Closure Trial in a Wine Bar: Consumer Preference and Alcohol
A candidate wants to know whether closure type affects consumer perception of reduction in Sauvignon Blanc. Over two evenings in a London wine bar, 120 patrons are recruited opportunistically as they order. Each participant receives three pairs of wines: one screwcap and one natural cork, matched by wine and vintage. Samples are poured 50 ml, presented in plain glasses, but the order within each pair is fixed (screwcap first) and the server knows the closure. Participants do not spit, and no water or bread is provided. The questionnaire asks them to rate liking and to say whether each wine is 'reductive', a term left undefined. The candidate has no laboratory access and a £1,500 budget. The research question is: 'Do consumers detect closure-related differences in reduction?' The candidate hopes to publish in a trade journal.
1. Critique the sampling and recruitment strategy for answering the research question. What can and cannot be inferred from 120 opportunistic wine-bar patrons?
Read the worked answer
The sample is a convenience sample of people already drinking in one wine bar on two evenings. It is not a probability sample of consumers, and the 120 figure is not a sample size in the inferential sense until repeated measures and clustering by session are accounted for. Patrons self-select into the venue, may be influenced by price, social context and alcohol, and are likely more engaged with wine than the general population. The fixed order (screwcap first) confounds closure with presentation order, and the unblinded server adds expectancy. The absence of spitting and water means fatigue and alcohol accumulate across six samples. Therefore the study can describe the responses of these patrons under these conditions; it cannot estimate a population preference or detection rate, and it cannot separate closure effects from order, setting and alcohol. A minimal fix is randomised balanced order, blind coding, small pours with expectoration, water and rest, and recording session as a cluster. Even then, the claim should be limited to 'in this venue, among these volunteers'.
2. Propose a minimal set of controls and design changes that could be implemented in the wine bar without laboratory facilities, and explain what each controls.
Read the worked answer
Use a within-participant design with each participant receiving both closures in each pair, but randomise the order within pair and across participants with a balanced Latin-square schedule; this controls order and carry-over. Code all glasses with three-digit random numbers and use a blind server; this controls expectancy and server bias. Use 20–25 ml pours, require expectoration, provide water and plain crackers, and impose a 60–90 second interval between samples; this limits fatigue and alcohol. Add a duplicate pair at the end to estimate within-person consistency. Use a forced-choice task, such as 'which sample shows more struck-match character?', rather than an undefined 'reductive' question; this improves construct validity. Collect basic demographics and wine involvement, not to weight the sample but to describe it. Check local licensing and ethics: no one under 18, no one driving, informed consent, and a maximum total alcohol intake. Record session as a random effect in analysis.
3. The questionnaire asks whether the wine is 'reductive' without a definition. Discuss the measurement problem and how to fix it.
Read the worked answer
'Reductive' is a technical term with multiple meanings - struck match, cabbage, sulfide, mercaptan - and may be unfamiliar to consumers. Asking for a yes/no judgement without a definition measures the respondent's interpretation of the word, not the wine's sensory state. This is a construct-validity problem: the item may correlate with wine knowledge, confidence or social desirability rather than with perception. Fix it by defining the term in plain language on the questionnaire, providing reference standards or example descriptors, using a forced-choice or intensity scale for specific attributes rather than a single technical label, and piloting the wording with 10–15 people from the target population. If the term must be retained for trade respondents, report the definition and treat the item as one indicator among several. Do not interpret a significant closure difference on this item as evidence that consumers detect reduction unless the measure is validated.