quality test.

Sources

Foundational references: Knaflic (Storytelling with Data); Minto (The Pyramid Principle); Few (Now You See It, Information Dashboard Design); Tufte (Visual Display of Quantitative Information); U.S. ODNI Analytic Standards (ICD 203) on confidence-level language; U.S. Army BLUF doctrine; Quarto / Project Jupyter / Wickham & Grolemund R for Data Science on notebook reports; Doumont Trees, Maps, and Theorems; HBR on executive summaries; Heath & Heath Decisive; Silver The Signal and the Noise.

The full SKILL.md (621 lines) carries the complete instructional content; this hub mirror is an index.