Week 01 · Marketing Research & Analytics
Davood Wadi, PhD
Marketing research provides evidence that reduces uncertainty in managerial decisions.
Marketing decisions are made under uncertainty.
A manager who launches a product cannot know in advance the level of demand it will attract.
A manager who changes a price cannot know in advance how customers will respond.
Marketing research reduces this uncertainty by providing evidence in place of assumptions.
The value of a study depends on the extent to which its findings change, confirm, or refine a decision.
Marketing research is the systematic and objective identification, collection, analysis, and dissemination of information for the purpose of improving marketing decisions.
Systematic means that the research follows a planned and documented procedure rather than relying on intuition.
Objective means that the researcher reports what the evidence shows, independent of the client's expectations.
The definition is oriented toward decisions: a study whose findings are not used in decision-making has not achieved its purpose, regardless of its methodological rigor.
Three related activities provide evidence to marketing managers.
Marketing research is project-based. It involves collecting new data to answer a specific question, such as why trial users of an application do not convert to paid subscriptions.
Marketing analytics is ongoing. It involves the analysis of data that the firm already generates, such as sales, web traffic, and customer records, in order to monitor and predict performance.
Marketing intelligence is the continuous monitoring of the external environment, including competitor prices, regulatory changes, and industry developments.
Most organizations require all three, as each addresses different questions over different time horizons.
Each case describes a source of evidence for marketing managers. Name the activity.
A project that collects new data to answer one specific question is marketing research.
Ongoing analysis of data the firm already generates, used to predict performance, is marketing analytics.
Continuous monitoring of competitors and regulation in the external environment is marketing intelligence.
New data are collected once, for one decision, so the activity is project-based marketing research.
The data are the firm's own customer records, analyzed continuously: marketing analytics.
Industry developments and regulatory changes are part of the external environment: marketing intelligence.
Which concept does this case show?
Problem-identification research detects problems that are not yet apparent.
Examples include market potential studies, market share tracking, and brand image research.
Problem-solving research informs the choice among alternative courses of action for a known problem.
Examples include pricing research, advertising testing, and product concept tests.
A single project may combine both types; identifying the type required determines the design of the study.
Sort each study by the type of research it represents.
Tap a card, then tap the group it belongs to.
Research requires financial resources and time, and its expected benefit must therefore exceed its cost.
Research is justified when the decision is consequential and the appropriate course of action is uncertain.
Research is not justified when the decision has already been made and the study would serve only to legitimize it.
Research is not justified when its cost exceeds the value of the decision it is intended to inform.
Research is not justified when its findings would become available only after the decision must be made.
The value of information is the expected improvement in the decision that the research makes possible.
Should the firm conduct research? Classify each situation.
The decision is consequential and the right course of action is uncertain, so research is justified.
The decision has been made; the study would serve only to legitimize it.
The study would cost far more than the value of the decision it is meant to inform.
The findings would arrive only after the decision must be made.
A large, uncertain decision: the expected benefit of research is likely to exceed its cost.
The price has already changed; the research would only legitimize a decision already taken.
Which concept does this case show?
Internal research departments, often called insights teams, operate within the firm and have direct knowledge of its data and decisions.
Full-service agencies design and conduct entire projects on behalf of clients.
Syndicated services collect data once and sell it to multiple firms, as in retail sales panels and media audience measurement.
Specialized suppliers provide a single component of the process, such as online panels, field interviewing, or data analysis.
Self-service platforms now enable managers to conduct surveys and experiments independently, which increases speed but does not in itself ensure quality.
Problem definition is the most consequential step in any research project.
A precise answer to an incorrectly defined question produces unwarranted confidence in a flawed decision.
Every subsequent step, from research design to analysis, depends on the definition of the problem.
Decision makers typically present a symptom rather than a problem.
A symptom is an observable change, such as declining sales, increasing complaints, or fewer repeat visits.
A problem is the underlying cause that produces the symptom.
Declining sales at a coffee chain may result from a new competitor, a price increase, slower service, or a change in commuting patterns.
