Your own idea can sound clear in your head while the people you hope to serve describe the problem in completely different words. Customer research helps you notice that difference before you spend weeks on a message, page or product. You do not need to become a professional researcher to begin. You do need a fair question, people who have relevant experience, careful notes and a decision you are willing to reconsider.
By the end of this article, you will be able to plan a small customer research session, ask questions that do not push for praise, separate what someone said from your interpretation, and choose one next test. The result is a one-page research note, not a claim that you have discovered what an entire market thinks.
Start with one decision you need to make
“Learn about customers” is too broad to guide a useful conversation. Which choice are you trying to make? Perhaps you need to know whether people understand an appointment form, how they currently choose a provider, or what information they require before booking. A research question is a question whose answer would help that decision. For example: “What information do people look for before requesting an appointment online?”
This is different from an interview question, which is the wording you say to a participant: “Think about the last time you looked for an appointment. What information did you check before contacting anyone?” The research question keeps you focused; the interview question makes it easy for them to describe an experience. Nielsen Norman Group recommends writing the research question first and connecting interview prompts to it (Rosala, 2021).
If you have already completed chapter 4, you may have a measurable goal. Research can help you find out what blocks the customer action behind that goal. If the goal concerns completed online requests, there is little value in asking only whether people “like” the website. Ask how they decided, what they expected to find and what happened when they tried. If you have not read chapter 4, this paragraph gives you enough context to continue: a goal is the result you hope to achieve; this chapter investigates the customer experience that could affect it.
Where customer research fits in the larger plan
PR Smith’s SOSTAC® framework begins with Situation: where are we now? It then moves through Objectives, Strategy, Tactics and Action to Control, which checks results and informs the next Situation. Customer research is one way to improve the Situation picture. It also helps you revisit an Objective if you find that it measures the wrong customer action. The diagram keeps the same six-stage layout as chapter 4, with Situation highlighted this time (SOSTAC, no date).
Figure 1 in words
Figure 1 in words: Situation gathers what is known about customers and the current experience. Objectives state the desired result. Later actions are measured in Control, and the learning returns to Situation. You do not need to fill a large formal plan before talking to people.
What counts as customer research?
Customer research is a deliberate attempt to learn from people and evidence relevant to a decision. It includes listening to someone describe a recent choice, observing how they use a page, reading records of repeated questions, or using a short survey when you already know what you need to count. A conversation becomes more useful when you know whom you asked, why they were relevant, what you asked and what you actually learned.
Primary information is collected for your current question, such as an interview or a test of your own booking page. Secondary information already exists for another purpose, such as a public report or your own historical service records. Both can help. Existing information may be quick and inexpensive but not answer your exact question; a direct conversation can explore it in detail but takes time and does not represent everyone. These are standard distinctions in marketing research (Gomez Albrecht, Green and Hoffman, 2023).
Figure 2
Figure 2 in words: Name the decision. State what you need to learn. Choose a method suited to that question. Record the evidence and its limits. Decide what to test or change. If the evidence is weak, the next decision may simply be to ask a better question.
You can begin with desk research: review public information or records you already have. This can reveal terminology, common alternatives and questions worth asking. It cannot tell you exactly why a particular person left your form unless that person tells you or you observe their experience. Do not treat a few online reviews as a representative survey of all potential customers. Business.gov.uk advises going beyond friends and family to hear from people who resemble future customers (Business.gov.uk, no date).
Dhruvi Infinity has a more advanced market research and startup validation guide. It combines broader market, competitor and customer evidence. This chapter concentrates on the beginner skill of asking and learning from people. You can read the advanced guide later; it is not a prerequisite.
Choose a method for the question
An interview is a conversation in which you ask, listen and follow up. It is useful for discovering the language, context and reasons behind an experience. A survey asks the same questions of multiple people, often to count responses. Observation means watching what someone actually does in a relevant setting or task, with appropriate permission. A usability test is a more specific observation: ask someone to try a page or product and watch where it helps or confuses them. Interviews gather reported experience; a usability test shows behaviour during the test. Neither method alone answers every question (Rosala and Pernice, 2023).
