A perfect product is not the same as a useful product. The practical way to discover what people need is to talk with potential users early, observe how they handle real tasks, and test small solutions before investing in a polished launch. This approach—often called customer discovery or user research—helps teams replace assumptions with evidence. It matters because product development can be costly: CB Insights’ analysis of failed startups found that “no market need” was the most frequently cited reason for failure, appearing in 35% of its post-mortem sample. The lesson is not to abandon quality, but to learn what quality should mean to the people a product is meant to serve.
Customer discovery reduces the risk of building the wrong thing
Customer discovery is the structured process of investigating who has a problem, how they currently address it, and whether a proposed solution would make a meaningful difference. Steve Blank, an entrepreneur and educator associated with the Lean Startup movement, describes customer development as a process of testing a business model’s hypotheses with customers rather than treating the initial plan as fact. In practice, discovery includes interviews, observation, surveys, prototype tests, and analysis of existing customer behavior.
These methods are related but not interchangeable. Interviews help explain motivations and context; observation reveals workarounds people may forget to mention; surveys can measure how common a pattern is across a larger group; and prototype tests show whether people can use a proposed solution. Combining methods makes it less likely that a team will mistake a confident opinion for evidence of a real need.
Customer interviews uncover context, not just opinions
A customer interview is a focused conversation designed to understand a person’s past experiences, decisions, and difficulties. Strong interviews ask about specific recent events—“Tell me about the last time you tried to do this”—rather than inviting speculation such as “Would you use our app?” People are generally better at describing what they did than predicting what they might do.
The interviewer should listen for concrete details: what triggered the task, which tools were used, where delays occurred, what the person tried next, and whether the problem had a cost. The Nielsen Norman Group’s guidance on user interviews emphasizes open-ended questions, follow-up questions, and avoiding leading prompts. A small set of conversations can reveal recurring themes, but it should be treated as qualitative evidence—not as a statistically representative measurement of an entire market.
Observation reveals the gap between stated needs and actual behavior
Contextual observation means watching people perform a real task in the environment where it normally happens. It can reveal interruptions, informal workarounds, confusing handoffs, and constraints that are easy to miss in a conference-room conversation. For example, a team designing scheduling software might learn by observing a clinic receptionist that the central problem is not finding an appointment slot but coordinating several systems while patients wait.
Observation is especially useful when a task is routine, complex, or difficult for users to describe. Researchers should obtain permission, protect private information, and avoid disrupting work. Notes should distinguish what was directly observed from the researcher’s interpretation, so that an early theory does not quietly become “the user’s stated need.”
Small experiments turn customer needs into testable decisions
Once a team has identified a plausible need, it can test its assumptions with the smallest credible experiment. Eric Ries defines a minimum viable product as a version that enables a team to collect the maximum amount of validated learning about customers with the least effort. This is not a license to release careless or unsafe work. It is a way to learn before spending heavily on features, infrastructure, or a full launch.
Experiments might include a paper sketch, clickable mock-up, concierge service, landing page, or limited pilot. The best test is determined by the riskiest assumption. If the uncertainty is whether people understand a workflow, test a prototype. If it is whether a problem occurs often enough to support a business, interview and survey a carefully defined audience. If it is whether a service can be delivered reliably, run a small operational pilot.
Prototypes test usability before a product is polished
A prototype is a representation of a proposed product or service that people can react to or use. It may be a paper drawing, a simple interactive screen, or a working feature with limited scope. By asking participants to complete realistic tasks, teams can identify confusion and breakdowns while changes are still relatively inexpensive.
The System Usability Scale, developed by John Brooke, is a widely used 10-item questionnaire for measuring perceived usability. It is a measurement tool, not a substitute for observing behavior or diagnosing a problem. Teams should pair scores with task completion, errors, time on task, and participants’ explanations. A prototype test should answer a defined question, such as whether users can find a setting, rather than simply asking whether they “like” the design.
A minimum viable product tests whether a solution delivers value
A minimum viable product is a deliberately limited version of a solution that tests an important customer or business hypothesis. It might serve one user group, support one essential task, or operate in one location. The purpose is to learn from real use while keeping the experiment manageable—not to label an unfinished product “minimum viable” and excuse missing essentials.
