Как делать research в продукте
Карьерник — квиз-тренажёр в Telegram с 1500+ вопросами для собесов аналитика. SQL, Python, A/B, метрики. Бесплатно.
Зачем это знать
Analytics only — incomplete. «Users doing X» not «why doing X». Research fills gap.
Senior analyst incorporates qualitative insights. Middle often focused only numbers.
Types
1. User interviews
1-1 conversation. Open-ended questions.
Best для: understanding motivations, pain points.
Cost: high (time-intensive).
2. Surveys
Questions multiple users.
Best для: quantitative validation, widespread feedback.
Cost: medium.
3. Usability testing
User performs tasks с product. Observe.
Best для: UX issues, friction points.
4. Analytics
Behavioral data.
Best для: what users do, scale.
5. Session recordings
Watch real sessions.
Best для: qualitative + behavioral.
6. Beta / dogfooding
Early access. Feedback.
Best для: pre-launch validation.
7. Competitive research
Analyze competitors.
Когда что
Exploratory
«New product — what to build?»
Interviews + user observation.
Confirmatory
«Hypothesis X correct?»
Analytics + A/B test.
Usability
«Does new design work?»
Usability testing.
Scale validation
«Everyone like this?»
Survey.
Mix methods. Triangulate.
User interviews
Preparation
- Target users recruit
- Schedule 30-60 min
- Prepare discussion guide (not script)
- Record (с consent)
Discussion guide structure
- Warmup (background)
- Current behavior
- Problems / pain
- Hypothetical scenarios
- Wrap-up
Open questions
- «Tell me about...»
- «Walk me through...»
- «Why is that important to you?»
- «What did you try before?»
NOT leading:
- «Don't you think X is great?» — bad
- «What do you think of X?» — better
After
- Notes / transcription
- Patterns across interviews
- Key quotes
Recruitment
- Внутри existing users
- Recruiting services (UserTesting)
- Social media (for specific segments)
Compensation — gift card, credit.
Surveys
Design
- Clear questions
- Short (< 10 min)
- Mix quantitative (Likert) + qualitative (open)
- No leading
Example Likert
«How satisfied с app? 1 (very unsatisfied) - 5 (very satisfied)»
Sample size
- Confidence: 95%, margin 5% → ~400 responses
- Smaller — directional только
Tools
- Google Forms (free)
- Typeform (nicer)
- SurveyMonkey (legacy but works)
Distribution
- In-app popup
- User panels
Response rates
~5-20% typical. Boost:
- Shorter
- Incentive
- Clear purpose
Usability testing
Moderated
Observer + user. Ask think-aloud.
Unmoderated
User performs tasks. Records self.
Tools
- UserTesting
- Maze
- Lookback
Metrics
- Task success rate
- Time on task
- Error rate
- Subjective satisfaction
Analytics integration
Use analytics first to:
- Form hypotheses
- Identify problems
- Segment users для interviews
Then research to:
- Explain behavior
- Inform решения
- Validate проблемы
Both inform each other.
Synthesize findings
Affinity mapping
Group notes by theme. Visual.
Personas (careful)
Synthesized user types. Useful, но don't overly simplify.
Customer journey map
Visual journey с stages, feelings, pain points.
Insight statements
«Users want X because Y, but current experience fails at Z».
Actionable.
Communicate
Research report
- Executive summary
- Methodology
- Key findings
- Quotes
- Recommendations
Share widely
Team-wide presentation. Video snippets impactful.
Action items
Connect insights к product changes.
Common mistakes
Talking too much
User должен speak 80%. You 20%.
Leading questions
«Don't you think...» — biased.
Asking for features
«What feature should we build?»
Better: «What problem trying solve?». They're poor designers.
Small sample decisions
«1 user mentioned — we must fix». Prioritize by frequency.
Survey + interview conflict
Data differs? Don't dismiss. Investigate.
Для analyst
Analyst owns research?
Sometimes. Often collaborates с Product Researcher (PR).
When no dedicated PR
Analyst может conduct basic research.
Invest в basics.
Training
- «Interviewing Users» — Steve Portigal
- «Talking to Humans» — Giff Constable
- NN/g courses
Ethics
Consent
Explicit. Recording — ok?
Incentives
Reasonable. Not coercive.
PII
Handle carefully. Anonymize in reports.
Harm
Minimize — questions not distressing.
В Russia
Culturally:
Interviews
Russian users often blunt. Less polite filler.
Can be advantage — direct feedback.
Surveys
Response rates similar West.
Paid compensation
Common. Gift cards, wallet credit.
Mistakes analytics-only
«We know why»
Analytics shows behavior. Motivation — elsewhere.
Ignore user voice
Users do strange things. Sometimes data wrong, user correct.
One-way
Analytics tell you. Research — you tell them hypotheses, validate.
На собесе
«Research в role?»
Specific:
- Type (interviews, surveys)
- Recruited users
- Sample size
- Findings
- Impact
Even informal research counts.
«When prefer research vs analytics?»
Qualitative — motivations. Quantitative — behaviors.
Both.
Связанные темы
FAQ
User interview — сколько?
5-8 reveal patterns. 20+ confidence.
Surveys online?
Common. Link email, in-app.
Research bought tools?
UserTesting, Maze paid. Google Forms / personal network free.
Тренируйте — откройте тренажёр с 1500+ вопросами для собесов.