Buyer research
Customer interviews or surveys: which will answer your question?
Choose interviews, surveys or a sequence of both using a practical decision table, a worked B2B example and lessons from Carbon’s buyer research.
The short answer
Use interviews to explore how a decision happens and uncover language or issues you do not yet understand. Use a well-designed survey to measure defined questions across a specified group. Often the useful sequence is interviews, questionnaire testing, then a survey—but the business decision should determine the method.
Start with what you need to decide.
“We should do a survey” often means “we need a clearer answer.” Those are not quite the same brief. The right method depends on what is unclear and how the answer will be used.
If you do not understand how customers choose a partner, a fixed list of answer options may simply measure your assumptions. If you already understand the possible issues and need to estimate how common they are within a defined audience, a handful of interviews will not give you that estimate.
Write the decision first: change the onboarding process, choose a positioning direction, improve the website or investigate why customers reconsider a partner. Then write the information that would change that decision.
Match the method to the uncertainty.
| Question you need to answer | Useful starting point | What to watch |
|---|---|---|
| What happened during the last supplier decision? | Interviews with people involved in recent decisions | Recall can be incomplete; ask about specific events and sequence. |
| Which concerns are we failing to understand? | Exploratory interviews across relevant customer situations | Do not recruit only your most enthusiastic customers. |
| How common is a defined concern in this audience? | A survey with clear measures and an appropriate recruitment design | A percentage describes the respondents and design; it is not automatically representative of a whole market. |
| Do people understand the proposed survey question? | Questionnaire pretesting or cognitive interviews | Ask how they interpreted the wording and chose an answer. |
| Which of several developed messages performs better on a defined measure? | A designed message test, often informed by earlier interviews | Control the comparison and define the measure; preference alone does not prove sales impact. |
| Why did this pattern appear in the survey? | Follow-up interviews with relevant respondents | Explore explanations without treating them as proven causal effects. |
A worked example: customers are satisfied, but renewals feel uncertain.
Imagine a B2B service business hearing positive feedback while more accounts ask for competitive proposals. The team suspects price, but it does not yet know what “better value” means to different customers. This is an illustrative research brief, not a client result.
Begin with interviews that cover different experiences: renewed accounts, recently lost accounts where access is possible, customers who reviewed alternatives and people involved in choosing a new provider. Ask about the most recent decision rather than asking people to describe an ideal partner in the abstract.
Useful prompts include: “What first prompted the review?”, “Who else became involved?”, “What did you need to explain internally?” and “What changed between the first conversation and the decision?” Follow the answer before moving to the next question.
The interviews may reveal distinctions worth measuring—for example, a price concern, uncertainty about results or difficulty explaining a provider’s contribution. Use those findings to develop neutral survey questions. Pretest the wording with suitable people before fielding the full questionnaire.
How we frame the research decision.
The practical question is: do we need to discover the possible answers, measure known answers, or understand a pattern we have already measured? That keeps the method tied to the business need.
For exploration, leave room for an answer you did not anticipate. “What made the change worth considering?” is more open than “Was the price too high?” For measurement, make sure the response options capture distinct ideas and that the question asks one thing at a time.
Pew Research Center’s questionnaire guidance explains why wording, answer options and order matter, and describes pretesting as part of survey development. Treat the first questionnaire as something to test, not a finished instrument.
The strongest design also accounts for access, budget, timing and the limits of the sample you can realistically recruit.
What Carbon’s buyer study helps clarify.
Carbon’s 2026 B2B Buyer’s Report findings show why separate questions matter. In the 189-response study, 177 of 188 respondents to the satisfaction question were somewhat or very satisfied. Separately, 110 of 186 were somewhat or extremely likely to reconsider at least one partner in the next 12 months.
Those findings can coexist. They do not tell us that 59% of satisfied customers will leave, because that combines different questions and implies an outcome the study did not establish. The public findings page includes the question wording, bases and methodology.
A survey can reveal a pattern that deserves investigation. Interviews can help explore how people describe the situations behind it. Neither method earns trust by claiming more than the evidence supports.
Ask for an answer you can act on.
A useful research readout should connect the question, method, audience, evidence, limitations and business implications. Keep observations separate from interpretation. Show where different groups disagree rather than smoothing every answer into one average.
Finish with a decision and the uncertainty that remains. You might be ready to change a message, test a customer conversation or investigate one issue further. The deliverable is not simply a large deck; it is a clearer basis for the next move.
Put it to work
Customer research method decision brief
Download the editable text file and use it with your team. No form needed.
Download the worksheetSources and further reading
Sources reviewed September 16, 2026.