Market Research: Watch What Customers Do, Then Ask Them Why

Nearly 200,000 sips

In the spring of 1985, The Coca-Cola Company had better research behind a decision than most companies ever have. It changed the formula of the world's best-known drink, "adopting a formula preferred in taste tests of nearly 200,000 consumers" ¹. New Coke launched on April 23. By June the company's consumer hotline was taking 1,500 calls a day, against 400 before. On July 11, 79 days after the launch, the original formula came back as Coca-Cola classic ¹.

The research was not wrong. People in the tests did prefer the new taste. The company's own history puts the problem in one line: "What these tests didn't show, of course, was the bond consumers felt with their Coca‑Cola" ¹. The tests answered the question they asked, which cola tastes better, and never asked the one that mattered: what will you do if we take the old one away?

Market research is the work of finding out what customers, competitors and markets are actually like, so that a decision rests on evidence rather than on what the people in the room assume. It includes studying existing data, asking people questions and watching what they do. Done well, it tells you what customers will do. Done badly, it tells you what they say.

Watch what customers do. Then ask them why.

This page covers what market research is, the main methods and what each can and cannot tell you, the gap between what people say and what they do, the free data most researchers ignore, and a step-by-step method built around a decision rather than a survey.

What market research is

The American Marketing Association's definition is the standard one: marketing research is "the function that links the consumer, customer, and public to the marketer through information," information used to "identify and define opportunities and problems; generate, refine, and evaluate actions; monitor performance" ². The words that matter there are used to. Research earns its cost only when it changes, confirms or kills a decision.

Two distinctions organize the field. Primary research collects new information for your question: interviews, surveys, observation, tests. Secondary research uses information that already exists: government statistics, company filings, industry reports, your own sales data. Separately, research is either qualitative, which explains why and how in depth with a few people, or quantitative, which measures how many and how much across many.

Market intelligence is the ongoing version of the same work: a standing watch on competitors, customers and the market, rather than a study commissioned for one decision. For competitors specifically, see the competitive analysis method.

Say versus do

People are honest in surveys, mostly. They are also human. The Pew Research Center, which runs some of the most careful surveys in the world, warns that "respondents understate alcohol and drug use, tax evasion and racial bias" and "may overstate church attendance, charitable contributions and the likelihood that they will vote in an election" ³. The bias, it notes, can be greater when an interviewer is present ³.

Clayton Christensen's milkshake study is the best-known illustration of the gap. A fast-food chain asked customers to "list the characteristics of an ideal milkshake." The answers were honest and the company acted on them, "But alas, milkshake sales did not improve." A researcher then spent a day in a restaurant watching who bought milkshakes and when, and found that "40 percent of the milkshakes were purchased first thing in the morning, by commuters who ordered them to go" ⁴. Christensen's explanation: "They faced a long, boring commute and needed something to keep that extra hand busy" ⁴. The customers had never said that, because nobody asked about the commute. They had been asked about milkshakes. The jobs to be done framework grew out of work like this ⁵.

The gap is often summed up with a line attributed to Henry Ford: if he had asked people what they wanted, they would have said faster horses. Quote Investigator, which traces quotations to their sources, "has not yet located compelling evidence that Henry Ford is responsible for this quotation" ⁶. It is a pity, because the line is usually used to argue against research at all. The real lesson is narrower. Do not ask customers to design the solution. Ask them, and better still watch them, to find the problem.

Start with the behavior you can already see

The cheapest research is often already published. Airbnb's founders did not need a survey to show demand for staying in strangers' homes. Their early pitch deck pointed to behavior that already existed: widely circulated reproductions of the deck cite 630,000 users on the temporary-housing site Couchsurfing.com and 17,000 temporary-housing listings in San Francisco and New York in a single week ¹², though the original deck itself has not been independently verified. People were already doing the thing. The open question was whether a better product could serve them.

For US markets, four free sources cover most of the ground before anyone designs a questionnaire:

• County Business Patterns (US Census Bureau): an annual series of economic data by industry, down to counties and ZIP codes; the number of establishments, employees and payroll in your market ⁸.
• The Economic Census (US Census Bureau): every five years, for years ending in 2 and 7, detailed statistics on businesses, including sales by industry ⁹.
• Consumer Expenditure Surveys (Bureau of Labor Statistics): what households spend, on what, by income and demographics ¹⁰.
• EDGAR full-text search (SEC): the full text of company filings since 2001, where competitors report their own revenue, segments, customers and risks in their own words ¹¹.

