PT
EN
There is a fairly common practice in Product Discovery.
A team talks with five or ten users, identifies some patterns, gathers everyone in a room and concludes:
“We already understand the problem. Now we just need to build the solution.”
Qualitative interviews are extremely important.
The problem begins when they start being treated as statistical evidence.
They are not.
They help understand behaviors, motivations, and narratives, but they are rarely sufficient to answer questions like:
- How many people actually have this problem?
- How intense is that pain?
- Does the behavior observed represent the majority of users?
- Is it worth investing millions in this direction?
These are different questions.
And they require different methods.
Qualitative Research Explains the “Why”
Individual interviews are excellent for discovering what we do not yet know to ask.
They allow exploring hypotheses, understanding language, identifying hidden needs, and revealing unexpected behaviors.
In just a few conversations it is possible to learn a great deal.
But there is an important limit.
The fact that five people report the same difficulty does not mean it is frequent in the population.
It only means it appeared in that small group.
It is an excellent hypothesis.
It is not yet a conclusion.
The Moment to Bring in Statistics
After hypotheses emerge, quantitative research enters.
A simple practice consists of turning the main learnings from the interviews into a structured questionnaire.
This form can measure:
- frequency of use;
- intensity of the problem;
- satisfaction;
- priorities;
- adoption intent;
- value perception.
When the research reaches approximately forty respondents or more, we start entering a scenario where inferential statistical techniques can be applied much more safely, provided the sample design is appropriate to the research objective.
This does not mean forty people represent an entire population.
It means it is already possible to start estimating relationships between variables and building confidence intervals for that sample universe.
Charts Tell Stories
Many teams limit their analyses to the average of responses.
The average rarely tells the whole story.
Histograms show how opinions are actually distributed.
Sometimes the average indicates moderate satisfaction. The histogram reveals a different reality: half the users love the product, the other half hate it. The average hides this polarization.
Scatter plots also provide valuable information. They help identify relationships between variables. For example:
- do more experienced users actually use a certain feature more?
- is higher usage frequency associated with greater satisfaction?
- do customers who use a specific feature show a lower cancellation rate?
Visualizing data tends to reveal patterns that go unnoticed in tables.
Confidence Intervals Matter
Another common mistake is interpreting any percentage as an absolute truth.
Every research study works with uncertainty.
When we calculate confidence intervals, typically at levels like 95%, we are making that uncertainty margin explicit.
This makes decisions far more transparent.
Instead of stating:
“72% of customers prefer this feature.”
We can say:
“With a given confidence level, we estimate that the preference falls within an interval consistent with the data collected.”
This difference seems small.
In practice, it completely changes the quality of the decision.
Every Sample Has Biases
There is another frequent trap.
Believing that a survey represents any audience.
It never does.
It represents only the population from which it was drawn.
Imagine a survey on the quality of banking apps. The results will be very different depending on who responds.
Software developers tend to observe architecture, stability, and technical experience. Financial sector specialists may value regulatory requirements. Regular customers typically assess ease of use, speed, and trustworthiness.
None of these groups is wrong. Each sees the product from a different perspective.
The problem is not choosing a specific sample.
The problem is forgetting its limitations.
Never Extrapolate Beyond Your Sample
This is perhaps one of the most important principles of scientific research.
Conclusions obtained from one group cannot be automatically extended to other groups.
Results obtained with technology professionals do not represent the entire population. Research conducted in one city does not necessarily describe another country. Studies conducted with premium customers hardly explain the behavior of low-income customers.
Every conclusion must respect the limits of the analyzed population.
Discovery Is Not Just About Listening to Customers
There is another frequently overlooked aspect.
Innovative companies do not live only by discovering existing needs.
They also create new markets.
In the dynamic capabilities proposed by David Teece, this appears clearly in the three major organizational capabilities:
- Sensing — identifying opportunities and changes in the environment.
- Seizing — capturing those opportunities through new products, business models, and investments.
- Transforming — continuously reconfiguring the organization to sustain new competitive advantages.
Many teams concentrate almost all their effort on just the first stage.
They listen to customers. Conduct interviews. Map journeys. Validate problems.
All of this is important.
But insufficient.
Innovation also requires imagining solutions that have not yet been explicitly requested by the market.
Not Everything Can Be Discovered by Asking
A phrase often attributed to Henry Ford summarizes this dilemma well:
“If I had asked people what they wanted, they would have said faster horses.”
Regardless of the exact authorship of the quote, the message remains relevant.
Customers are excellent at explaining their problems.
They cannot always imagine radically new solutions.
It is precisely in this space that strategic vision, experimentation, and innovation enter.
Conclusion
Qualitative interviews remain one of the most important tools in Product Discovery.
But they represent only the beginning of the investigation.
Product decisions gain far greater robustness when qualitative hypotheses are complemented by quantitative research, statistical analyses, and a clear understanding of sample limitations.
At the same time, companies cannot depend exclusively on what the market can verbalize.
Listening to customers is essential. Measuring correctly is indispensable. But innovating also requires creating possibilities that customers themselves cannot yet see.
The balance between discovery, validation, and transformation is what differentiates organizations that merely follow the market from those that help shape its future.