June 9, 2013 Section 2, Team 9

Managing Capacity and Lead Time at Littlefield Technologies – Team 9’s Summary
The purpose of this simulation was to effectively manage a job shop that assembles digital satellite system receivers. The objective was to maximize cash at the end of the product life-cycle (270 days) by optimizing the process design.
REVENUE
25000 20000 15000 10000 5000 1 9 17 25 33 41 49 57 65 73 81 89 97 105 113 121 129 137 145 153 161 169 177 185 193 201 209 217 225 233 241 249 257 265 Total Revenue Demand DAYS 0

Figure 1 : Revenue and demand

DEMAND 25 20 15 10 5 0

As shown by the figure above, total revenues generally followed the same trend as demand. The few sections of negative correlation formed the basis for our critical learning points. Although the process took a while to completely understand during the initial months of the simulation, the team managed to adjust, learn quickly and finish in 7th place with a cash balance of $1,501,794. For the purpose of this report, we have divided the simulation into seven stages after day 50, explicating the major areas of strategically significant decisions that were made and their resulting effects. Initial Operations Strategy - Days 0 to 50 The goal was to act conservatively and avoid making reactionary decisions. The strategy was to reduce lead-time to 0.25 days (6 hours) as quickly as possible. The rationale was: • To move to contract option two which allowed us to maximize revenues per job; • To maintain work-in-process (WIP) inventory of less than 100 units to avoid turning away new customer orders until this backlog was cleared. A second stuffer was purchased as this was highly utilized during the first 50 days of the simulation. We also switched to smaller batch sizes (two lots of 30 kits each) to monitor its effects on lead times. Stage 1 – Days 50 to 88 On day 59, with the average daily job lead-time at 0.3, we recognized an opportunity to generate higher revenues via contract two. By day 71, the tuner had reached maximum capacity and a second one was purchased. However, lead times began to rise quickly so we moved back to contract one. Once operations stabilized by day 73, we successfully reduced lead times by decreasing the batch size to 3x20. By day 84, the tester was almost over utilized and the queues were increasing. In order to expedite jobs, we gave priority to jobs slated for step four. Stage 2 – Days 89 to 114 As demand increased, the queue for the tester reached 109 jobs. We increased the lot size 1x60 to clear this queue. The tester utilization was expected to continue on its upward trend and so another tester was bought on day 108. The batch size was also reduced back from 1x60 to 3x20. Economic Order Quantity (EOQ) At day 114, we tried to determine the appropriate batch size to minimize the inventory costs. Our research lead us to Economic Order Quantity. This required a forecasted demand to calculate the EOQ over various periods in the simulation. We plotted number of jobs arriving against the days and added a Page 1 of 3

June 9, 2013 Section 2, Team 9 regression line to predict demand at day 150, when the demand was supposed to stabilize. Our regression line gave us an equation of Demand = (0.0651 x Day Number) + 1.0651. At day 150 (assumed to be the peak), our formula predicted an average demand of 10.8 jobs per day. We approximated this to 11 for the sake of simplicity. 2 �� �� ��

������ =

D = Forecasted annual demand = 11×365 = 4015 orders per year S = Set up cost/order (assumed to be the shipping cost) =
!!!""!#$ !"#$# !"#$!!" !" ! !!!"#$%&

=

!""" !"#

= $8.33

H = Holding cost/order = assumed at 1/3rd the cost of the order = . 333 ×$600 = $200 Therefore EOQ = 18.3. We decided that the batch size of 3x20 was the optimal size and would further reduce lead times by reducing the WIP inventory. On day 114, the batch sizes were switched once again to 3x20 kits. At this time, the tuner was at low utilization (approx. 20%) so there seemed to be sufficient capacity to handle this switch in batch sizes. Now that we had two tuners running, the results we had experienced around day 73 were not anticipated. Stage 3 – Days 115 to 128 On day 116, we performed an analysis of our data points up to and including day 114 and concluded that contract two would be more lucrative if average lead times were kept under 0.33 days. While this would have meant $0 revenues on some days, our net profits would still be higher. Due to increased demand and an inability to process 3x20 units, the queues at the stuffer and tuner spiked on day 121. We knew that demand would continue to increase and so we purchased another stuffer. The negative correlation between revenues and demand at this time is seen in Figure 2.