Each cause requires a different decision; the researcher must therefore identify the underlying cause before designing the study.
An online grocery service is losing business. Sort each statement: is it a symptom or a possible underlying problem?
Tap a card, then tap the group it belongs to.
A problem audit is a structured examination of the decision situation, conducted by the researcher together with the decision maker.
It establishes the event that created the need for action and the history of the problem.
It identifies the courses of action under consideration.
It specifies the information required to choose among them.
It determines how, and by whom, the findings will be used.
It records the decision maker's existing beliefs, so that the research tests those beliefs rather than confirms them.
The management decision problem concerns what the decision maker should do. It is action-oriented.
The marketing research problem concerns what information is required and how it can be obtained. It is information-oriented.
“Should the firm introduce a premium tier of its streaming service?” is a management decision problem.
→“What proportion of current subscribers would pay a higher price for premium features, and which features determine that willingness to pay?” is the corresponding marketing research problem.
“Should the retailer redesign its store layout?”
corresponds to“How does the current layout affect the time shoppers spend in the store and the categories they visit?”
The research problem must be broad enough to capture the underlying issue and specific enough to guide the research design.
Match each management decision problem with its corresponding marketing research problem.
Management decision problem
Marketing research problem
Tap a term, then the example it goes with.
The research problem is divided into specific research questions.
Each research question specifies one element of the information the decision requires.
A hypothesis is a testable statement that proposes an answer to a research question.
“Subscribers under 30 are more willing to pay for premium features than subscribers over 30” is a hypothesis.
Hypotheses increase the precision of a study, because the data can either support or contradict them.
Discussion: A fitness club reports that new members now cancel within three months at twice the rate of the previous year. The general manager requests a survey of member satisfaction. Which other causes could account for this symptom, and how should the management decision problem and the marketing research problem be formulated before the survey is commissioned?
Rigorous research follows a structured sequence of steps.
The steps are presented in sequence, although in practice researchers return to earlier steps when new evidence changes the question.
Step 1 is defining the problem, through the problem audit and discussions with decision makers.
Step 2 is developing an approach to the problem, including the theoretical framework, research questions, and hypotheses.
Step 3 is formulating the research design, which specifies the type of study, the measures, and the sample.
Step 4 is collecting the data, in the field, online, or from existing sources.
Step 5 is preparing and analyzing the data.
Step 6 is preparing and presenting the report, with findings and recommendations linked to the decision.
A juice brand wants to understand weak sales of its new line. Put the project's activities in the order of the research process.
Tap a step, then the number where it belongs.
Artificial intelligence, and large language models in particular, is now used at every step of the research process.
AI increases the speed at which research can be conducted. It does not alter the criteria by which the validity of research is judged.
The researcher's task is therefore to determine at which steps AI is appropriate and how its output is to be verified.
In problem definition, AI can be used to summarize background documents and to propose alternative explanations for a symptom.
In research design, AI can be used to draft survey questions, interview guides, and screening criteria.
In data collection, AI can be used to moderate interviews at scale and to translate research instruments into other languages.
In analysis, AI can be used to code open-ended responses, to write analysis code, and to summarize large volumes of text.
In reporting, AI can be used to draft summaries and visualizations for different audiences.
Match each use of AI with the step of the research process in which it occurs.
Step
Use of AI
Tap a term, then the example it goes with.
AI systems can generate facts, statistics, and references that appear credible but are false.
AI systems can reproduce biases present in their training data.
AI output tends to conform to the framing of the prompt, so that a leading prompt yields a confirmatory answer.
Responses from AI-simulated consumers are not observations of real consumers. They may appear plausible while differing systematically from the responses of real people.
AI output is expressed with the same apparent confidence whether it is accurate or not, so its fluency provides no evidence of its accuracy.
Each case shows a limitation of AI in marketing research. Name the limitation.
The reference appears credible but is false.
The model reproduces a bias present in its training data.
A leading prompt yields a confirmatory answer.
