What It Is
The Four Types of Research is a classification framework from Erika Hall's Just Enough Research that categorizes all research activities by their purpose. Instead of starting with a method (survey, interview, A/B test), you start by identifying what type of question you need to answer, then select the appropriate methods. The four types are Generative, Descriptive, Evaluative, and Causal.
When to Use It
- At the start of any research initiative, to ensure you are using the right approach for your question.
- When a stakeholder says "let's do a usability test" and you suspect the real question is more fundamental.
- When planning a research roadmap and you need to sequence different types of research across a product lifecycle.
- When a team is unsure whether they need research at all — this framework clarifies what they would learn.
How It Works
1. Generative Research (Exploratory)
Question: "What exists? What are the opportunities? What problems do people have?"
Purpose: Discover new insights, uncover unknown needs, explore a problem space before defining solutions.
Methods: Open-ended interviews, contextual inquiry, field observation, diary studies, ethnographic research.
When: Early in a project, before you have a defined solution. Also useful when entering a new market or domain.
2. Descriptive Research (Documenting)
Question: "What is happening? How do people currently behave? What does the landscape look like?"
Purpose: Document and characterize existing behavior, attitudes, or market conditions. Build a factual picture of the current state.
Methods: Surveys, analytics review, competitive analysis, content audits, log analysis, personas (data-driven).
When: After generative research has identified areas of interest, and you need to quantify or map them. Also useful for establishing baselines.
3. Evaluative Research (Testing)
Question: "Does this solution work? Can people use it? Does it meet their needs?"
Purpose: Assess a specific design, prototype, or product against criteria. This is where usability testing lives.
Methods: Usability testing, heuristic evaluation, A/B testing (for design variants), concept testing, accessibility audits.
When: After you have a solution (or prototype) to test. This is the most commonly used type and the one teams over-index on.
4. Causal Research (Measuring)
Question: "Did this change cause that outcome? What is the impact?"
Purpose: Establish cause-and-effect relationships between variables. Determine whether a specific intervention produced a measurable result.
Methods: Controlled experiments (A/B tests with statistical rigor), multivariate tests, longitudinal studies, pre/post analysis.
When: After launch, when you need to measure impact. Requires sufficient sample sizes and controlled conditions.
Key Principles
- Type before method. Always ask "what type of question am I answering?" before asking "what method should I use?" The method follows from the type, not the other way around.
- Most teams skip generative research. They jump from an idea to evaluative testing, never validating whether the problem is real or significant. This is the single biggest research mistake in product teams.
- The types build on each other. Generative informs what to describe, descriptive informs what to evaluate, evaluative informs what to measure causally. Skipping a stage means the next stage's inputs are weaker.
- No type is "better." Each answers a different question. An A/B test cannot tell you what to build, and an interview cannot tell you which variant converts better.
- Match rigor to stakes. Not every question needs a controlled experiment. Use the lightest-weight approach that gives you sufficient confidence for the decision at hand.
Common Mistakes
- Using evaluative research for generative questions. Running a usability test to "understand user needs" conflates two types. A usability test tells you whether a specific design works, not what problems people have.
- Treating all surveys as descriptive. A survey with leading questions about your product is evaluative (or just bad research). Surveys are only descriptive when they document behavior and attitudes without presupposing a solution.
- Skipping straight to causal research. Running an A/B test before doing generative and evaluative work means you are measuring the impact of something you never validated was the right thing to build.
Source
Erika Hall, Just Enough Research (2nd edition, 2019), Chapter 2. The four types are introduced early as the organizing framework for the rest of the book, which then covers methods appropriate to each type.