REV

Figure 2: Stage 3

DEMAND

25000 20000 15000 10000 5000

18 16 14 12 10 8 6 4 2 118 120 122 124 126 128 130 132

0 0 Additionally, revenue was decreasing as lead-time had spiked to DAYS almost 24 hours. As such, we switched back to contract one to Total Revenue Demand reduce our losses. This strategy seemed successful and by day 125, lead times were decreasing and we switched back to contract two. Demand, however, continued to increase and by day 128, our tuner was over utilized and our queue for the tuner was in excess of 200 units. The team had overlooked the fact that the tuner took relatively the same time to process a 1x60 unit order as a 3x20 unit order; this oversight was starting to cost us deeply. There was no way to “purge” the system of our queues by halting orders; the alternative was to switch the batch size to 1x60 kits to clear the queue. The eventual recovery can be seen from days 128 to 132 in figure 2.

Stage 4 – Days 129 – 142 Purging the system was a slow process, but was essential to bring down lead times. The effects of the change in lot size to 1x60 were seen within a few days and the queues at the tuner were back down to manageable levels. The team realized the tuner would be unable to handle our calculated batch size of 3x 20. We conducted a cost benefit analysis using estimated lead times and revenues to determine if we should buy a third tuner. We decided that our demand was almost at its peak and our analysis concluded that no significant benefit would be realized from purchasing another tuner. We decided to optimize the process using a batch size of 2x30. Page 2 of 3

June 9, 2013 Section 2, Team 9 Stage 5 – Days 143 – 176 This period was a reality check for the demand forecast. The incredible variability in the demand at the peak levels was not accounted for. The number of arrivals reached 20 on some of the days, far in excess of the predicted value of 11. The tuner was again unable to handle the demand, leading to station three queues in excess of 200 units. As shown in Figure 3, revenues dropped as demand increased. The lot size was again changed to 1x 60 implementing our “purging” strategy from Stage 3. Once queue sizes were back to normal and the revenues started to match the demand, we switched the lot size back to 2x30.
REV

25000 20000 15000 10000 5000

Figure 3: Stage 5

DEMAND

25 20 15 10 5

Stage 6 – Days 177 – 220: Exit Strategy 0 0 As demand began to decrease at day 180, we maintained our DAYS approach (batch sizes of 2x30 kits and contract two). The team Total Revenue Demand considered the variability in demand that was previously overlooked and decided to only retire one stuffer at day 218 and one tester at day 219, ensuring we would have enough capacity to maintain a low lead time even at a demand of ten orders per day. This strategy proved effective and we maximized profits through the remainder of the simulation.
163
166 169 175 160

Lessons Learned The overarching takeaway from this exercise is the importance of being pro-active rather than re-active. This is elucidated through the main points below: 1) Demand forecasting: We initially neglected demand variability and so failed to estimate the capacity of the process. This also impacted the EOQ calculation. We should have considered using a 95% confidence interval. The standard deviation at day 114 was around 3 which would have given us a maximum demand range from 5 – 17 orders/day (���������� = ������������������ ������������ ± 2 ���������������� ��������������������) 2) Economic Order Quantity: Through this process, we were introduced to the EOQ concept. The EOQ calculation would have been beneficial at the start of the process (day 50). Even though we would have had to make several assumptions, it would have provided a better starting point. 3) Criticality of a thorough Analysis: An analysis of the capacity and utilization of the tuner would have made it obvious that a batch size of 20 would be too much for two tuners to handle in conjunction with three stuffers. Using the times given the stations and incorporating an allowance for the variability in the demand and the queues (buffers) before each station, we would have realized that our setup of 3-2-2 using 2x30 units would have been optimal from the start. Furthermore, an analysis of the two contracts at the outset and the determination of the breakeven point for the lead times would have made the decisions of switching contracts fairly easy. 4) Interdependency of operational measures: As we decreased the batch size, we reduced leadtime by minimizing WIP inventory (lead time was proportional to lot size). However, this is not the only factor affecting lead. We also needed to consider the capacity of the machines and the variability in inputs in collaboration with our analysis. 5) Decisions made in haste are rarely good ones: While the initial strategy was to avoid making hasty decisions, it was easy to get caught up in the ‘competition’. There were periods where contracts and lot sizes were switched back and forth without any real analysis. It would be prudent to maintain a control system to ensure hasty decisions are not made. Overall, the Littlefield simulation provided a valuable lesson in plant operations, particularly as they relate to issues of variability, demand forecasting and the interdependence of operational measures. Page 3 of 3

172

178