Steve is interested in doing research on how customer perceptions of various service factors influence customer satisfaction. Steve thinks that customers that shop in discount stores are more satisfied then those that shop in full-line stores. Also, he believes that satisfaction decreases with age. When it comes to customer satisfaction, Steve believes that it is composed of 3 parts - satisfaction with sales help, satisfaction with products and satisfaction with store environment (e.g. ease of parking, cleanness of store, etc.). Steve has access to customers that shop for a major retailer with hundreds of locations, some discount and some full-line.
a. Articulate two research hypotheses.
b. Develop a portion of a survey that measures the three variables (customer satisfaction, age and store type (discount/full-line).
c. Comment on how Steve could go about collecting an adequate and representative sample.
d. What statistical technique would Steve use to test the two research hypotheses?
Solution: (a) Steve wants to test the following hypothesis:
In order to do this, he needs to analyze the correlation coefficient of the variables and check if it's significantly different from zero (negative), or perform a regression analysis an to analyze the sign and significance of the regression coefficients.
Another research hypothesis that Steve wants to check is
(b) The best way to develop the survey is to design a consumer satisfaction evaluation method. The alternatives are basically a quantitative score and ordinal categories like "Dissatisfied", "Satisfied", "Very Satisfied", etc. The rest of the variables are standard, so they should be included as simple categorical question "discount or full time" (for store type) and a number for Age.
(c) Steve should define first his population. Is the population all the clients they have now? Or the population should include current and prospective clients? Based on the factor, he should decide what sampling technique is best for the situation. Is the store small and it has only one branch? In that case he should for a simple random sampling. Is the store big enough to have more complex structures in his clientele? Then he probably should use stratified sampling. Or does the store have too many branches geographically spread out? Then cluster analysis could be the solution.
(d) If could use correlation for quantitative variables, Spearman correlation for ordinal data, and linear regression with dummy variables in general.