MKTG3001 Chap.6 Sampling Methods and Measurement Scales
Sampling Methods and Measurement Scales
Week 6 joins two questions every marketing survey must settle: who responds and how their answers are measured. On the sampling side, you work through the three decisions of a sampling design (target population, sampling procedure, sample size), including inclusion and exclusion criteria and the screening questions that enforce them.
The procedures split into a probability family (simple random, systematic random, stratified random, cluster) and a non-probability family (convenience, purposive, snowball, quota), each with its own trade-off between representativeness and practicality.
On the measurement side, you move from concept to construct to variable, learn the four levels of measurement (nominal, ordinal, interval, ratio), compare Likert, semantic differential and comparative formats, and see why single items, overlapping options, unbalanced scales and flipped directions damage data. The chapter closes on reliability as consistency and validity as relevance plus coverage.
What this chapter covers
- 01
Sampling design and its three decisions
- 02
Target population, inclusion and exclusion criteria, screening questions
- 03
Probability sampling: simple random, systematic, stratified, cluster
- 04
Non-probability sampling: convenience, purposive, snowball, quota
- 05
Factors that shape sample size
- 06
Levels of measurement and scale formats
- 07
Scale design pitfalls, reliability and validity
Worked example · free
Choosing a non-probability branch procedure when random selection is impossible
- 2Without any complete list of eligible managers, random selection cannot happen, so the study belongs on the non-probability branch. That route also suits in-depth interviews, where relevant and detailed insight matters more than representativeness.
- 2Snowball sampling is the best fit: begin with two or three managers known to stock sour beers, then ask each one to refer peers who do the same. Referrals open doors to a small, hard-to-access trade group that cold contact would miss.
- 2Its main weakness is limited diversity, because referrals travel through personal networks and may cluster in one suburb or venue style. Seeding the chain with managers from different areas and venue types reduces that risk.
Key terms
- Target population
- The entire group of people a study wants to learn about, defined through inclusion and exclusion criteria.
- Screening question
- An item placed at the start of a survey that checks eligibility so ineligible respondents leave early.
- Stratified random sampling
- A probability procedure that splits the population into subgroups by a shared characteristic and then selects randomly within each one.
- Quota sampling
- A non-probability procedure that fills a required number of participants for each subgroup without random selection.
- Semantic differential scale
- A rating format with a pair of contrasting adjectives as its two anchors, where respondents mark the point between them that matches their view.
- Internal consistency
- The degree to which several items measuring one construct produce similar response patterns from the same person.
Sampling Methods and Measurement Scales FAQ
What is the difference between stratified and quota sampling?
Both divide the population into subgroups and make sure each subgroup is represented. Stratified sampling then draws members at random within every stratum, so it belongs to the probability family and needs a full list. Quota sampling simply keeps recruiting available people until each subgroup target is met, so the chance of selection is unknown and bias can creep in.
Is a Likert scale ordinal or interval data?
Strictly, Likert responses are ordered categories, and nobody can prove that the step from Disagree to Neutral equals the step from Neutral to Agree, which makes them ordinal. In practice many marketing researchers treat multi-item Likert totals as approximately interval so they can compute means. State whichever assumption you adopt and justify it briefly in an exam answer.
Can you use probability sampling in qualitative research?
Yes. Probability methods are more often paired with surveys, but this is a tendency rather than a rule. A researcher could randomly draw interviewees from a customer list when spread matters. Most interview studies still favour non-probability approaches, since they look for people with specific experiences who can offer rich, detailed accounts.
How do you check if a survey scale is valid?
Review relevance and coverage. Relevance asks whether every question clearly links to the construct; items that drift off topic should be revised or dropped. Coverage asks whether the set of questions spans all important aspects of the construct, adding items where a facet is missing. A scale can be consistent yet still invalid if it measures the wrong thing.
Exam move
Treat this chapter as two linked toolkits and revise each from memory. For sampling, sketch the randomisation fork with four procedures on each side, then practise saying one strength and one weakness for every procedure without notes. Rehearse the decision logic a scenario question can probe: is there a complete population list, must findings generalise, are subgroups to be compared, is the group hard to reach?
For measurement, recite the four levels as a cumulative ladder and attach your own example to each, then name the three scale formats and when each fits. Memorise the four design pitfalls by drafting a deliberately flawed survey item and repairing it.
Finally, write two timed answers of about 220 words, one recommending a sampling procedure for an invented business and one critiquing a scale for reliability and validity, finishing each with a clear recommendation.
Working through Sampling Methods and Measurement Scales in MKTG3001? Sia is AskSia’s AI Marketing tutor — ask any MKTG3001 Sampling Methods and Measurement Scales question and get a clear, step-by-step explanation grounded in how MKTG3001 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.