Hypotheses Measurement Case Study 4: Theory-testing research: testing a probabilistic relation

used including the loading and unloading times as well as driving and waiting time, the number and types of vehicles used, and the number of roundtrips, into a monetary value. The variable costs are mainly based on costs per hour and cost per kilometre. We validated the costs with all retailers, and adapted them slightly in case the retailers felt this would give a better representation of the actual costs. The costs for overtime are higher per hour than in the normal situation. The five dimensions of distribution strategy were determined as follows. ■ Stops per roundtrip. This was measured by calculating the aver- age number of stops per vehicle roundtrip during a week. This equals the average number of different shop deliveries that are combined in one vehicle. This can vary from full- truckload FTL deliveries, in which a vehicle only makes one stop per roundtrip, to less-than-truckload LTL deliveries, implying that a vehicle makes more than one delivery per roundtrip Stock and Lambert, 2001. ■ Vehicle capacity. Based on McKinnon et al. 2003 we distin- guished six different vehicle types. We sorted these types on increasing load factor, starting with the smallest capacity and ending with the largest vehicle capacity see Table 7.1. We calculated each retailer’s average vehicle fleet capacity based on the number of vehicles in each category. ■ Stopping time per vehicle. Stopping time can be split into two parts: a fixed stopping time per stop and a variable stopping time per Table 7.1 Vehicle types sorted on capacity Value Type Characteristics Example 1 Small rigid 2 axles, ⬍7.5 tons 2 Medium rigid 2 axles, ⬎7.5 and ⬍18 tons 3 Large rigid ⭓2 axles, ⬎18 tons 4 City semi-trailer articulated, 3 axles 5 Articulated vehicle articulated, ⬎3 axles 6 Drawbar combination combination, ⬎3 axles stop. The fixed time is the result of activities, such as looking for a place to park, parking the vehicle, and notifying a shop’s staff of the driver’s presence. Variable stopping time depends on the amount of product carriers that has to be unloaded. For example, one retailer case 13 uses detachable swap bodies, which can be unloaded in the absence of the vehicle. The vehi- cle brings a full container and picks up the empty container of the previous delivery in less than 10 minutes. Another way to speed up the unloading process at the shops is to have shop staff available to assist the driver during the unloading of the vehicles. ■ Distance between shops and distribution centre DC. This dimen- sion is measured by the average distance in kilometres between the retailer’s shops and its distribution centres. ■ Self-imposed time windows. We use a three-point ordinal scale for this dimension. Self-imposed time windows are strict, normal, or long. A reason for a retailer to impose a strict time window is, for example, to separate the shopping public from the sup- plying activities, so the shelves are full before the shop opens. Other retailers want to make sure a vehicle leaves a govern- mental time window restricted area, before the time window ends. Usually this is the case if the area is physically closed, by for example rising posts, outside the time window period. If the vehicle does not leave the area in time it results in a fine and a delay. Retailers that have normal self-imposed time win- dows supply their shops during the hours that staff is available to receive the goods. The retailers that use long self-implied time windows even deliver their stores at times when no staff is present in the shops. In these cases the driver possesses a key to the shop or the shop’s depot to deliver the goods.

7.2.9 Data presentation

Table 7.2 shows the values for the initial distribution strategy dimen- sions for all 14 cases. In Figure 7.1 we depict, for each of the five dimensions of distribu- tion strategy, how the increase of this dimension increases the average total distribution costs. The graphs in Figure 7.1 differ from those in the previous chapter in the following way. The lines represent the different values on distribu- tion strategy, not time window length.

7.2.10 Data analysis

Our hypotheses state that it is likely that a retailer’s sensitivity to time window pressure in cases with comparatively high values on distribution strategy dimensions is higher than in cases with comparatively low val- ues on these dimensions. We tested them by classifying the cases into four groups for each dimension of distribution strategy separately: number of stops, vehicle capacity, stopping time, and distance to shops. We used the following procedure. First, we sorted the cases by increas- ing value on that dimension, and then for each dimension we formed Group 1 by taking the three cases with the lowest value, Group 2 con- sisting of the next four cases, Group 3 consisting of the subsequent four cases, and finally Group 4 with the three cases with the highest value on the dimension. For the dimension self-imposed time windows, we grouped all cases into three groups corresponding to the three values available on our measurement scale for this variable see Table 7.2. After having formed these groups, we compared the steepness of the rise in distribution costs resulting from the increase in the number of time access windows from level 1 to 6 between these groups. Table 7.2 Distribution strategy dimensions per case Case Stops per Vehicle Stopping time Distance between Self-imposed roundtrip capacity per vehicle stores and DC time windows 1 5.4 3.9 64 110 normal 2 1.2 5.6 122 127 normal 3 2.4 3.5 155 103 normal 4 3.4 4.9 83 76 normal 5 1.0 4.9 63 89 strict 6 4.2 5.5 185 116 normal 7 9.1 5.0 181 198 long 8 8.3 1.8 165 103 long 9 7.2 1.0 72 86 normal 10 10.2 2.6 256 102 long 11 1.2 4.9 47 71 normal 12 1.3 4.7 78 42 normal 13 1.1 3.0 17 32 normal 14 6.3 4.9 134 42 normal