Как написать аналитический отчёт
Карьерник — квиз-тренажёр в Telegram с 1500+ вопросами для собесов аналитика. SQL, Python, A/B, метрики. Бесплатно.
Зачем это знать
Вы сделали deep analysis, нашли important insight. Но если отчёт плохой — никто не поймёт и не применит. Analysis без good report = wasted work.
Writing skills — undervalued в аналитике. Лучшие analysts — clear writers.
Структура отчёта
1. Title
Descriptive: «Churn analysis Q1 2026: mobile users drive 60% of losses».
Not «Q1 report». Title — first message.
2. Executive summary
1 paragraph.
- Problem / question
- Key finding
- Recommendation
- Impact estimate
Exec reads only this. Stand-alone.
3. Context
Why analysis? Background.
Stakeholder concerns, business goals.
4. Methodology
What data, how analyzed. Brief.
Data sources, time periods, filters.
5. Findings
Main section. Each finding:
- Statement
- Supporting evidence (chart, number)
- Interpretation
Typically 3-5 main findings.
6. Recommendations
Actionable next steps:
- What to do
- Priority
- Owner (if known)
- Expected impact
7. Limitations
Be honest. What couldn't cover.
Builds trust.
8. Appendix
Detailed methodology, additional charts, raw data.
Skip для exec version.
Шаблон (Google Docs)
# Title
## TL;DR
- Finding 1
- Finding 2
- Recommendation
## Context
Background paragraphs.
## Methodology
Data, timeframe, analysis type.
## Findings
### Finding 1: [Statement]
Chart + explanation + interpretation.
### Finding 2: ...
## Recommendations
1. **Do X:** [details]
2. **Do Y:** [details]
## Limitations
- Cover this, не cover that
## Appendix
[Additional details]Writing tips
Short sentences
Long sentences → confusing.
Active voice
«We analyzed X» > «X was analyzed».
Concrete
«Revenue fell $500k» > «Revenue declined significantly».
No jargon
Business language. «CR» → «conversion rate».
Bullets
Easier scan than paragraphs.
Headings
Scannable. Reader jump к section.
Charts в report
Each chart
- Clear title
- Label axes
- Legend or direct labels
- Source / notes
Chart sequence
Each supporting narrative. Not «here are 20 charts».
Data tables
When numbers matter more than trend.
Audience
Executive
- Short (1-2 pages)
- Visual-heavy
- No technical details
- Clear recommendations
Product team
- Details
- Specific metrics
- Technical methodology ok
- Implementation suggestions
Engineering
- Technical depth
- SQL / code
- Data pipeline concerns
Cross-functional
Balance — common language.
Workflow
1. Before writing
Structure outline. Know conclusion.
Writer clarity ↔ reader clarity.
2. Draft
Get it down. Don't polish.
3. Review
Sleep on it. Read next day.
4. Edit
Tighten. Cut unnecessary.
5. Peer review
Colleague reads. «Does this make sense?».
6. Stakeholder
Share. Iterate based on questions.
Common mistakes
Numbers without context
«CR 5%». So what? Compare к target, period, industry.
No recommendation
«Data shows X». Что делать?
Too long
30-page отчёт → ignored.
Aim 3-5 pages для most analyses. Plus appendix.
Burying insight
Insight на page 10 → nobody reaches.
Lead with conclusion.
Defensive
«This might not be statistically significant but...».
Be confident. Or don't include.
Weak titles
«Analysis results» → no one reads. «Mobile drives 60% of churn — fix iOS onboarding» → читают.
Editing checklist
- Executive summary stands alone?
- Each chart supports narrative?
- Numbers have context?
- Recommendations specific?
- Limitations mentioned?
- Grammar / typos?
- Links / references?
Delivery
Format
- Google Docs collaborative
- PDF fixed / shareable
- Notion для internal wiki
- Slides для executive present
Length
Depends на audience. Exec: 1-2 pages. Detailed: 5-10 pages. Rarely more.
Follow-up
After share:
- Schedule discussion
- Collect feedback
- Iterate
Templates
Analysis report
As above.
Weekly / monthly metrics
Summary + dashboard link + callouts.
Post-mortem
Event → impact → RCA → actions.
Experiment results
Setup → hypotheses → results → decision.
Research proposal
Questions → methodology → timeline → deliverables.
Writing samples
Good
«iOS users driving 60% of churn. Recent app update broke Apple Pay на 15% of devices. Rollback recommended. Estimated impact: +3% conversion ($120k annual revenue)».
Bad
«Our analysis examined various factors влияющие на customer retention across multiple cohorts and segments. Several interesting patterns emerged, particularly around mobile users in certain geographic regions».
First: specific, actionable, quantified. Second: vague, filler, no commitment.
Version control
For important reports:
- Version numbers
- Track changes
- Reviewers
- Final vs draft
Data referenced
Links
Query / notebook linked. Future reference.
Raw data
Access for verification. Don't paste raw tables — link.
Replicability
Someone could reproduce? Key quality indicator.
Storage
Searchable
Notion / Confluence wiki. Tagged.
Accessible
Relevant team members access.
Indexed
Title, summary searchable.
На собесе
«Write analysis report?»
Walk through structure. Show thinking process.
Feedback to improve
Stakeholder questions
Track what asks. Next time — include in report proactively.
Success metric
Did decision happen? Based on your report?
Impactful reports — valuable. Non-read reports — signal.
Связанные темы
FAQ
Templates существуют?
Yes. Internal team может иметь. Или Notion templates.
Как improve writing?
Read great writing (FiveThirtyEight, Stratechery). Practice. Get feedback.
One format все situations?
No. Adapt context.
Тренируйте — откройте тренажёр с 1500+ вопросами для собесов.