Interviews matter, but they are not enough to understand an opportunity. Sometimes we need to watch what happens, review records, consult statistics, or listen to people who are not yet customers.

Choose the method based on the question, not the tools at hand

Opportunity research combines primary sources, created through direct study of a situation, and secondary sources, drawn from existing records, studies, or statistics. The SBA distinguishes these approaches and explains that existing data help answer broad questions, while direct research adds details specific to the business and its customers [R01]. GOV.UK includes interviews, observation, record reviews, and earlier studies in discovery [R02].

A popular tool is not necessarily the right one. A rigorously designed survey may quantify certain responses but fail to explain why someone abandoned a task. Observing a workday can uncover interruptions and adjustments, but on its own it cannot estimate how many establishments experience them.

If the question calls for reconstructing experiences and motivations, learn to plan customer discovery interviews.

Method 1. Direct and contextual observation

Contextual observation examines how someone performs an activity in their usual environment: a shop, office, home, or workplace. You may discover interruptions, improvised tools, dependencies between people, and steps the participant forgets to mention. GOV.UK describes contextual research as a way to understand real tasks, constraints, and solutions used in the person's environment [R05].

Before observing, define the process you want to understand, obtain permission, agree on how visible data will be handled, and avoid interfering with the activity. Record the sequence, friction points, exceptions, and outcome. Afterwards, ask about decisions you could not infer. Watching someone switch among three screens does not prove the process is bad: that combination may respond to constraints you have not yet understood.

Method 2. Desk research and secondary data

Consult public statistics, industry reports, censuses, business registries, methodological literature, authorized internal data, and official documentation. For every source, record who produced it, its date, territory, population covered, definitions, units of measurement, and limitations. A nationwide report may help establish context while representing a neighborhood poorly; a figure from ten years ago may not describe today's market.

In Peru, for example, the INEI offers Perú en Cifras, with indicators by territory based on censuses, surveys, and administrative records [R15]. In the United States, the Census Bureau provides tools such as Census Business Builder for studying populations, locations, and economic activity [R11]. These are examples of data sources, not automatic evidence of demand for any particular idea.

If you will use those data to size an opportunity, you also need an explicit market estimation method.

Method 3. Exploratory surveys

A survey can help organize responses from a defined group and detect subjects worth deeper research. Its purpose should be clear: “How often do participating establishments handle reservation changes?” is more useful than “Would you like an innovative app?” Avoid double-barreled questions, ambiguous scales, and options that force positive answers.

Survey quality depends on the sampling frame, recruitment, wording, response rate, and possible nonresponse bias. Sharing a form with followers or friends may generate leads but usually does not justify conclusions about an entire population. Do not confuse calculated percentages with statistical representativeness: 80% of 20 volunteer responses does not necessarily describe 80% of the market.

Method 4. Analysis of existing records and behavior

Where you have legitimate access, examine support tickets, order histories, operational logs, cancellations, service times, or internal search activity. These can reveal bottlenecks that do not emerge when you ask “What problem do you have?” But records may be incomplete: a complaint that was never filed will not appear in support data, and a lost order may leave no trace.

First establish what each field represents, who is missing from the dataset, and how the records were created. Do not collect unnecessary personal data or repurpose information in ways incompatible with its original use. Consent and privacy matter for notes, recordings, and research records alike [R09].

Method 5. Mapping the current experience

Reconstruct a task from trigger to resolution. For reservations: request, checking availability, confirmation, change, arrival, and completion. At each step, note the person involved, channel, effort, waiting, and observed errors. The map can combine interviews, observation, and documents; it helps reveal where the problem begins and what information is missing.

Do not confuse the current journey with designing a future product. First describe what actually happens. GOV.UK's discovery research guidance finds experience mapping useful for understanding needs and obstacles [R02].

For more detail, see how to design the customer journey and relationships.

How to combine methods without calling everything “validation”

Triangulation means comparing what you learn through different sources or methods. If you interview customers who mention delays, observe waiting times, and review operational records, you gain complementary perspectives. Three weak pieces of evidence do not automatically create a strong truth; they may share biases or originate from the same underlying source.

To distinguish when several methods strengthen a conclusion from when they repeat the same bias, see how to collect and evaluate evidence without confirmation bias.

This matrix can help plan a research round:

Question Primary method Possible cross-check Caution
What causes service delays? Observation Employee interview Do not infer causation automatically
What options do people use today? Interview Public review of alternatives Do not equate availability with real use
How does demand vary by month? Historical records Official industry data Check seasonality and coverage
Which groups experience the problem? Interviews across segments Well-designed survey Do not generalize a volunteer sample
How many establishments exist locally? Official directory or statistics Observation and count Check duplicates, closures, and dates

Exploratory research versus experimentation

Watching how a restaurant works today aims to understand a situation. Inviting it to use an alternative system for two weeks to measure errors introduces an intervention: that is an experiment requiring clear criteria. In pillar 1, you can learn what should be measured and why; pillar 3 covers hypotheses, criteria, experiments, and decisions based on results [R10].

The distinction does not depend on whether a form or digital tool is used; it depends on the question and study design. An exploratory survey can identify reported behavior; it does not automatically prove real willingness to buy.

To move from understanding the context to examining a specific claim, use the experiment selection guide.

Example: a natural products service

A store suspects some customers forget to restock products used regularly. It interviews buyers about their most recent replenishment, observes how they check dates or containers, and—where permitted—reviews anonymized purchase histories. It finds regular purchases but also pauses in consumption and purchases for other people. A survey may suggest how common a difficulty is among respondents, but it cannot be projected to the whole market without an appropriate sample.

Before proposing automatic reminders, the actual finding is that different replenishment behaviors exist. The next questions are who exhibits them, when they change, and what consequences forgetting a purchase has. Testing a solution later would require a dedicated experiment.

Criteria for selecting and documenting methods

For each method, ask: “What decision will this answer change?”, “Can we observe what we need?”, “Which people or cases will be excluded?”, “What resources and consent are required?”, and “What complementary evidence would help?” Record date, sample, instrument, limitations, and responsible people. A simple method carried out well is more valuable than a sophisticated combination that answers no meaningful question.

Common mistakes

Using surveys to replace all observation; treating global trends as local behavior; interpreting website visits as buyers; ignoring missing data; observing without permission; confusing research with a commercial experiment; and claiming certainty from sources based on the same original information.

What you should have when you finish

A research-methods plan with questions, sources, recruitment, risks of bias, consent where relevant, recording methods, and reasons for combining techniques. With that plan, you can investigate without turning each tool into an end in itself.

Main sources: [R01] SBA; [R02] GOV.UK, discovery; [R05] observation; [R09] privacy; [R10] Strategyzer; [R11] Census Bureau; [R15] INEI. The example is hypothetical.

Bibliography and references