Teams can track measures such as task completion, repeat use, time saved, support requests, conversion, or customer retention, depending on the hypothesis. A metric becomes useful when it is tied to a decision: continue, revise, or stop. For instance, a pilot that attracts sign-ups but sees few users complete the core task may indicate that the promise is appealing while the experience is not yet useful.
Evidence-based iteration makes learning part of the work
Iteration is the cycle of testing an idea, examining evidence, and changing the product or the underlying assumption. It connects discovery to delivery: research should not be a one-time project before design begins, but an ongoing practice as teams learn about different users, situations, and barriers.
The Government Digital Service’s service-design guidance encourages teams to understand users and test services with them throughout development. This principle is relevant beyond government: a product that works for an early enthusiast may fail for someone using assistive technology, working on a slow connection, or dealing with a time-sensitive task. Repeated testing helps reveal those differences before they become expensive defects or exclusions.
Inclusive research helps teams avoid designing for an imagined average
Inclusive research deliberately recruits people with varied needs, abilities, backgrounds, and levels of experience. The goal is not to treat one participant as a spokesperson for an entire group, but to uncover different contexts and make design assumptions visible. The World Health Organization estimates that 1.3 billion people—about 16% of the global population—experience significant disability, underscoring why accessibility is a core product concern rather than a niche add-on.
Recruit participants who reflect the people expected to use the service, including those who may face access barriers. Offer accessible research materials and accommodations, explain how recordings or notes will be handled, and compensate participants fairly where possible. Accessibility checks should also be paired with usability testing: standards can identify many technical barriers, while people’s experiences show how those barriers affect real tasks.
A repeatable feedback loop turns findings into action
A feedback loop is a recurring process for collecting user evidence, sharing what it means, making a change, and checking whether the change helped. A simple loop is: identify a question, recruit relevant participants, observe a task, record patterns, choose an action, and test again. This keeps research connected to product decisions instead of leaving findings in a report that nobody uses.
To make the loop practical, maintain a short research log with the question, participants, observations, limitations, and resulting decision. Separate repeated evidence from one-off comments, and keep track of unanswered questions. If findings conflict, investigate differences in user context rather than selecting only the evidence that supports the team’s preferred solution.
How to start asking real people what they need
A team does not need a large research department to begin learning. A small, disciplined first round can clarify whether a problem is real and what to investigate next.
- Write down the assumptions behind the idea, including who has the problem and how they solve it now.
- Choose one high-risk question that could change the direction of the work.
- Recruit people who have recently experienced the relevant task or problem.
- Ask about actual past behavior, and observe the task when appropriate.
- Summarize patterns and uncertainty separately; do not overstate what a small sample proves.
- Create the simplest safe test that can resolve the next important question.
- Share the result with the team and make an explicit decision about what to change or test next.
For example, a small-business team considering a new invoicing feature might first speak with owners who recently sent invoices. They could watch a few people complete the task, identify where errors or delays occur, and test a basic prototype of the most troublesome step. The team would then compare observed task success and user feedback with its original assumptions before building a full feature.
Conclusion: Learn before you polish
Customer discovery helps teams define the real problem; interviews and observation uncover its context; prototypes and minimum viable products test possible solutions; and inclusive, repeatable feedback loops help improve them. The 35% “no market need” finding in CB Insights’ startup post-mortem analysis illustrates the cost of building without sufficient evidence, while accessibility data from the World Health Organization highlights the importance of learning from people with varied needs.
Perfection is not the starting point for useful work. Start with a clear question, talk to people who encounter the problem, and test a modest solution. Then use what you learn to decide what deserves more time, investment, and polish.
Sources: CB Insights, The Top 12 Reasons Startups Fail, 2021, https://www.cbinsights.com/research/report/startup-failure-reasons-top/; Steve Blank, The Four Steps to the Epiphany, K&S Ranch; Nielsen Norman Group, User Interviews: How, When, and Why to Conduct Them, https://www.nngroup.com/articles/user-interviews/; Eric Ries, The Lean Startup, Crown Business, 2011; John Brooke, SUS: A Quick and Dirty Usability Scale, 1996; Government Digital Service, Service Standard, https://www.gov.uk/service-manual/service-standard; World Health Organization, Disability, https://www.who.int/news-room/fact-sheets/detail/disability-and-health.
