Research methods
Customer research methods: pick the one that answers your question
The main types of customer research, what each can and cannot tell you, and a simple sequence for a small team deciding what to build.
Customer research methods fall into five groups: talking to people (interviews), watching them work (field visits and usability sessions), asking many people the same questions (surveys), measuring what they do (product analytics and experiments), and reading records you already have (support tickets, sales notes, reviews). Choose by the question you need answered. If you need to know why something happens or how people do a job, use a qualitative method such as interviews or observation. If you need to know how many people do something or how much, use a quantitative method such as a survey or analytics.
Here is a hypothetical example. You are exploring software that helps dental practices fill last-minute cancellations. You have four open questions:
- How does the front desk handle a cancellation today?
- How often do cancellations leave a chair empty?
- Would the office manager, the dentist or a group owner decide to pay for a fix?
- Can a receptionist use your prototype during a busy morning?
Each question needs a different method. The first and third call for interviews. The second is a counting question, and the practice's scheduling records will answer it better than anyone's memory. The fourth needs you to watch someone use the prototype.
The main types of customer research
| Method | Good for | Weak at |
|---|---|---|
| Interviews | How a job gets done, who decides, what people tried before | How common something is |
| Field visits and contextual inquiry | Seeing the real workflow, workarounds and interruptions | Scale; they take time to arrange |
| Usability sessions | Whether people can complete a task with your product or prototype | Whether they want the product at all |
| Surveys | How many people report an attitude or experience, once you know what to ask | Explaining why; follow-up questions |
| Analytics and A/B tests | What people actually do in a live product, at volume | Reasons behind the behavior; they need a product and traffic |
| Existing records (support tickets, sales and churn notes, public reviews) | Recurring complaints and the words customers use | Needs of people who never contacted you |
Nielsen Norman Group's guide to research methods maps 20 methods on the same lines: qualitative methods suit questions about why something happens or how to fix it, and quantitative methods suit questions about how many and how much.
Separate what people say from what they do
The same guide separates attitudinal research (what people say) from behavioral research (what people do) and notes that the two often differ. A receptionist may tell you she always calls the waitlist when someone cancels. The call log might show that she only does it on quiet days.
This matters when you choose a method. Interviews and surveys collect accounts. Observation, analytics and records show behavior. When a decision rests on behavior, such as whether people will change a habit or pay, look for evidence of what they have actually done: a recent example in an interview, a record, or a real purchase.
Match the method to your stage
Early on, you are deciding what to build and for whom. NN/g groups interviews, field studies, diary studies, surveys and concept testing as generative methods for this stage. Many early teams start with interviews because they can be arranged quickly and let you follow an answer you did not expect.
Once you have a design, usability testing shows where people get stuck. After launch, analytics, A/B tests and surveys measure how the product performs, and interviews explain the numbers.
Surveys are easy to send and hard to do well. NN/g's article on when to run a survey argues they fit questions that are both quantitative and about attitudes, such as what share of users feel confident after a task. A question that asks why belongs in a qualitative method. Our interviews versus surveys guide goes through that choice in more detail.
Start with what you already have
Before recruiting anyone, check what your company already holds: support tickets, sales call notes, reasons for cancelling, and public reviews of competing products. These cost nothing to read and often show the customers' own words for the problem. They are skewed toward people who bothered to write, so treat them as leads for your interviews and not as a measure of how common a problem is.
A small-team sequence that works for most early questions
- Write the decision you face and the one or two claims you are unsure about.
- Read the records you already have and list what they cannot tell you.
- Interview people who did the relevant job recently. For the dental example, that means receptionists, office managers and someone who owns several practices. The finding customers guide covers where to find them, and the interview count guide helps you decide how many to plan.
- If a question about frequency remains, count it with records or a survey written in the vocabulary you heard in interviews.
- When you have something to show, watch people use it.
You may not need every step. A small decision can rest on a handful of conversations and the records you already hold.
When you need people outside your network
Interviews are the method most teams can start this week, but only if you can reach people who do the work. For the dental example, you could start from a directory page such as operations professionals in dental services or describe the role in a people search. Instant Expert finds people whose work matches your question. You review them, it sends the invitations, and you pay for each call that gets booked. Before you invite anyone, write down the one question each conversation needs to answer.