Market Research Methodology

Primary Versus Secondary Research

Secondary research uses data that already exists — industry reports, government statistics, competitor filings, published academic work. Fast and cheap. Limited by whether anyone has asked your question.

Primary research generates new data — surveys, interviews, observation, experiments. Answers your specific question at the cost of time and money.

Exhaust secondary sources before commissioning primary. A significant proportion of research budgets is spent rediscovering published findings.

Qualitative and Quantitative

Qualitative methods — depth interviews, focus groups, ethnography — explain motivation and surface questions you did not know to ask. They cannot tell you how common anything is.

Quantitative methods — surveys, behavioural data, experiments — measure prevalence and test hypotheses. They cannot tell you why.

The sequence that works: qualitative to generate hypotheses, quantitative to test them at scale. Running quantitative research first means testing hypotheses you assumed rather than discovered.

Survey Design Errors

Leading questions. "How much did you enjoy the new feature?" presumes enjoyment.

Double-barrelled questions. "Was the service fast and friendly?" cannot be answered when it was one and not the other.

Stated preference versus revealed preference. What people say they would pay and what they do pay diverge substantially. Treat willingness-to-pay survey data with heavy scepticism.

Sampling bias. Surveying your existing customers tells you about people who already chose you. It says nothing about why others did not.

Sample Size and Significance

For most consumer research, several hundred responses per segment supports reasonable conclusions. Below roughly a hundred, differences between subgroups are usually noise.

For qualitative work the logic differs — you are looking for saturation, the point at which additional interviews stop producing new themes. That typically arrives between eight and fifteen interviews per segment.

Competitive and Desk Research

Useful sources that cost nothing: competitor pricing pages and their change history, job postings that reveal strategic direction, review sites for unmet needs in the category, search volume trends, and customer-facing staff who hear objections daily.

The Questions That Produce Useless Answers

Most bad research is bad at the question level, and no amount of sample size repairs it.

Do not ask people to predict their own behaviour. "Would you buy this at ₹500?" reliably overstates intent, because agreeing is free. Ask what they did last time instead — what they actually bought, what they actually paid, when.

Do not ask them to explain their preferences. People construct plausible reasons after the fact, fluently and sincerely, and those reasons are not the cause.

Do not ask two things in one question. "Was the service fast and friendly?" cannot be answered by someone who found it fast and rude.

Do not put the answer in the question. "How much did you enjoy…" has already assumed enjoyment.

The reliable pattern is past behaviour, specific and recent. It is less exciting than asking about the future and it is the only part of a survey that predicts anything.

Knowing When You Have Enough

Two different questions get confused: how many people do I need, and how many conversations do I need. They have different answers.

For qualitative work, the test is saturation — you stop when new interviews stop producing new themes. For a reasonably homogeneous group that usually arrives sooner than people expect; for a varied one, later. The number is an outcome, not an input, and running interviews to hit a target after saturation is spending money to confirm what you know.

For quantitative work, the size depends on the decision, not on a rule of thumb. Detecting a large difference needs far fewer responses than detecting a small one, so the useful question is how small a difference would change what you do. If a five-point gap would not alter the decision, you do not need to be able to detect it.

And representativeness beats volume. Two thousand responses from a self-selected list describe that list. Two hundred from a properly drawn sample describe the market. A large unrepresentative sample is more dangerous than a small one, because the numbers look authoritative.