Unit 2: Research design and observation
Marketing Research notes · PTU syllabus (MBA 302-18)
On this page
- Unit summary
- Causality and experimentation
- Extraneous variables and sources of error
- After-only and before–after designs
- Time series designs
- Statistical designs: randomised block, Latin square and factorial
- Ex post facto research
- Test marketing
- Direct and indirect observation research
- Developing a research proposal
- Key terms
- Quick revision
- Important questions
Unit summary
Causal designs test whether a marketing action actually changes outcomes, while observation records what people really do. This unit covers causal research designs — after-only, before–after, time series, Latin square, factorial, ex post facto and randomised block — direct and indirect observation research, and developing a research proposal.
After this unit you can
- Explain validity and errors in experiments
- Explain pre-experimental, quasi-experimental and statistical designs
- Explain direct and indirect observation methods
- Develop a research proposal
PTU syllabus topics
- Causal research designs (after-only, before-after, time series, Latin square, factorial, ex-post facto, randomized block)
- direct and indirect observation research
- developing a research proposal
Topic 1
Causality and experimentation
An experiment manipulates one or more independent variables (the cause), controls other variables and measures the effect on a dependent variable.
- Conditions for causality: concomitant variation (X and Y vary together), time order (X occurs before Y), elimination of other possible causes.
- Experiment terms: independent variable (treatment, e.g., price), dependent variable (sales), test units, control group, extraneous variables.
- Internal validity: was the change really caused by the treatment? External validity: can results be generalised?
- Symbols: X = treatment, O = observation, R = random assignment, EG = experimental group, CG = control group.
Example
A shelf-display experiment: a biscuit brand uses an eye-level display (treatment) in 20 stores and a normal display in 20 similar stores (control) for four weeks. Sales rise 15% in treatment stores and 2% in control stores, so the display likely caused about a 13% increase.
Exam tip
Learn the three conditions of causality: concomitant variation, time order and elimination of other causes.
Topic 2
Extraneous variables and sources of error
Extraneous variables are variables other than the treatment that affect the result and threaten internal validity. Others include testing, instrumentation, statistical regression, selection bias and mortality (drop-outs).
- History: outside events during the experiment.
- Maturation: changes in subjects over time (fatigue, learning).
- Testing effect: the first test influences the second.
- Instrumentation: changes in measuring tools or observers.
- Selection bias: groups that differ from the start.
- Mortality (attrition): subjects dropping out.
- Statistical regression: extreme scores moving toward the average.
Using a control group and random assignment reduces these errors.
Example
History spoils a test: during a two-week price test for a cold drink, an unexpected heatwave raises sales in all shops. Without a control group, the rise would be wrongly credited to the new price.
Exam tip
Control groups and random assignment are the main ways to control extraneous variables.
Topic 3
After-only and before–after designs
After-only (one-shot case study)
EG: X O1
No pre-measure or control — cannot rule out other causes
One-group before–after
EG: O1 X O2
History, maturation and testing effects possible
Before–after with control group
EG: R O1 X O2; CG: R O3 O4
Treatment effect = (O2 − O1) − (O4 − O3); controls most threats
After-only with control group
EG: R X O1; CG: R O2
Effect = O1 − O2; avoids testing effect
Example
Sales in test stores rise from 500 to 620 units after a display (O2 − O1 = 120) while control stores rise from 480 to 520 (40). Display effect = 120 − 40 = 80 units.
Topic 4
Time series designs
Time series designs are useful when a proper control group is not possible, such as when an advertising campaign runs across the whole market.
- Time series (quasi-experiment): O1 O2 O3 X O4 O5 O6 — repeated measures before and after the treatment on the same group (e.g., weekly sales before and after an ad campaign).
- Multiple time series: adds a control group measured at the same points — stronger evidence.
- Use: when random assignment is not possible; consumer panels make it practical.
Example
Tracking an ad campaign: a brand records its monthly market share for six months (O1–O6), launches a TV campaign (X), and keeps measuring for six months more. A clear jump that stays after the campaign suggests an effect, though history (a competitor's stock-out) must be ruled out.
Exam tip
Write the notation O1 O2 O3 X O4 O5 O6 and say it is a quasi-experimental design.
Topic 5
Statistical designs: randomised block, Latin square and factorial
Statistical designs allow the researcher to measure the effect of more than one variable, or to control specific extraneous variables, in a single experiment.
Completely randomised design
Treatments assigned randomly to test units
Randomised block design
Units grouped into blocks on one extraneous variable (store size), treatments randomised within blocks
Latin square design
Controls two extraneous variables (store and week) with a square arrangement — each treatment once in each row and column
Factorial design
Tests two or more independent variables and their interaction (price × advertising)
- Latin square example: three price levels tested across three stores over three weeks so that each price appears once in each store and each week.
- Factorial example: 2 × 3 design — two package designs × three price levels = six treatment combinations; reveals whether the best price depends on the package.