Figure 3
Figure 3 in words: Interview when you need to understand what happened and why the person thinks it happened. Observe a real task when you need to see where a process breaks. Survey when you have a clear question and need comparable answers from a suitably selected group. Combine methods if the decision requires both explanation and a count.
Do not choose a survey just because it produces a chart. If you do not know the options people consider, a multiple-choice list can force answers into your assumptions. Conversely, an interview with a few people cannot tell you “63% of the market wants this.” Each method has a purpose and a limit. The OpenStax marketing text describes exploratory interviews and secondary data as ways to learn about a problem before larger descriptive research (Gomez Albrecht, Green and Hoffman, 2023).
A direct example without invented customers
Suppose the question is “Why do visitors leave an appointment form?” A short interview about the last time someone booked a service can reveal what information they expected. Watching someone try your actual form can reveal a button they did not notice or an error message they could not understand. A survey could later ask a clearly defined group which missing information matters most, once you know the likely choices. These are proposed methods, not reports of interviews or results. You would still need real participants and a functioning test to learn anything about your own page.
Ask people who can speak from relevant experience
The first recruitment question is “Who has faced the decision I am studying?” If the topic is making an online appointment, people who have recently looked for such a service may be relevant. Existing customers can tell you about their experience with you; people who considered but did not choose you can reveal different obstacles. Neither group stands in for everyone.
Sampling means deciding which people will be included in your research. A small set of willing participants is usually a convenience sample, not a random sample. It can reveal useful questions and patterns, but you cannot claim its percentages represent the whole population. Note who was included and who was missing. Five conversations with close friends are five conversations with people who know you; they are not evidence that all future customers share their views.
Figure 4
Figure 4 in words: Recent users can describe an actual experience; people who chose an alternative can explain a different decision; people who abandoned a process can help identify an obstacle; friends can help pilot the wording of your questions. Each view has a limit. Recruit for the question and record how each person was found.
For a first learning round, you might invite a few relevant people and treat the conversations as exploration. There is no magic interview count that validates a market. If answers differ sharply, investigate why rather than averaging the differences away. A person with urgent needs, a long-time customer and a first-time visitor may face different decisions; chapter 6 will show how to describe a useful first audience without pretending everyone is alike.
Be honest in the invitation: explain the topic, likely length and that you are learning, not selling. Make it easy to decline. If you offer an incentive, record it because it can influence who responds. Avoid recruiting only enthusiastic followers and then treating praise as independent evidence of demand.
Write questions that invite an honest answer
The most common problem is a leading question: wording that suggests the answer you want. “Wouldn’t a simpler booking page be helpful?” already points towards “yes.” A neutral prompt asks about a real occasion before presenting your solution: “Tell me about the last time you tried to book a service online. What happened?” Then ask, “What were you looking for at that point?” Nielsen Norman Group explains that leading questions can affect the validity of the answers, and recommends broad, open prompts followed by focused questions (Rosala and Pernice, 2023; Rosala, 2021).
An open question lets the person explain in their own words. “What happened after you opened the form?” is open. “Did you click the blue button?” is closed; sometimes it is useful for a precise check, but it can end the conversation too early. A probe is a follow-up that asks for detail without inserting your explanation: “What made that difficult?” or “Can you walk me through the next step?” A05-05-question-rewrites-table.webp77.3 KB
Figure 5 in words: Replace “Would you use our service?” with “How did you handle this the last time it came up?” Replace “Is our form easy?” with “Show me how you would request an appointment.” Replace “Do you like this feature?” with “What information did you need before deciding?” The revised questions seek specific behaviour and context rather than approval.
Future intentions are especially easy to overread. A person may sincerely say they would use a service but act differently when time, money or another person’s approval is involved. Ask what they did recently, what alternatives they considered and what the outcome was. If you later want to test a proposed solution, show it clearly and observe the response; do not claim that a polite interview answer proves willingness to pay (Osterwalder, 2018).
A short guide you can actually use
Write four or five prompts, not a rigid speech. Start with context: “When did you last face [the situation]?” Move to behaviour: “What did you do first, and what happened next?” Ask about alternatives: “What else did you consider?” Explore obstacles: “What, if anything, slowed you down?” End with an open invitation: “What did I fail to ask that matters here?” Use the same main topics with each person, but follow up where their experience is specific or surprising. Pilot the guide with one person and rewrite confusing questions before you rely on it (Rosala, 2021).
