2. Marco teórico
2.1.4. Importancia del gobierno de las TIC
Situation: “Carrier contract is being considered”
“Please, since the free cash flow of the company is low, avoid contracting external carriers that require advanced payment of tariffs”
Requirement: “when free cash flow is low, prefer trucks from the com- pany’s fleet over external carriers”
“Percentage of company’s cash
comiited to pay freight tariffs”
Target: 10% Current: 25%
Figure 9.9 Context information including Finance’s strategy and indicators
recommendation of the Finance department. If the user’s decision shows to be unjustified, the Finance department can identify such pattern.
9.7
Simulation of the Scenario
To evaluate the gains from the SA-BPM approach in the FLM’s new strategy, we performed a simulation of the scenario.
The scenario was set up with the following settings:
• shipment areas: The city was divided in four “shipment areas”. Each area corresponds to the area covered by one truck in a single trip;
• truck fleet: FLM has a fleet of 6 trucks, each with a load capacity of 20 items. Regardless of the area, a truck expends one hour in average and exactly $50 dollars to perform the delivery over the area;
• order dispatch: orders are grouped by area. When the number of items for an area reaches 20 items, one of the available trucks is loaded and dispatched to the area;
• carriers: a carrier covers all four areas in a single trip. Its load capacity is 60 items. There are both cheap and expensive carriers in the market. A cheap carrier costs exactly $100 dollars, while an expensive carrier costs exactly $300 dollars. A carrier is always available when needed, but, at certain moments, cheap carriers are not available;
9.7 SIMULATION OF THE SCENARIO 160
Table 9.4 Cost per Product Delivered (cpp)
Max. Load Cost cpp
Fleet’s truck 20 $50 $2.50
Cheap carrier 60 $100 $1.67
Expensive carrier 60 $300 $5.00
• carrier assignment rule: when the sum of pending orders for all areas reaches 60 items, a carrier is contracted to deliver these orders.
The following statistical information was employed:
• order arrival: 266 items are ordered per day. The arrival was modeled as a Poisson process with a 33.3 arrival rate (hourly). A day is considered to span 8 work-hours. Items can be ordered by retailers in any of the four areas. The distribution of items among areas is uniform. Hence, any item has equal probability to have been ordered from each of the four areas;
• carrier prices: carrier prices change periodically from cheap to expensive. At any time, there is a 50% chance that a cheap carrier is available;
• shipment time: the time necessary for a truck to ship all items in its area has an exponential distribution with mean equal to 1 hour.
The scenario was modeled using Generalized Stochastic Petri Nets (GSPN) [71].
Two setups were simulated: a setup with no adapter and a setup where the shipment cost adapter (D.2) was active. The presence of the adapter determines the carrier assignment rule. When the adapter is active, the employees avoid contracting expensive carriers, regardless of the number of orders pending. These orders are delivered by the company’s fleet until the carrier price drops.
Observe that the adapter makes its recommendation based on an estimation of the carrier prices. This estimation may, however, be wrong. A delay between the decision moment and the actual contract of the carrier may provoke the raise of the price in the meantime. Thus, even when the adapter is active, it may occur that expensive carriers are contracted.
9.7.1 Results
The GSPN was simulated using the TimeNet tool [37] with a 99% confidence level and 5% of error tolerance. The following metrics where computed:
• Throughput – u: the throughput of the GSPN in terms of items delivered per day. The throughput is split in three flows:
– uf leet: items delivered through the company’s trucks;
– uchp: items delivered by cheap carriers;
– uexpen: items delivered by expensive carriers;
• Number of Deliveries – d: the number of deliveries per day. It is computed from the throughput on the basis of how many items are shipped at once in each delivery. It is split in:
9.7 SIMULATION OF THE SCENARIO 161
Table 9.5 Setup 1 - No adapter. 99% confidence; 5% error.
Values per day Fleet Chp. carrier Expen. carrier Total
Items delivered (u) 123.2 86.4 86.4 296
Number of deliveries (d) 6.16 1.44 1.44 9.04
Deliveries cost (c) $308 $144 $432 $884
Table 9.6 Setup 2 - Adapter D.2. 99% confidence; 5% error.
Values per day Fleet Chp. carrier Expen. carrier Total
Items delivered (u) 228.8 48.0 19.2 296
Number of deliveries (d) 11.44 0.80 0.32 12.56
Deliveries cost (c) $308 $80 $96 $748
– df leet: number of fleet’s trucks deliveries; – dchp: number of cheap carriers deliveries;
– dexpen: number of expensive carriers deliveries;
• Deliveries Cost – c: the daily costs of delivery, computed from the delivery cost of each shipment method and the number of deliveries per day. It is similarly split in three costs, for each shipment method.
The results of the simulations are presented in Tables 9.5 (setup 1) and Table 9.6 (setup 2). Observe that, without the adapter, the carrier option is adopted more frequently, with only 123 items in a day being delivered with the company’s fleet, while 173 items are delivered through carriers. Observe, also in setup 1, that both cheap and expensive carriers are contracted with the same chances, regardless of their price. The total cost for the company is of $884 per day in this setup.
In the setup 2, the use of the company’s fleet raises. A number of 229 items are delivered per day with the fleet’s trucks, while only 67 are delivered by way of carriers. As a result, the fleet’s trucks make much more deliveries in a single day (aprox. 2 deliveries per truck for a total of 11 deliveries). It can be observed that, when carriers are contracted, most of the time cheap carriers are contracted. Expensive carriers are contracted when the prices change between the time the recommendation is computed and the moment the carrier is actually contracted. At the end of the day, the FLM expends $748 with deliveries.
The performance indicator targeted by Adapter D.2 is the average cost of delivery per product, which we denote by cpp(AV G). This indicator can be computed through the Eq. 9.1.
AV G(cpp) = uf leetcppf leet+ uchpcppchp+ uexpencppexpen uf leet+ uchp+ uexpen
(9.1) Through Eq. 9.1 we can compute the average costs of each setup as being:
Setup 1 (no adapters):