Start with the decision, not the method
A useful research plan begins with a decision: what might you change once you have an answer? “Understand our customers” is too broad to guide a study. “Decide which part of checkout needs improvement” points toward a specific question, such as where people hesitate and what they think is happening at that point. The clearer the decision, the easier it is to choose a method that can produce relevant evidence.
Surveys are well suited to collecting comparable answers from a defined group. Interviews help you explore how people explain their needs, choices, and experiences. Observation shows what happens during an activity, including steps people may forget to mention. None is inherently more rigorous than the others. Each becomes useful when its strengths match the question and its limitations are accounted for.
Choose a survey when comparable answers matter
A survey is a good starting point when you know what you want to ask and need to compare responses across people. You might ask recent buyers which delivery concern mattered most, or compare how different customer groups rate the clarity of a product description. Consistent wording and response options make patterns easier to inspect, provided the people who respond reasonably represent the group you want to understand.
Surveys are less helpful when you do not yet know the right answer options. A question asking why customers abandoned a purchase may miss a reason you never thought to include. Open-text questions can help, but brief written answers rarely provide the same opportunity to ask follow-up questions as a conversation does. Before distributing a survey, test its wording with a few people to check whether they interpret questions as intended. Keep the results tied to who was invited, who responded, and what the questions actually asked; a precise-looking percentage does not fix a poorly framed question.
Choose interviews to explore meaning and motivation
Interviews are useful when a decision depends on how people understand a problem. For example, if customers describe a service as “complicated,” a conversation can reveal whether they mean its price, the steps required to start, or the language used to explain it. Ask for specific examples: “Tell me about the last time you tried to do this” often produces more useful detail than “Would you use this feature?” Follow-up questions can uncover expectations and trade-offs that were not obvious when the study began.
The flexibility of an interview is also a source of risk. An enthusiastic participant may not reflect the wider audience, and a leading question can steer an answer. Use a short guide covering the same core topics with each participant, while leaving room to probe unexpected points. Treat repeated themes as ideas to investigate, not as a reliable count of how common a view is. Interviews can explain possible reasons behind a pattern; they generally cannot establish its prevalence on their own.
Choose observation to understand what people do
Observation is especially useful when the activity itself matters. Watching someone compare products, navigate an unfamiliar form, or use a tool in their normal workflow may reveal pauses, workarounds, and environmental constraints that do not come up in a survey or interview. Decide in advance what actions to note, and separate a description of what happened from your interpretation of why it happened. A pause before clicking is visible; confusion is a possible explanation, not an observed fact.
Context matters. People may behave differently when they know they are being watched, and a short session captures only a slice of their experience. When appropriate, ask afterward what they were trying to accomplish at key moments rather than guessing. Explain what you will observe, obtain appropriate permission, and avoid recording personal information you do not need. If the normal setting is not accessible, a task-based session can still expose friction, but describe it as a test situation rather than everyday behavior.
Combine methods when one answer is not enough
Many research questions have more than one part. Suppose a team wants to improve a sign-up process. Observation could show where people stop or backtrack. Interviews could explore what they expected to happen there. A survey could then ask a broader group about a clearly defined concern that emerged. The order can change: survey responses might reveal a pattern that calls for interviews, or interviews might help write better survey questions.
You do not need to use every method. Choose the smallest sensible combination that addresses the uncertainty behind your decision. Before collecting data, write down what each method should reveal and what it cannot tell you. Afterward, look for both agreement and disagreement between findings. If people say a step is easy but struggle to complete it, that gap is worth examining. A useful conclusion states what you learned, where the evidence is limited, and what action or further test follows.
