Research: Methods and When to Choose

The method is chosen after the research question. Interviews reveal context and motivation, observation reveals real behavior, quantitative data reveals scale and distribution, and experiment reveals causal effect; one method rarely covers all four tasks.

Why you need research at the Discovery stage

Identifying problems and testing ideas

Discovery is about finding what’s really important to people. This minimizes the risk of making a product that no one will use. Research methods help to see the real tasks of users, rather than working for the sake of beautiful hypotheses.

** Example:**
Fintech often want to make another feature in the application. Interviews with users show that people are not waiting for new animations, but for clear notices of write-offs.

When to Start Discovery Research

Discovery should be launched in two cases: when you suspect that you do not fully understand the problem, or there are several ideas and analysis is necessary before choosing a direction.

Read more about discovery steps

The main methods of discovery-research

Qualitative methods: interviews, observation, diaries

Qualitative techniques are appropriate when it is important to understand motivation, process, barriers, or context. They work well at the start, when there are still a lot of ideas and little information.

User Interview (User Interview)

With the help of interviews, it is easy to find out what tasks people have, what prevents them, how they solve problems now. They are used to validate problems, prioritize and search for unexpected insights.

** Example:**
The delivery service asks real users how they choose delivery times and what prevents them from placing an order. It turns out that the biggest pain is inaccurate forecasts of couriers, not the speed of delivery.

Observational Research (Observational Research)

It’s important to see what people are doing, not just what they’re saying. This is important if there is a risk of social desirability bias – when users pass off wishful thinking.

** Example:**
In SaaS for Business, the team observes how customers’ employees work with reports and sees that the information search section is being used in a way that is not intended. Based on this, a new hypothesis for revising the interface appears.

Diary studies

Participants in diary studies record their actions or emotions according to the script. The method helps to see patterns that are difficult to fix through a one-time interview.

Description of qualitative methods

Quantitative methods: surveys, analytics, segmentation

Quantitative methods are used to test the scale of a problem or idea on a broad auditorium. They help to separate important pains from secondary ones.

Polls (Surveys)

They allow you to quickly collect statistics: how relevant the problem is, how often something happens, compare segments with each other.

** Example:**
After a series of interviews, the B2B team builds a survey of 10 questions and sends it to 100 customers to record what type of problems are most common.

Behavioral Analytics (Product Analytics)

It is used to search for patterns in an existing product: where users fall off, what scenarios are in demand.

Conclusion:
If there is a product with traffic, be sure to use analytics along with the interview.

Best Practices in Product Research

Quick Ideas Testing: Prototyping and Problem-solution fit

Prototypes and Concept Testing

Quick prototypes (sketchies, clickable layouts) are usually tested with 3-7 people from a key segment. It helps to find out whether people understand the idea, whether they see value, which is problematic.

** Case:** The mobile banking application makes a clickable prototype and gives it to users for testing. 4 out of 5 people do not find the main function. The team changes navigation before launch.

Problem-Solution Fit

Check whether there is a match between what hurts people and how the team proposes to address it in the product. For such verification, it is convenient to do “problem interviews” (without presenting a solution) and “solution interviews” (prototype display).

Problem-solution fit method (https://leanstartup.co/)

How to Choose the Right Research Method

Problem Search vs. Solution Verification

If you want to find a real problem, bet on qualitative methods (interview, observation). If you want to assess scale and priorities, use quantitative approaches (survey, analytics).

** Simple diagram:**
First you get into context with an interview, then you zoom in through a survey. If you have a product, add analytics.

Settings for the task, not according to the template

The method is chosen not according to fashion, but based on a specific question. It is important to clearly formulate a research goal before starting.

** Example:**
Early on, the idea seems interesting to IT professionals and students. To understand who the problem is most real for, the team does 3 interviews for each segment, and then sends out a short survey and catches the difference in numbers.

Standards of Method Selection for Discovery

Mistakes and anti-patterns in product research

Major errors

1. Doing surveys before interviews – there is a risk of asking about things that are not relevant to real problems
2. Ignore positive and negative patterns in analytics data
3. Confusing Idea Testing with Problem Testing: Testing UI before finding out if there really is pain

Recommendations on anti-patterns

Conduct a minimum of 5-7 interviews before any quantitative surveys Always combine qualitative and quantitative data Fixing all the original assumptions: what exactly you know is a hypothesis

Errors in Discovery

FAQ: Questions and answers

**Why do several types of research at once? **
When you combine methods, you see the full picture: not only what people do, but why.

**Can you start with analytics if the product is already on the market? Yes, but don’t just limit yourself to numbers - the analytics will show the symptom and the interview the cause.

**How do you know if the interview was successful? You got specific stories from real experience, not abstractions or wishes.

**How long does it take to validate an idea? Usually 5-7 per segment; large numbers increase data saturation, but diminishing returns come quickly.

**What’s wrong with interviews without interviews? Polls without immersion in context lead to false results: you risk asking the wrong question.

Where to look for benchmarks for discovery metrics? Platforms like Nielsen Norman Group, Product Talk and Lean Startup publish guides and examples.