Marketing research should help you choose an action: reorganising a catalogue, entering a market, reviewing an offering or testing a campaign. Collecting keywords and opinions without a starting question produces information that is hard to use. The objective is to reduce uncertainty around a decision, recognising what the data supports and what remains unconfirmed.
This process combines existing information with research involving people. The following five stages offer a working approach for small businesses, online stores and local businesses. The example of a store considering a category expansion is hypothetical: it does not represent RedeShop client data or results.
1. Define the decision and research question
Replace a vague question such as ‘What do customers want?’ with one that can inform a decision. For example: ‘What information do people need when choosing a size, and why does that uncertainty generate support requests?’ Identify the decision-maker, the audience covered and when the answer is needed.
Separate the observed problem from its presumed explanation. Receiving few quotation requests is an observation; concluding that the price is too high is a hypothesis. The problem may also concern the visitors attracted, clarity of the offering or form submission. Record alternative explanations before selecting questions so the research does not merely confirm the first idea.
- Decision: what action will this research help select?
- Question: what needs understanding to make that decision?
- Audience: existing customers, potential customers or people who abandoned the process?
- Evidence: what information could support or contradict the hypothesis?
- Boundary: which decision is outside this work?
2. Choose sources and prepare a feasible plan
Start with existing records: support questions, return reasons, declined proposals, internal store searches and landing pages. The Small Business Administration distinguishes research using existing sources from direct collection through interviews or questionnaires. These methods complement each other: the former provides context, while the latter helps explain reasons that traffic reports cannot.
In Search Console, queries and pages show how a website appears in Google Search, within the report’s limitations. In Google Trends, interest is relative and normalised; a high index does not represent an absolute number of buyers. Record the country, period, filters and tool. Research on Portugal should not automatically be presented as a description of another market.
Choose a method the team can carry out and analyse. Interviews can explore questions; a questionnaire can compare responses to consistent questions. Define in advance who will be invited, how responses will be stored and who will have access. A small sample can reveal useful themes without necessarily representing all buyers or supporting a market-size estimate.
3. Collect information without leading the answers
Ask about recent, concrete situations. ‘How did you choose the last window you bought?’ can reveal criteria and obstacles; ‘Don’t you think better insulation is important?’ already suggests an answer. Ask for examples and distinguish what happened from what someone imagines they would do. Stated purchase intent is not an order.
In the category example, interview people with different levels of product familiarity. Record which measurements they seek, where they check compatibility and what information they needed to confirm. Include conflicting questions and successful purchases. Interviewing only satisfied customers can leave out the difficulties faced by those who abandoned the process.
4. Analyse patterns and test alternative explanations
Organise responses by theme and retain their source. A simple table can include question, observation, source, date, interpretation and confidence level. Avoid treating an isolated opinion as frequent behaviour. Compare different audiences before combining them: professionals and consumers may need different information.
Cross-check the themes against the website. If measurement questions recur, check whether a guide exists, whether people can find it and whether it matches the products. Research may recommend improving links to existing content rather than creating more pages. A matching pattern does not prove causality; stock changes, campaigns and seasonality can also explain differences between periods.
5. Turn findings into a verifiable next step
Present the recommended decision, supporting evidence and uncertainties. In the example, the proposal could be to revise a sizing guide and place it beside the product selection. Define who implements it, which pages are covered and how operation will be checked. Establish a baseline before claiming improvement and keep a record of changes.
The final document should enable action while making limitations visible: question context, methods, participants, findings, alternatives and validation plan. If the data is insufficient, a useful outcome may be to postpone a decision or run a smaller test. Research does not need to justify an investment already decided; it needs to make the choice better informed.
To connect observed demand to website structure, explore our keyword research service. RedeShop can help define questions and priorities for content, SEO or advertising, with a scope suited to the available data.