Theory Methodological reflection on Case Study 3

6.3.2 Research objective

The study’s objective was to test a new proposition about the relation between time access window pressure and distribution costs for retailers.

6.3.3 Research strategy

The proposition specifies a deterministic relation between time window pressure and distribution costs. A deterministic proposition can best be tested in a serial experiment in which it is demonstrated in a single experiment that each manipulated change of the independent vari- able results in a corresponding change in the value of the dependent variable. Such an experiment is not feasible for this study, because it would require that local governments would vary time window pres- sure for the purpose of this research which is not possible in practice. If an experimental research strategy is not feasible, as in this study, a longitudinal single case study is commendable. Case Study 3 discusses this possibility and concludes that it is not feasible because it is too difficult to collect accurate historical data on both time window pressure and distribution costs. However, this problem is solved by formulating real- istic scenarios with different levels of window pressure and by using the retailers’ current distribution data to calculate the distribution costs for each of these scenarios. In this way, Case Study 3 imitates a longitu- dinal single case study. Furthermore, the study was designed as a parallel longitudinal case study. As discussed above in 5.3.3, an advantage of the parallel case study approach is that the chance of finding a rejec- tion of the proposition in one round of parallel testing is higher than with a test in a single case. A disadvantage of this approach is that more tests will be performed than are necessary. Also, with this large number of parallel cases, the danger that a probabilistic approach will unwittingly creep into the analysis is present.

6.3.4 Candidate cases

The universe of instances of the object of study to which the theory is applicable consists of all distribution activities of all large retailers in Western Europe with shops that are at least partly located in shopping areas in cities in which time access windows could be installed. It is correctly stated in 6.2.2.4 that Dutch retailers are instances from this domain.

6.3.5 Case selection

Fourteen Dutch retailers were selected from the set of candidate cases. Retailers with different competitive strategies were selected: discounters lower end of the market, retailers that focus on cost middle segment of the market, and retailers that focus on response or differentiation higher end of the market. This attempt to select a “representative” sample of cases also shows that the study was designed as a parallel case study. This representativeness was not needed for this study.

6.3.6 Hypotheses

Because the proposition in this study specified a deterministic relation, and implicitly a continuously increasing relation, the hypothesis stated that the rank order of measurement points according to the observed values of the dependent variable was exactly the same as the rank order of meas- urement points according to the observed values of the independent vari- able. Because the independent variable time access window pressure had two independent dimensions, number of time window restricted areas and time window length, two hypotheses needed to be formulated.

6.3.7 Measurement

In order to generate realistic roundtrip data for all retailers under the conditions defined by the 18 scenarios, actual current roundtrip data needed to be collected. It is described how multiple data sources were used, namely interview data face-to-face, e-mail, telephone, question- naire data, and documents. The quality of the collected data cannot be evaluated because no further detail is provided. The four dimensions of distribution costs for all 14 cases in all conditions defined by the 18 scenarios were generated by a mathematical model. A reference is pro- vided to another article in which the model is described and discussed in more detail.

6.3.8 Data presentation

Figure 6.2 presents the data generated for the different scenarios in one case, as an example. The mode of presentation is such that it directly provides the rank orders needed for testing the hypothesis.