Company filings are the most underused of the four. A public competitor's annual report states its segment revenue, its largest customers' share of sales and the risks its lawyers thought worth disclosing. That is market research someone else was legally required to do. For sizing a market from these sources, see TAM, SAM, SOM and the market sizing calculator; for an industry's structure, see industry analysis.

How to run a market research project

Most research goes wrong before the first question is asked, because it starts from a topic ("let's understand our customers") rather than a decision. The steps below start from the decision.

1. Name the decisionWrite the choice the research will inform, with a date: launch or not, which segment first, what price. If no decision depends on the answer, do not commission the research.
2. Write down what would change your mindBefore any data arrives, agree what result would make you choose differently. This turns the study into a test, and stops a team from reading any result as support for what it already wanted.
3. Do the secondary research firstCensus data, filings, your own sales records and existing studies often answer the sizing and structure questions for free, and sharpen the questions only primary research can answer.
4. Watch before you askLook at what people already do: what they buy, how they work around the problem, what they abandon. Behavioral data and observation set the baseline that interviews then explain.
5. Interview about the pastAsk people to walk through the last time they faced the problem: what they did, what it cost, what they tried. Plan for ten to fifteen interviews in one segment, and stop when new ones stop telling you anything new.
6. Measure with a survey only when you know what to countUse a survey to size what the interviews uncovered, not to discover it. Ask about recent behavior rather than future intentions.
7. Test with something at stakeWhere you can, make people act: a pre-order, a pilot with a signed letter, a landing page with a price. Commitment is the closest thing to the real decision customers will make.
8. Report the decision, not the dataLead with the recommendation and the threshold it met or missed. The charts belong in the appendix.

The research plan

Fill in one row per question the decision depends on. A question without a method, a source and an owner is not yet part of the plan.

A worked example

The company below is invented, to show the plan in use.

Every number the decision turned on came from behavior or published data. The one thing the team almost did, asking people whether they would buy, is the one method that could not have answered the question.

Research has a shelf life: the Edsel

Ford's Edsel is remembered as a design failure, but its makers did more research than most. When Ford ended the line in November 1959, Time tallied the "Total cost of research, design, tooling, expansion of production facilities: $250 million" and called the car "a prime example of the limitations of market research" ⁷. Ford had run studies to give the car precisely the right personality. The problem was time: "The flaw in all the research was that by 1957, when Edsel appeared, the bloom was gone" ⁷. The research described the market of the mid-1950s; the car arrived in a different one. Our analysis of the Edsel follows the decision in detail.

Every market study is a photograph. The longer the gap between the research and the launch, the more the decision should rest on behavior you can keep watching rather than on opinions captured once.

Before you commission research: a checklist

• The research is tied to a named decision with a date.
• We have written down, in advance, the result that would change our mind.
• We have checked what existing data, filings and our own records already answer.
• At least one method observes behavior rather than asking for opinions.
• Interview questions ask about past behavior, not future intentions.
• Survey questions measure something the interviews already uncovered.
• Where possible, people are asked to commit something: time, money or a signature.
• The findings will be reported as a recommendation against the agreed threshold.

Monday morning: take the next decision your team will make about customers and write down the one result that would change it. If nobody can, the team does not need research yet. It needs to agree what it is deciding.

New Coke's researchers asked nearly 200,000 people the wrong question with great care. The craft is not in asking more people. It is in finding the question whose answer customers will prove with what they do.

Questions

What is market research?

Market research is the work of finding out what customers, competitors and markets are actually like so that a decision rests on evidence. It combines existing data, questions to customers and observation of what they do.

What is the difference between primary and secondary research?

Primary research collects new information for your own question, through interviews, surveys, observation or tests. Secondary research uses information that already exists, such as government statistics, company filings, industry reports and your own sales data.

What are the main market research methods?

The main methods are secondary data analysis, customer interviews, surveys, observation and behavioral data, and tests such as pilots, landing pages and price experiments. Each answers a different question, so most good studies combine two or three.

How do you do market research on a small budget?

Start with free secondary sources such as Census business statistics, the BLS Consumer Expenditure Surveys and competitors' SEC filings, then interview ten to fifteen customers about the last time they faced the problem, and test demand with something that asks for a commitment.

What is the difference between market research and market intelligence?

Market research is usually a study commissioned to inform one decision. Market intelligence is the ongoing watch on competitors, customers and the market that runs whether or not a decision is pending.

Where to go next

• TAM, SAM, SOM, Done Honestly
• Market Sizing Calculator: Top-Down vs. Bottom-Up
• Competitor Analysis: A Practitioner's Method for Predicting Moves
• Jobs to Be Done
• Market Segmentation