A stakeholder says, “Sales are down. Can we run a survey?” Before anyone opens a questionnaire tool, there is a more useful question to answer: what will you do differently once you know why?
This article covers the first stage of the market research process: what it is, how to do it well, and a worked example. If you only remember one thing, remember that a study can be flawlessly fielded and still be useless if it answers the wrong question.
Key takeaways
- The first step in the marketing research process is defining the problem and the research objectives before choosing any method.
- Write down three things: the decision the research will inform, the questions that decision depends on, and what you already know.
- Separate the symptom from the problem: “sales are down” is a symptom, and several causes may be behind it.
- If every possible answer leads to the same action, you do not need the study.
What is the first step in the marketing research process?
The first step is to define the problem (or opportunity) and the research objectives. In practice that means writing down three things: the decision the research will inform, the questions that decision depends on, and what you already know. Only then do you choose a method.
The full process has five steps. The first one sets the direction for the other four.
- Define the problem and the research objectives. State the decision, the questions it depends on, and what you will do with each answer.
- Develop the research plan. Choose primary or secondary research, the population and sample, the method, the timeline and the budget.
- Collect the data. Field the survey, run the interviews, or gather existing data.
- Analyze the data. Turn responses into findings: crosstabs, significance tests, segment comparisons.
- Present the findings. Report the result, the evidence behind it and the recommended action.
Some sources split these into six or seven steps, but defining the problem comes first in every version.
Why does defining the problem matter more than any other step?
Every later choice is made against the problem statement. The sample you recruit, the questions you ask and the cut of the data you report all depend on what you are trying to decide. If the statement is vague, each of those choices is a guess.
A vague problem also hides a basic issue: you may not need new research. An earlier study, a tracker or internal sales data may already answer part of the question, and finding that out in week one is cheaper than after fieldwork.
How do you define the problem well?
Work through four moves.
1. Start from the decision, not the topic
“Understand our customers” is a topic. “Decide whether to relaunch the pack or cut promotional depth” is a decision. Ask the stakeholder what action the research will inform and who will take it. If nobody can name an action, the project is not ready.
2. Separate the symptom from the problem
“Sales are down” is a symptom. The problem is whatever is causing it, and several causes may be in play: fewer buyers, less frequent buying, a price gap, a distribution loss. List the plausible causes before you decide which ones research can test.
3. Write objectives as answerable questions
An objective should be a question a respondent or a dataset can answer, such as “What share of last year’s buyers did not buy in the past six months?” Avoid objectives like “explore perceptions,” which cannot be finished or checked.
4. State hypotheses where you have them
A hypothesis is a testable guess about the answer, such as “Lapsed buyers switched to a cheaper competitor.” Writing hypotheses forces you to say what evidence would confirm or reject them. If you cannot form any yet, that points to exploratory work first; the article on deductive and inductive reasoning in market research explains when to start from a hypothesis and when to start from observation.
What does a worked example look like?
The numbers below are illustrative, not from a real client.
A packaged-snack brand sees granola bar sales fall from $50.0 million to $47.0 million year over year. That is a decline of $3.0 million, or 6% ($3.0M divided by $50.0M). The brand manager’s request is: “Find out why sales are down.”
Here is the same request after the first step.
| Element | Vague version | Defined version |
|---|---|---|
| Problem | Sales are down | Sales fell 6%; we do not know whether we lost buyers, lost frequency or lost price position |
| Decision | None stated | Choose one of three actions for next year: relaunch the pack, cut promotional depth, or hold course |
| Objective 1 | Understand why | Measure what share of last year’s buyers did not buy in the past six months |
| Objective 2 | Understand customers | Among lapsed buyers, identify where they went instead |
| Objective 3 | Explore price | Compare price perception against the two nearest competitors |
| Success test | A report | Each objective maps to one of the three actions |
Now the hypotheses, each tied to a decision.
| Hypothesis | Evidence that supports it | Action it points to |
|---|---|---|
| H1: Buyers switched to a competitor | Lapsed buyers name a specific brand they now buy | Relaunch the pack |
| H2: Buyers stayed but buy less often | Buyer share is stable, purchases per buyer fell | Hold course and test frequency drivers |
| H3: The price gap widened | Price perception is worse than competitors among lapsed buyers | Adjust promotional depth |
Notice what happened. A request for “why” became three testable explanations, each linked to a decision. The research plan nearly writes itself: a short survey of past buyers, a purchase-frequency cut, and a price perception question.
A comparison with last year’s results would also help, and the guide to comparing past and present survey results covers how to do that without misleading yourself. If the brand is tracked continuously, you can also see what changed in brand health and against competitors, and why.
How do you know the problem is defined well enough?
Run a quick check before you move to the research plan. If you can answer yes to all five, you are ready.
- Can you name the decision and the person who owns it?
- Can you list the possible answers and what you would do for each?
- Are the objectives questions that data can answer?
- Have you checked what existing research or internal data already says?
- Does the decision owner agree in writing?
The “what would we do for each answer” test is the sharpest. If every possible result leads to the same action, you do not need the study.
What are the common mistakes?
- Starting from the method. “We need a conjoint” or “let’s run a survey” is a solution, chosen before the problem is known. Pick the method after the objectives, not before.
- Defining the problem as a symptom. Falling sales, low awareness and weak engagement are things to explain, not problems to solve.
- Writing objectives nobody can fail. “Gain insight into attitudes” always succeeds, so it guides nothing.
- Letting scope grow. Each stakeholder adds a favorite question, and a 10-minute survey becomes 30. Tie every question to an objective, and cut those that are not.
- Skipping the check for existing data. Many questions are partly answered already, and a quick review can shrink the study. Connecting findings across studies and waves makes that review faster, because teams reuse what they know instead of starting over.
How do you carry the problem definition through the rest of the process?
Keep the one-page problem statement visible. Trace each questionnaire item to an objective, structure the analysis around the objectives and hypotheses, and lead the report with the decision and your recommendation.
For the analysis and reporting steps, tools that produce crosstabs with weighting and significance testing, charts and shareable report pages help. Halo Reports on mTab does that; see Halo Reports for what it includes.
Later steps depend on what the objectives call for. If the decision is about trade-offs between features and price, conjoint analysis may fit; if it is about ranking a long list of messages, MaxDiff may.
Before your next project, write the problem statement on one page: the decision, the owner, three to five objectives and any hypotheses. Share it with the decision owner and ask them to change anything that is wrong. That ten-minute exchange will save far more time than it costs.