No single method owns the truth. Social listening captures raw language and public intent. Surveys summarize preferences at scale. Interviews reveal why behind the numbers. Strong teams stack them deliberately - not randomly.
Complements customer research methodology and PMF interviews vs community evidence.
Decision matrix
| Method | Best for | Weak for | Typical n |
|---|---|---|---|
| Social listening | Phrases, objections, alternatives | Private workflows, exact budgets | Hundreds of posts scanned |
| Surveys | Priority ranks, segment splits | Unknown unknowns | 50–500 responses |
| Interviews | Why, procurement, politics | Speed, representativeness | 8–20 deep calls |
Social listening (public conversations)
Sources: Reddit, HN, Stack Overflow, GitHub, YouTube comments, etc.
Strengths:
- Unfiltered language buyers use when vendors are not in the room
- Recurring pain across strangers (frequency signal)
- Competitor and "alternatives to X" context
Blind spots:
- Loud minorities; astroturfing in some categories
- No budget authority; missing silent majority
- Enterprise buyers often discuss in private channels
Tools: Manual search, Trending Problems, Search for ranked multi-platform results.
Surveys
Strengths: Count how many prefer A vs B once you know the questions.
Blind spots: You only measure what you thought to ask; low response rates skew results.
Use when: You have 10+ interview themes to quantify, or need price sensitivity bands before packaging change.
User interviews
Strengths: Follow-up questions, procurement steps, politically sensitive truth.
Blind spots: Courtesy bias; small-n; sponsor effects in enterprise.
Use when: Deal size justifies hours, or threads and surveys disagree.
Practical sequence (early B2B)
Week 1–2: Listen → 30–50 phrases from public threads
Week 3: Survey → prioritize top 5 themes (optional)
Week 4+: Interview → 8–12 calls on highest-intent segment
Ongoing: Listen → weekly phrase refresh
Pre-PMF teams often stop after listen + 5 interviews - enough to rewrite homepage, not enough to pretend you have statistically valid WTP.
When methods disagree
| Situation | Likely cause | Next step |
|---|---|---|
| Threads loud, interviews quiet | Wrong segment in calls | Refine ICP filter |
| Interviews loud, threads quiet | Category not discussed publicly | Sales-led motion |
| Survey vs threads conflict | Survey audience ≠ thread audience | Segment survey |