New product launches, market-entry decisions, existing-product rebrands — major business decisions are typically preceded by market research.
But "what kind of research, at what scale?" is rarely obvious. This article covers the foundational distinction between quantitative and qualitative research and when to use each.
Quant vs. qual
Quantitative research
Numbers, ratios, statistics. "How many people, what, how much."
- Methods: surveys, web behavior log analysis, transaction data analysis
- Sample size: hundreds to tens of thousands
- Analysis: averages, distributions, statistical tests
- Strength: statistically defensible, year-over-year comparable
- Weakness: shallow on "why"
Qualitative research
Words, behaviors, emotions. "Why this thinking, why this behavior."
- Methods: interviews, focus groups, ethnography, behavioral observation
- Sample size: a few to a few dozen
- Analysis: narrative analysis, coding
- Strength: deep understanding, hypothesis discovery
- Weakness: statistical representativeness is limited
Which first?
The conventional order is qualitative → quantitative:
- Qualitative for hypothesis discovery and depth
- Quantitative for validation and market sizing
Reverse order ("see the big quantitative pattern, then deep-dive for meaning") is also valid (quant → qual). Pick based on purpose.
Quantitative research design
Sample size benchmarks
For "speaking statistically about the overall market," required sample size depends on population and acceptable margin:
| Population | ±5% margin | ±3% margin |
|---|---|---|
| 1,000 | 278 | 516 |
| 10,000 | 370 | 964 |
| 100,000 | 383 | 1,056 |
| Infinite | 384 | 1,067 |
The practical rule of thumb: ~400 is the floor for "statistically defensible".
Sampling design
Random sampling is the ideal; quota sampling (mirroring population proportions by age, gender, region) is the practical norm.
Example (US adult population proxy):
- Gender 49:51
- Age 25% / 25% / 25% / 25% across 25–34 / 35–44 / 45–54 / 55–64
- Region weighted to census distribution
Question design
1. Lock the hypothesis
State "what are we trying to confirm" up front. Without it, even lots of questions produce no conclusion.
2. Include benchmark questions
At least one industry-standard question (e.g., NPS). Enables cross-comparison with other companies / studies.
3. Screening up front
Limit the survey to qualified respondents:
Q1. Have you used [product type] in the past 6 months?
○ Yes → continue
○ No → end
4. Competitive comparison
Q. Please compare our product to competitors.
Rate satisfaction with each (5-point):
- Our product A: ___
- Competitor B: ___
- Competitor C: ___
Reveals positioning in one question.
5. "Should not" — what they don't want
Asking what they don't need is also useful. Surfaces features to cut:
Q. Which of these features feel unnecessary? (Multiple)
□ Feature A
□ Feature B
□ Feature C
□ All are needed
Qualitative research design
Sample size
5–15 interviews typically reach saturation (where new findings stop emerging and the same patterns repeat).
Question design
Open-ended
- How do you generally think about X?
- Why did you choose △△ in the past?
- If ×× were the case, what would you do?
Questions that can't be answered "yes/no".
Behavioral observation
"How do you think about it" is less reliable than "actually do it." Show a hypothesis-validation prototype and observe reaction.
Interviewer notes
- Don't lead: "X is the case, right?" is off
- Don't fear silence: deep answers often come 5 seconds after the pause
- 5 whys: drill from surface answer to root motivation
Hybrid patterns
Common combinations in practice:
Pattern 1: Screen-and-dive
Quantitative survey identifies high / low responders → interview them.
Pattern 2: Hypothesis-then-validate
Small-n interviews generate hypotheses → quantitative validates.
Pattern 3: Parallel
Quant and qual concurrent, cross-referenced during analysis.
What not to do
❌ "Voice of the market" from n ≤ 30
Statistically meaningless.
❌ 100 responses of open text only
Analysis cost too high; ends up unused.
❌ Survey only your network
Snowball sampling is convenient but heavily skewed in attributes.
❌ Speak about older audiences via web-only
Skewed toward digitally-accessible segments.
Summary
Market research fundamentals:
- Lock purpose and hypothesis first
- Qualitative for hypothesis discovery and depth
- Quantitative for validation and sizing
- Multiple channels to correct sampling bias
- Frame with decision-makers before reporting
Repoan's market-research template for new products mirrors this article's screening + competitive structure.
Past 400 responses, manual open-text reading breaks down. AI reports (see AI-driven response analysis) auto-classify open text by industry and use case — surfacing unexpected use cases in minutes. A/B testing (see survey A/B testing) lets you try multiple question variants, continuously improving the research design itself.