The simulated responses appeared plausible but differed systematically from those of real people.
The output has the same apparent confidence whether it is accurate or not.
Which concept does this case show?
The researcher, not the tool, is responsible for every finding in the report.
Every AI output that informs a decision must be verified against a source, a sample of human judgments, or empirical data.
Researchers should document where and how AI was used, so that others can evaluate and replicate the work.
The use of AI shifts the researcher's effort from producing the work to verifying it.
An AI agent is a system built on a large language model that can search for information, compare alternatives, and complete transactions on behalf of a person.
A consumer who delegates a purchase to an AI agent makes the agent a surrogate consumer: the consumer states the goal, and the agent acquires product information and selects an alternative.
In agent-mediated purchasing, the agent rather than the consumer encounters the product page, the price presentation, and the sponsorship label.
Research questions therefore extend from consumers to their agents: which information agents acquire, how they weight product attributes, and whose interests their recommendations serve.
New management decision problems follow, such as how to present product information to AI agents, and whether an agent is suitable for deployment to the firm's customers.
The evaluation of an AI model or agent is a research problem: the researcher defines the behavior of interest, measures it under controlled conditions, and estimates it with known precision.
Standard AI benchmarks report accuracy on large sets of general questions, such as mathematics or programming. They do not establish how a model behaves in a specific marketing task.
A behavioral evaluation applies the methods of this course to the AI system: validated measures (Week 4), adequate samples of responses (Week 5), hypothesis tests (Week 7), and controlled experiments (Week 8).
The same model can produce different responses to the same prompt, so that a single response is not a reliable basis for conclusions about its behavior (Wadi & Fredette, 2025).
The researcher who evaluates an AI system is accountable for the evidence in the same way as for research on consumers.
A retailer plans to deploy an AI shopping assistant on its website. The vendor reports that the underlying model scores 91 percent on a widely used benchmark of general knowledge and reasoning. A manager also asked the assistant once which of two coffees offered better value, and the answer was correct. Which conclusion does this evidence support?
Marketing research involves four stakeholders: the researcher, the client, the respondent, and the public.
Each stakeholder holds rights and bears obligations.
Ethical violations reduce public trust in research, on which all researchers depend.
Respondents give informed consent: they are told the purpose of the study and how their data will be used, and that they may withdraw at any time.
Respondents' privacy is protected: personal data are kept confidential and reported only in aggregate or anonymized form.
Respondents are not harmed, deceived without justification, or pressured to participate.
Research is never used as a pretext for selling. Soliciting sales under the guise of research is known as sugging and constitutes an ethical violation.
Uploading respondent data to an external AI tool constitutes a use of those data and must therefore fall within the scope of the consent that respondents gave.
Sort each research practice: does it meet the obligations to respondents, or violate them?
Tap a card, then tap the group it belongs to.
Researchers report findings accurately, including findings that are unfavorable to the client.
Researchers disclose the limitations of the method, the sample, and the tools used, including AI.
Clients do not misrepresent findings to the public, for example by citing results out of context in advertising.
Clients do not commission studies designed to reach a predetermined conclusion.
Discussion: A marketing manager proposes to replace a planned focus group study with a set of AI-generated personas that respond to the same questions within minutes. What would the firm gain, what might it lose, and under what conditions, if any, would the substitution be methodologically defensible?
Marketing research provides evidence for decisions made under uncertainty, and it is justified only when its findings can inform a decision in time.
Research, analytics, and intelligence are complementary sources of evidence.
The most consequential step is the translation of a symptom and a management decision problem into a precise marketing research problem.
The research process comprises six steps, from problem definition to reporting, to which researchers often return iteratively.
AI can accelerate every step of the process, but its output must be verified, and the researcher remains accountable for every finding.
Ethical research protects respondents, reports accurately to clients, and respects the public interest.
AI agents that acquire information and select products on consumers' behalf are a new subject of marketing research, and the evaluation of AI systems applies the methods of measurement, sampling, and experimentation developed in this course.