MKTG90011 · Marketing Research
Research Designs
Three research designs map onto three levels of problem clarity, and you pick the design from what you still don't know. An ambiguous problem needs exploratory research — loosely-structured, qualitative, hypothesis-generating (interviews, focus groups, observation). A somewhat-defined problem needs descriptive research — structured surveys on large samples that describe who, what and how often (cross-sectional or longitudinal/panel). A clearly-defined cause-and-effect question needs causal research — an experiment that manipulates the independent variable while controlling everything else. Causality is the hard one: establishing that X causes Y needs three conditions — covariation (they move together), time order (the cause precedes the effect), and no plausible alternative explanation (achieved by randomisation and control). Correlation alone never proves causation. In MKTG90011 the project's Qualtrics survey is a descriptive design that feeds the quantitative tests, so this week wires the methods canon directly to your H1–H6.
What this chapter covers
- 015.1 The design taxonomy: exploratory, descriptive, causal
- 02Matching problem clarity to design
- 03Descriptive designs: cross-sectional vs longitudinal / panel
- 04Causal designs: experiments, manipulation and control
- 05The three conditions for causality
- 06Why correlation alone fails to prove cause
Worked example: match the problem to a design
- +1(a) Ambiguous → exploratory. “No idea why” signals a fuzzy problem, so use qualitative, hypothesis-generating work (interviews/focus groups) to clarify it.
- +1(b) Somewhat defined → descriptive. “What proportion… how does it vary” is a who/what/how-often question, answered by a structured survey on a large sample.
- +1(c) Clearly defined cause-effect → causal. “Does the discount increase repurchase” is an X→Y claim, so run an experiment: manipulate the discount, control other factors, randomise.
- +1State the causality conditions. To claim the discount causes the lift you need covariation, time order (discount before repurchase) and no alternative explanation — which is why a controlled, randomised experiment, not a correlation, is required.
Key terms
- Exploratory design
- A loosely-structured, qualitative approach used when the problem is ambiguous — interviews, focus groups, observation — to build understanding and generate hypotheses, not to confirm them.
- Descriptive design
- A structured approach (typically surveys on large samples) that describes a market — who, what, how often — without testing cause. It is cross-sectional (one snapshot) or longitudinal/panel (repeated over time).
- Causal design
- An experiment that tests whether one variable causes another by manipulating the independent variable and controlling other factors; requires randomisation and control to rule out alternative explanations.
- Cross-sectional vs longitudinal
- Cross-sectional surveys measure a sample once (a snapshot); longitudinal/panel designs measure the same units repeatedly over time, so they can track change but cost more and risk panel attrition.
- Conditions for causality
- The three requirements to claim X causes Y: covariation (they move together), time order (cause precedes effect), and elimination of alternative explanations (via control and randomisation).
Research Designs FAQ
How do I choose between the three designs?
By how clearly the problem is defined. Ambiguous → exploratory (clarify it qualitatively); somewhat defined → descriptive (survey to describe it); clearly defined cause-effect → causal (experiment to test it). Choose the design from what you still don't know, not from what data is easiest to get.
Why isn't a correlation enough to prove causation?
Because covariation is only one of the three conditions for causality. You also need correct time order (the cause must precede the effect) and the elimination of alternative explanations. A correlation can arise from reverse causation or a lurking third variable, so only a controlled, randomised experiment supports a causal claim.
What is the difference between cross-sectional and longitudinal research?
Cross-sectional research measures a sample once — a single snapshot, cheaper and faster. Longitudinal (panel) research measures the same units repeatedly over time, so it can detect change and trends, but it costs more and suffers from attrition as panel members drop out.
Is the project's survey exploratory, descriptive or causal?
Descriptive. The Qualtrics survey describes the sample and supplies the metric and categorical variables that feed the quantitative hypothesis tests (H1–H6). The earlier qualitative stage is exploratory; the survey itself is a descriptive design.
Exam move
Practise the “match the problem to a design” item until it's reflex — it recurs every year — by reading the trigger words (why/explore, describe/what-proportion, cause/increase). Memorise the three conditions for causality as a unit (covariation, time order, no alternative explanation) so you can earn the causal mark and explain why correlation alone fails. Keep the cross-sectional vs longitudinal distinction handy, and remember that your project's survey is a descriptive design feeding the tests — that link is examinable and grounds your method section.
