Validation & Idea Testing
Primary research means collecting data yourself. Learn the five main methods, see examples for a new business, and know when existing data is enough.

Primary research is information you collect yourself, straight from the people or situations you want to understand. For a new business, it answers the questions no published report can: whether your customers want your specific offer, how they solve the problem today, and what they would pay.
The methods are old and well tested. What changes for a founder is the order you use them in and how small you can start.
Primary research is data gathered firsthand to answer your own question. You write the questions, choose who to ask or watch, and own the raw results. When the goal is a business decision, people often call it primary market research. The definition is the same.
Secondary research works with data someone else already collected: government statistics, industry reports, competitor websites and published studies. It is faster and often free, but it was built for someone else's question.
The U.S. Small Business Administration draws the same line in its market research guidance. Existing sources save time and energy but may not be specific to your audience. Asking consumers yourself gives a more nuanced picture of your target customer, at the cost of time and sometimes money.
Primary data comes in two flavors:
Observation can produce either, depending on whether you take notes or tally behavior.
The two are partners. Secondary research frames the market, and primary research tests your specific idea inside it.
| Factor | Primary | Secondary |
|---|---|---|
| Who collects it | You | Someone else |
| Cost | Your time, sometimes small incentives | Often free |
| Speed | Days to weeks | Hours |
| Specificity | Your customer and your offer | Broad market, industry or region |
| Best for | Reactions, reasons, willingness to pay | Market size, demographics, trends |
A sensible order is secondary first, then primary. Secondary market research tells you whether the market is large enough to bother and who already serves it. For example, the Census Bureau's County Business Patterns series reports the number of establishments by industry and employment size, for areas below the national level. It only covers establishments with paid employees, so one-person operators will not show up in the count.
That kind of data gives you a starting map of competitors. A competitor analysis fills in what they charge and how they position themselves. Neither tells you whether your customers will switch, which is where firsthand research takes over.
Each method below includes what it is good for, where it falls short, and an example from one hypothetical founder: someone planning a mobile dog-grooming van in their city.
An interview is a one-on-one conversation where you ask open questions, listen and follow up. In UX work it is called a user interview. In market research and social science, the same method goes by in-depth interview, semi-structured interview or qualitative interview.
Interviews are the best tool for finding out whether a problem is real, how often it happens and what people do about it now. Their weakness is that people report their behavior from memory, and memory is unreliable. A skilled interviewer also avoids leading questions, because those bend the answers toward what you hoped to hear.
Tip: ask about the last time something happened, not what someone would do in the future. "Walk me through your dog's last grooming appointment" gets a story. "Would you use a mobile groomer?" gets a polite yes.
For example, the grooming founder talks to ten dog owners. Seven mention the drive and the two-hour wait at the shop. Two say they would pay more to skip it.
A survey asks many people the same fixed questions so you can count the answers. Surveys work best after interviews, once you know which questions matter and which answer options are realistic.
Question wording decides whether the results mean anything. Pew Research Center warns against double-barreled questions, which ask about two things at once and produce answers that are hard to interpret. It also notes that agree-disagree questions can trigger acquiescence bias, a tendency for some respondents to agree with whatever statement they see.
Tip: pretest every survey on a handful of people before you send it widely. Watch where they hesitate or ask what a question means.
For example, the grooming founder writes an eight-question survey and shares it in local neighborhood groups. It asks how often owners book grooming, what they pay now (in price ranges) and how far they drive. Keep in mind who answers: people in online groups are a self-selected slice of your market.
Observation means watching what people actually do, in the place where they do it. It closes the gap between what customers say and what they do, which interviews and surveys cannot close on their own.
Observation can be as simple as mystery shopping, standing behind a counter for a day, or watching someone use a competitor's product. Take notes on timing, workarounds and moments of friction.
For example, the grooming founder calls five local groomers as a customer and asks for the next open appointment. Three cannot book a full groom for more than a week. That is a capacity gap no survey answer would reveal.
A focus group is a moderated discussion with a small group of target customers. It is useful for hearing reactions to a concept, a name or a set of packages, and for listening to customers debate trade-offs among themselves.
The limits are real. A confident participant can steer the room, and the group still reports opinions rather than behavior. Use focus groups after interviews, to react to options you have already shaped.
For example, the grooming founder invites six survey respondents who agreed to follow up. They see three service bundles at different prices. The group splits on whether nail trims should be included, which tells the founder to offer them as an add-on.
A field test puts a real offer in front of real buyers and measures what they do. A smoke test is a lightweight version: a landing page with a booking button, a pre-order form or a pop-up at a market. These are the strongest signals in primary research because they measure commitment instead of opinion.
Be honest with the people who respond. If the service is not live yet, say it is launching soon and refund any deposits if you do not go ahead. The idea validation experiments guide walks through smoke tests, fake doors and paid pilots in more depth.
For example, the grooming founder posts a simple page offering launch-week slots with a refundable deposit and shares it in the same local groups. The share of visitors who leave a deposit is the strongest demand signal so far.
Start from the decision you need to make and choose the method that can answer it.
| Your question | Method | Why |
|---|---|---|
| Is the problem real and frequent? | Interviews | Stories of the last time it happened |
| How many people share it? | Survey | Counts across a larger group |
| What do people do today? | Observation | Behavior instead of recall |
| Which version of the offer lands? | Focus group or interviews | Reactions and reasons |
| Will they pay? | Smoke test or pre-order | Commitment with money or time |
| How big is the market? | Secondary research | Someone has already measured it |
Online signal, such as review mining and forum threads, sits between the two categories: the data already exists, but you collect and code it yourself. The bootstrapped market research methods guide covers those free channels. For a fill-in structure that ties your assumptions, competitors and interview questions together, use the market research template.
A few habits quietly ruin results, even when the method is right.
AI is useful for the preparation work around primary research. You can ask it to turn your assumptions into an interview guide, rewrite survey questions that ask about two things at once, or group a stack of interview notes into themes.
It cannot answer the questions for your customers. A model summarizes text that already exists, which makes its output secondary research at best. The answers that count still come from real people. EntraWorld's AI business tools include market research alongside business plans, financial models and brand identity, and they work by asking you guided questions.
Pick the one assumption that would sink the business if it were wrong. Write down the result that would make you continue, then book five interviews with people who match your customer and are not your friends.
By the end of the week you will know whether the problem is real enough to survey and test. If it is, the next step is a smoke test with a real offer. When you are ready to turn what you learned into a plan, join EntraWorld free and build it with the tools and founder network in one place.
Start free. The first 5,000 Premium memberships include a full year of every tool.
Join EntraWorld free →