Figure 6
Figure 6 in words: Explain why you are asking and how notes will be used. Invite a recent, concrete occasion. Follow the sequence of what the person did. Ask what else they tried and what got in the way. End by checking your understanding, not by pushing your product. You can change the order when the conversation naturally takes you there.
During the conversation, listen long enough for an unexpected answer. Do not fill a pause with your preferred solution. If someone says “It took too long,” ask what part took time, what they expected and how they responded. Do not translate it immediately into “They need an app.” The same words can point to very different problems. A brief recap such as “Have I understood that the timing was uncertain?” lets the person correct you without demanding that they agree.
If the participant asks what you are building, answer honestly. You can say what you are exploring and explain that you would like to understand their experience before describing a proposed fix. You do not need to hide your interest or pretend to be an independent academic researcher. Equally, do not spend the remainder of the session persuading them to approve the idea. If you later want feedback on a prototype, make that a clearly identified second part: first learn about their situation, then ask them to try something concrete. Keeping these tasks distinct helps you understand whether they had the need before they saw your solution.
You may hear a word you do not understand, or discover that the participant uses a term differently from you. Ask what they mean in this situation. The same label, such as “reliable” or “quick,” may refer to a response time, a guaranteed appointment, a clear confirmation message or something else. Write down the detail, not just the attractive adjective. This is especially useful when you later write marketing copy: use customers' meaning accurately rather than lifting a dramatic phrase out of context.
Record evidence without turning guesses into quotes
As soon as practical, write the participant’s actual words where they matter, then your interpretation separately. If you did not record a verbatim quotation, paraphrase it and label it as your note. Do not add dramatic language that the person did not use. A simple note can include the research question, date, broad participant context, how they were recruited, the relevant action or reported experience, a short accurate excerpt if you have one, your interpretation, and an open question.
Figure 7
Figure 7 in words: Layer 1 records what the person said or did. Layer 2 notes the situation and your interpretation. Layer 3 marks what is still unknown or could have another explanation. Layer 4 proposes a next check. Do not rewrite layer 2 as if it were the participant’s quotation.
Here is the distinction without inventing a participant response. Observation would be “the participant submitted the form and saw an error message,” if you actually watched that happen. Self-report would be “the participant said the form failed last week,” if that is what they told you. Interpretation would be “the error may reduce completed requests.” The interpretation is a hypothesis to test, not an additional observation.
An interview can reveal what someone remembers, believes or says they did. It cannot show every detail of their past behaviour. Observing a test can reveal what happens in that test; it does not prove that every visitor will behave the same way. This distinction keeps the evidence useful instead of making it sound stronger than it is.
Respect people and their information
Tell participants what the discussion is for and how you intend to use notes. Ask separately before recording audio or video; a written note may be enough. Keep only information relevant to the question, protect it appropriately and decide when to delete it. The UK Information Commissioner’s Office explains data minimisation as holding personal data that is adequate, relevant and limited to what is necessary for a specified purpose (ICO, no date). Participation permission is not a shortcut around a business’s wider data protection responsibilities; check the ICO’s small organisation guidance if you will collect identifiable information.
Avoid copying a customer’s sensitive account, health or financial details into a general research notebook when you do not need them. If a person asks not to be recorded, stop recording. If they choose not to answer a question, move on. Good research depends on trust and accurate context; pressure tends to produce worse evidence.
Turn several notes into a decision
After each conversation, write one sentence answering: “What did this add or challenge?” At the end of a small round, compare notes by theme—a recurring issue or need—while preserving differences. A theme is not a vote count. If three people mention timing and one says price is the real obstacle, the fourth person may represent an important group rather than noise. Look for the situation in which each issue appeared.
Figure 8
Figure 8 in words: Record a possible recurring pattern, any conflicting evidence, what you still cannot conclude and the smallest useful next test. Leave the cells blank until you have real notes. A handful of interviews can suggest where to look next; it cannot calculate the percentage of an entire market with that need.
