BUSI 614 Week 3 Qualitative Assignment 2 Example

Measuring Why Owners Switch: Scales and Questionnaire Design for a Pet Food Survey

Constructs the Survey Must Measure

The depth interviews with Northfield Pet Foods customers and former customers identified several reasons younger owners switch to fresh or subscription dog food: a conviction that fresh meals make dogs healthier, the visible appeal of real ingredients, the convenience of home delivery, a view of the dog as a family member who deserves the best and frustration with the price of fresh services. Northfield's managers must decide whether to launch a refrigerated product, start a subscription program or change their marketing. The survey must therefore measure, for a representative sample of owners aged 25 to 44, current and past brand use, the beliefs and attitudes that drive switching, the importance of convenience and price, and the intention to buy each possible offering.

These constructs call for different levels of measurement. Brand used most often and the type of food fed are nominal categories. Monthly spending on dog food and the number of dogs in the household are ratio measures, best collected as numbers rather than ranges so that averages and totals can be computed. Attitudes and intentions are measured on rating scales that are usually treated as interval data, which allows the means, correlations and regressions the later SPSS assignments use.

Choosing Scales

Most constructs here are attitudes and beliefs, measured on interval-type rating scales. Likert scales, which ask respondents to agree or disagree with statements, suit beliefs such as "fresh food helps dogs live longer," because the statements can be taken directly from interview language. Semantic differential scales, with bipolar adjectives at each end, suit brand image comparisons, for example rating Northfield and a fresh competitor from "processed" to "natural." An itemized rating scale with labeled points suits purchase intention. I would use seven points for attitude items, balanced around a labeled neutral point, because many owners will genuinely have no view on some claims, and forcing a choice would add error.

Whether to measure a construct with several items or one is a key decision. Churchill (1979) set out a procedure for developing measures of marketing constructs that begins by specifying the construct's domain, generates items from the literature and qualitative research, purifies them using coefficient alpha and factor analysis, and then assesses reliability and validity with new data. The procedure argues for multiple items because any single item contains specific error. Bergkvist and Rossiter (2007), however, found that for constructs with a concrete, singular object and attribute, such as attitude toward a particular brand, single-item measures predicted outcomes as well as multi-item scales. The choice between one item and several should follow the nature of the construct, not habit.

For Northfield, perceived health benefits of fresh food is abstract and has several facets, so it calls for a multi-item scale developed along Churchill's lines. Attitude toward the Northfield brand is concrete and can be measured with a single, well-worded item, which shortens the questionnaire.

Sample Items

Every item is worded in the owners' own terms from the interviews, such as "fresh" and "real ingredients," rather than in the company's product language. Purchase intention is asked separately for each offering, a refrigerated product and a subscription plan, so that managers can compare demand for the two options they are weighing instead of reading one general measure of interest.

Order, Wording and Respondent Effort

The questionnaire would use a funnel order: screening questions on dog ownership and age first, then current feeding behavior, then general attitudes, then brand-specific ratings and finally intentions and demographics. Placing brand questions after general attitudes avoids priming respondents to think about Northfield when describing their beliefs. Items would avoid double-barreled wording, such as "fresh food is healthy and affordable," and leading phrases, such as "most experts agree."

Respondent effort also matters. Krosnick (1991) described satisficing, in which respondents reduce the effort of answering by selecting the first acceptable option, agreeing with statements, giving the same rating to every item or choosing "don't know." Satisficing becomes more likely as questions grow harder, as respondents' ability falls and as their motivation fades. The questionnaire should therefore stay under 15 minutes, use plain language, vary the direction of some statements and place the most important questions early. Most younger owners will answer on a phone, so rating scales should display vertically or as compact sliders that fit a small screen, and grids of many items, which encourage giving the same answer down a column, should be broken into short sets of three or four statements.

Guarding Against Method Bias

When predictors and outcomes are measured in the same questionnaire from the same respondents, their correlations may be inflated by the measurement method itself. In a broad review, Podsakoff et al. (2003) traced this common method bias to several sources, including respondents' desire for consistency and the influence of item context, and recommended procedural remedies such as separating the measurement of predictor and outcome variables, protecting respondents' anonymity and varying scale formats. For Northfield, purchase intentions would be measured near the end, after an unrelated block of demographic questions, using a different response format from the attitude items, and the invitation would assure respondents that answers are anonymous and that there are no right answers. A short marker variable, an item theoretically unrelated to the other constructs, such as preference for a particular season of the year, would also be included so that analysts can check afterward whether unrelated items correlate more than expected.

Pretesting

Before fielding, the questionnaire would be pretested with 20 owners from the target population, some of them interviewed while completing it and asked to think aloud, to identify confusing items, missing response options and fatigue. A soft launch to 100 panel respondents would then check completion times, item nonresponse and the reliability of the multi-item scales, allowing final revisions before the full survey.

Once the full data are in, the multi-item scales would be checked for reliability, with coefficient alpha of at least .70 as the usual minimum, and for validity. Items intended to measure perceived health benefits should correlate more strongly with one another than with items measuring the dog-as-family construct, evidence of discriminant validity, and the scale should relate to reported switching in the direction the interviews suggested, evidence of predictive validity. Items that fail these checks would be revised before the survey is repeated, following the purification step in Churchill's procedure.

References

Bergkvist, L., & Rossiter, J. R. (2007). The predictive validity of multiple-item versus single-item measures of the same constructs. Journal of Marketing Research, 44(2), 175-184. https://doi.org/10.1509/jmkr.44.2.175

Churchill, G. A., Jr. (1979). A paradigm for developing better measures of marketing constructs. Journal of Marketing Research, 16(1), 64-73. https://doi.org/10.1177/002224377901600110

Krosnick, J. A. (1991). Response strategies for coping with the cognitive demands of attitude measures in surveys. Applied Cognitive Psychology, 5(3), 213-236. https://doi.org/10.1002/acp.2350050305

Podsakoff, P. M., MacKenzie, S. B., Lee, J.-Y., & Podsakoff, N. P. (2003). Common method biases in behavioral research: A critical review of the literature and recommended remedies. Journal of Applied Psychology, 88(5), 879-903. https://doi.org/10.1037/0021-9010.88.5.879