Measurement Case Study 3: Theory-testing research: testing a deterministic relation

The measurement process followed the same procedure for all cases, and consisted of four steps: ■ open interview with the distribution or logistics manager to get familiar with each retailer’s operations and urban freight transport activities and the current or likely retailer’s reaction on time window pressures; ■ a questionnaire to collect detailed data on each retailer’s oper- ational level; ■ company documents and additional information with infor- mation on each retailer’s entire transport planning for one week; ■ e-mail andor telephone contact for additional information needed. Collected data were put into a mathematical model that generated the distribution costs in all four dimensions, for a given time access win- dow pressure. In this model we needed to solve a number of vehicle routing problems with time windows. The number of extra vehicles was kept to a minimum. To plan the new roundtrips we used the vehicle routing software SHORTREC 7.0, developed by Ortec Consultants. From the new calculated retailers’ roundtrip planning, we derived the values for the dimensions of distribution costs. For a detailed descrip- tion of the collection of actual retailers’ distribution data as well as of the model we refer to Quak and De Koster, 2007.

6.2.9 Data presentation

We filled all 18 cells of Table 6.2 for each retailer and for each of the four dimensions of distribution costs, resulting in 56 4 ⫻ 14 tables. The tables can also be represented in graphs, as is shown in Figure 6.2 for one of the 14 retailers case 8. The two time window pressure dimen- sions are represented in Figure 6.2 as follows: the x-axis represents the number of time window restricted areas resulting from each scenario for this retailer. The different values of time window length are represented by a line for each scenario A, B, and C.

6.2.10 Data analysis

Hypothesis 1 states that for each of the 14 retailers and for each of the four dimensions of distribution costs the six values 1–6 in each of the three rows A, B, or C in this table are in a perfect order of increasing costs. Hypothesis 2 states that for each of the 14 retailers and for each of the four dimensions of distribution costs the three values A, B, or C in each of the six columns 1–6 in this table are in a perfect order of increasing costs. We tested both hypotheses in each of the 56 tables by looking at the actual numbers, and use Figure 6.2 here only as means of presentation. It shows that the value of all four dimensions of cost, increased with the number of shops affected by a time access window, an effect that is clearly visible as a rise in each of the lines if one goes from left small number of restricted areas to right high number of areas. In each graph the line for scenario C is consistently higher than the line for scenario B, which is consistently higher than the one for A, which rep- resents the fact that the value of all four dimensions of cost increased with the decrease of length of the time access windows. Because no instance was found in which, for a given value of number of restricted areas, the value of a dimension of cost was higher for scen- ario A than for scenario B or C, and because the value for B never exceeded the one for C, and because no instance was found in which a Number of roundtrips Scenario A Scenario B Scenario C 20 40 60 80 100 120 20 40 60 80 100 Stores affected 20 40 60 80 100 Stores affected 20 40 60 80 100 Stores affected 20 40 60 80 100 Stores affected Percentage increase 20 40 60 80 100 120 Percentage increase 20 40 60 80 100 120 Percentage increase 20 40 60 80 100 120 Percentage increase Number of vehicle kilometres Scenario A Scenario B Scenario C Total time used Scenario A Scenario B Scenario C Number of vehicles used Scenario A Scenario B Scenario C Figure 6.2 Distribution costs as an effect of time window pressure Case 8