- Analysed with ANOVA.
Example
A factorial design: a juice brand tests two prices (₹25, ₹30) × two pack designs (old, new) = four combinations, each in different stores. The design shows the effect of price, of pack, and their interaction — for example, the new pack works only at ₹25.
Exam tip
Randomised block controls one extraneous variable, Latin square controls two, and factorial measures interactions.
Topic 6
Ex post facto research
- Ex post facto ("after the fact") research studies an effect that has already occurred and looks back for causes; the researcher cannot manipulate the independent variable.
- Example: comparing heavy and light users of a brand to see whether exposure to a past campaign differed.
- Limitation: cannot establish causality firmly — self-selection and other variables may explain differences.
Example
Looking back for a cause: a researcher finds that customers who joined a bank's loyalty programme have higher deposits. Since customers chose to join, the programme may not be the cause — wealthier customers may join more often. Ex post facto designs show association, not proof of causation.
Exam tip
Its main weakness: the researcher cannot manipulate the independent variable or randomly assign subjects.
Topic 7
Test marketing
Test marketing is a controlled experiment in a limited but carefully chosen part of the market, used to predict sales or profit of a new product or marketing plan before a national launch. Simulated test markets use laboratory settings and models.
- Standard test market: launch in selected cities through normal channels. Controlled test market: research firm handles distribution in selected stores. Simulated test market: lab store and purchase-intention models.
- Uses: forecast sales, test the marketing mix; risks: cost, time, competitor interference, revealing plans.
Example
Testing before a national launch: an FMCG company launches a new shampoo only in two test cities chosen to represent its target market. Strong trial but low repeat purchase reveals that buyers dislike the fragrance, which is changed before the national launch.
Exam tip
Mention the drawbacks: cost, time, competitors watching or disturbing the test, and the risk of copying.
Topic 8
Direct and indirect observation research
Direct
Observing behaviour as it happens — shoppers in a store, mystery shopping
Indirect
Observing records or traces of past behaviour — pantry audits, garbology, website logs
Structured or unstructured
Pre-specified checklist vs open recording
Disguised or undisguised
Respondents unaware or aware of being observed
Human or mechanical
Observers vs cameras, eye-tracking, scanners, people meters, web analytics
- Merits: records actual behaviour, no respondent bias, useful with children or non-verbal behaviour. Limits: cannot observe motives or attitudes, costly, ethical concerns about privacy.
Example
A retailer's eye-tracking study shows shoppers rarely look below knee level, so high-margin products are moved to eye-level shelves.
Topic 9
Developing a research proposal
A research proposal is the formal plan of the project, agreed between the researcher and the client before work starts. It ensures both agree on the problem, method, time and cost.
- 1
Executive summary
- 2
Background and problem definition
- 3
Research objectives and hypotheses
- 4
Research design and methodology
Design, sampling, data collection, instruments
- 5
Data analysis plan
- 6
Time schedule
Gantt chart
- 7
Budget
- 8
Report format and deliverables
- 9
Appendices
- Purpose: agreement between client and researcher on scope, method, cost and time; a basis for evaluating the project.
Example
A short proposal outline: for a study on online grocery adoption in Ludhiana — background (growth of quick-commerce), objectives (adoption drivers and barriers), design (survey of 300 households), sampling (stratified by locality), analysis (factor and regression analysis), timeline (8 weeks) and budget.
Exam tip
List the contents in order; examiners often ask this as a 5-mark question.
Key terms
- Internal validity
- Confidence that the treatment caused the observed effect
- Control group
- Group not exposed to the treatment
- Latin square design
- Design controlling two extraneous variables
- Factorial design
- Design testing two or more variables and their interaction
- Ex post facto research
- Study of causes after the effect has occurred
Quick revision
- Causality: concomitant variation, time order, eliminating other causes.
- Errors: history, maturation, testing, instrumentation, selection, mortality, regression.
- After-only, before–after, with control groups; time series designs.
- Randomised block, Latin square, factorial; ex post facto; test marketing.
- Observation: direct, indirect, structured, disguised, mechanical; research proposal contents.
Important exam questions
Practice questions written to the PTU exam pattern for this unit's syllabus: short answers (Section A style) and long answers (Sections B and C style).
Short-answer questions
- Q1.State the conditions for causality.
- Q2.Distinguish internal and external validity.
- Q3.Write the notation for a before–after design with control group.
- Q4.What is a Latin square design?
- Q5.What is ex post facto research?
- Q6.Distinguish direct and indirect observation.
Long-answer questions
- Q1.Explain experimental designs used in causal research with notation.
- Q2.Explain randomised block, Latin square and factorial designs with marketing examples.
- Q3.Discuss observation methods in marketing research.
- Q4.Explain the contents of a marketing research proposal.
Stuck on this unit?
Message SBS on WhatsApp for help with Marketing Research, or to ask about studying MBA at Synetic.