One sensible next test might be to revise an unclear sentence and observe a few relevant people trying to complete the task. Another might be to ask a broader and better selected group a short survey question. If your first interviews reveal that you recruited the wrong group, the next step is to recruit differently. A decision to gather better evidence is still a decision.
If you already have an operating business, compare interview notes with what customers actually ask in service messages, returns or support conversations. Those records may show repeated friction, although they tend to include people who contacted you and miss those who quietly left. If you are prelaunch, your first useful evidence may simply be a clearer account of the current alternatives and the situation in which the problem occurs. Neither starting point requires inventing a survey percentage. Record the source of every observation and the group it can speak for.
Do not erase negative or contradictory responses. Write what would change your mind: “If people can find the price but still leave because the available times do not fit, changing the price label alone will not solve the problem.” This is a conditional test plan, not a claim that such behaviour has already occurred.
Figure 9
Figure 9 in words: Start with a decision, collect relevant accounts or observations, identify a possible explanation, test the smallest useful change, and update the question. The cycle can also end in “stop this idea” if the evidence points there. Learning does not have to become a positive story.
What the findings cannot tell you yet
“Everyone wants it.” A few interviews do not support a claim about everyone. Say who you spoke to, how you found them and what they described. If you need an estimate of how common a need is, plan a larger, well-designed measurement step.
“They said they would buy.” A stated intention is one kind of evidence, but a real choice with time, money and alternatives is stronger for that particular decision. Do not demand a purchase during an exploratory interview; separate learning about a problem from a later offer test (Osterwalder, 2018).
“We discovered the answer in one conversation.” An unusual response may be a valuable clue. Check it with other relevant people or a different method before building a whole strategy on it. Equally, do not discard it just because it is inconvenient.
“We interviewed users, so the site is usable.” An interview asks people to describe experiences. A usability test asks them to try a task while you observe. If the problem concerns a button, form or page, watching the task is often more direct than asking whether it looks easy (Rosala and Pernice, 2023).
Try it: prepare a first research note
You can do this exercise before you have any participants. Keep your answers short enough to fit on one page.
Decision: Write the choice you need to make. “Which information belongs on the first booking screen?” is more useful than “understand our market.”
Research question: What do you need to learn to make that choice? Write it as a question you do not already know the answer to.
Relevant people: Describe who has recently faced the situation and how you might invite them without pressuring them. Note who your initial group will miss.
Method: Choose interview, observation or survey and give one reason. For a broad unknown, start with a few conversations; for a specific page problem, watching a task may be clearer.
Interview prompts: Write four or five open prompts about a recent experience. Rewrite any question that invites praise or predicts a future purchase.
Notes: Create separate spaces for what was said or observed, your interpretation, uncertainty and a next check. Decide what personal data you actually need.
Decision rule: State what you might change after seeing the evidence and what would make you ask more people instead.
If you are not ready to approach someone, pilot your prompts with a willing person solely to test whether the wording makes sense. Mark that as a question pilot, not customer validation. When you do speak to relevant people, record the date and method. Keep the original notes so that you can revisit your interpretation.
Check your understanding
Consider the question: “Would you love a faster booking app?” Why is it weak? It introduces a proposed solution, suggests a positive answer and asks someone to predict future behaviour. A better starting point is “Think of the last time you booked this type of service. What did you do, what took time and what alternatives did you consider?” If the decision concerns a particular app screen, invite the person to try a task on the real design as a separate observation.
Watch, if video helps you
The written method above is complete without video. These optional explanations show interview mistakes and a fuller professional discussion.
Start with common mistakes:Five User Interview Mistakes to Avoid (in 5 Minutes), NNgroup, 2022, approximately six minutes according to the publisher’s page. It demonstrates the difference between useful follow-up and leading a participant. This is a professional UX education source, not a peer-reviewed paper.
Go deeper:How To Talk To Users | Startup School, Y Combinator, 2022. Gustaf Alströmer discusses whom to approach, how to interview and how to interpret answers. The discussion is oriented towards startups; adapt the method to your business and skip any examples that do not fit.
Learn how to choose a first audience using evidence, distinguish a segment from a persona, and make a practical persona sheet with assumptions and unknowns clearly labelled.
A beginner-friendly route through 26 short lessons on customers, offers, channels, content and measurement. Start anywhere, with no account or paid tools.