"Should we run user interviews or surveys?" is one of the most common questions in early-stage customer research programs.
The framing is wrong. They're not in competition — they're a depth vs. scale division of labor, and they're meant to be combined. This article lays out the real differences, when each one is the right tool, and the patterns for running them together.
The core difference — depth vs. scale
| Dimension | User interviews | Surveys |
|---|---|---|
| Sample size | 5–20 | 100–10,000 |
| Time per respondent | 60–120 min | 3–10 min |
| Data captured | Conversation with context | Structured data |
| Question answered | "Why? How?" | "What? How much?" |
| Strength | Depth, discovery | Scale, repeatability |
| Weakness | Doesn't scale | Thin on context |
| Output style | Stories, hypotheses | Statistics, trends |
They capture different kinds of truth.
Strengths and weaknesses
Interview strengths
1. Context and emotion
When someone says "it's hard to use," you also get when, in what situation, and what they were expecting. With surveys you just get the three words.
2. Unexpected findings
Because you haven't locked the question set, topics you didn't expect come up. That's where insight lives.
3. Reading non-verbals
Facial expressions, tone, hesitations — non-verbal signal you'd never capture in a survey.
4. Strong for hypothesis generation
Generating brand-new hypotheses and new angles: interviews win, hands down.
Interview weaknesses
1. No scale
n=5–20 carries no statistical weight. You can't speak for "all customers."
2. Effort
60–120 min per session + prep + analysis = 2–4 hours per respondent. 20 respondents = 80 hours.
3. Interviewer skill dependency
Output quality varies wildly with the interviewer's skill. Good interviews take craft.
4. Hard to circulate internally
You can't drop "8 conversations" into a slide deck the way you drop a chart. Findings move slowly through the org.
Survey strengths
1. Scale
n=hundreds to tens of thousands — statistically meaningful patterns.
2. Repeatability
Same questions next quarter = trend over time. The foundation of continuous research.
3. Efficiency
3–10 min per respondent. Fast, high-volume capture.
4. Easy to socialize
Numbers and charts slot into board decks and internal reports.
Survey weaknesses
1. You only get answers to your questions
Anything outside the designer's framing doesn't come back. Low discovery value.
2. Thin context
"Dissatisfied" tells you nothing about why or in what situation. Lots of interpretation room.
3. Open text gets neglected
Past ~100 responses, eyeballing breaks down and open text quietly becomes dead weight in the pipeline.
4. Designer bias
Without discipline, the designer subtly steers questions toward the conclusion they want.
When each is the right tool
Use interviews when
- You're trying to form new hypotheses
- You want to understand why
- You're looking for the unexpected
- You want deep understanding of a few customers
- The goal is insight discovery
Use surveys when
- You're validating a hypothesis at scale
- You're monitoring a metric (NPS, CSAT)
- You're tracking segment-level patterns
- You're looking at change over time
- You need statistically meaningful sample sizes
Either works
- Simple satisfaction check → survey is fine
- Customer base overview → survey is more efficient
Neither alone is enough
- Deep churn analysis → survey ("what's increased?") + interview ("why?")
- New feature concept → interview (hypothesis) + survey (scale check)
- Continuous customer understanding → ongoing surveys + quarterly interviews
The hard part — kill the "which is better" debate
The "interviews vs. surveys" framing is a category error. They serve different purposes and capture different truths.
To people who say "interviews are enough"
- n=5–20 can't legitimately describe "all customers"
- Hypotheses need scale validation
- "We talked to 5 people and..." doesn't carry much weight in an exec staff room
To people who say "surveys are enough"
- Surveys only return what you asked
- You're making decisions without understanding "why" or "how"
- Insight is hard to extract from numbers alone
The answer isn't either/or — it's "do both." Having both capabilities in the org is essentially the baseline for serious customer understanding in the AI era.
Four working patterns
Pattern 1: Quant-first (the default)
1. Survey to read overall trends
2. Find an anomaly or curious area
3. Interview the relevant segment (n=5–10)
4. Form a hypothesis
5. Survey to validate at scale
6. Ship a change
7. Survey to measure the effect
Pattern 2: Qual-first
1. Interview to generate hypotheses (n=5–10)
2. Survey for scale validation
3. Ship a change
4. Survey to measure
5. Follow up with more interviews if needed
The default for new business / new feature work.
Pattern 3: Parallel
1. Run survey and interviews simultaneously
2. Reconcile findings against each other
3. Look at the disagreements for new insight
4. Integrate into the decision
Time-constrained efficiency pattern.
Pattern 4: Continuous observation
1. Monthly survey running on autopilot
2. One qualitative interview per quarter
3. Watch numerical drift alongside voice change
Best for mature products in steady improvement mode.
Effort sizing
Rough sizing for combined operations:
| Tier | Effort per cycle |
|---|---|
| Light (n=300 survey + n=3 interview) | 20–30 hours |
| Standard (n=500 survey + n=5 interview) | 40–60 hours |
| Heavy (n=1,000 survey + n=10 interview) | 80–120 hours |
On a quarterly cycle, that's roughly 10–40 hours per month — one dedicated owner, or split across multiple people.
Org structure
Combining both modes has org implications:
Setup 1: One person does both
The lean version. A marketer / PMM / UX researcher runs both.
Setup 2: Split quant and qual
Data analyst (quant) + UX researcher (qual). Requires a formal venue for the two to compare notes.
Setup 3: Split + integration role
Data team + research team + PMM/PdM as integrator. Three layers, common at mid-size and up.
Across all three, what matters most is a recurring meeting where both sides reconcile findings into decisions.
Where Repoan fits
Repoan is survey infrastructure, but it's designed with the interview interface in mind:
- Pre-interview screening surveys — pick the right n=5
- Post-interview scale validation — convert hypotheses into surveys
- Continuous-survey dashboards — track change over time
- AI open-text analysis — structure interview-style free-form data
- AI-generated questions — design the next survey based on interview findings
Summary
Interviews vs. surveys is a depth vs. scale split:
- Interviews: 5–20 people, context, emotion, the unexpected, hypothesis generation
- Surveys: 100–10,000, scale, repeatability, pattern detection, hypothesis validation
- Which is "better" is the wrong question; how to combine them is the right one
- Four patterns: quant-first, qual-first, parallel, continuous observation
"Do both" is the modern default for customer research. Drop the versus framing and run them as complements — that's the fastest path to deep customer understanding.