Littlefield Simulation Game

A Report Submitted to

Instructor: Prof. Sachin Jayaswal

In partial fulfilment of the requirements of the course

OM-II (2013-14)

Group: D-15
Team Name: xibalba
Section: D
On
19 March 2014

INDIAN INSTITUTE OF MANAGEMENT, AHMEDABAD

1
Overall Analysis and Strategy 𝐷𝑒𝑚𝑎𝑛𝑑 = 4.367 + 0.397 ∗ 𝑋

Key decision questions before the game Putting X = 60, we forecasted the stable
started: demand to be around 35 orders per day. In
1. Demand Prediction retrospect, due to lack of sufficient data,
2. Capacity Planning we fell short of actual demand by 15 units,
3. Inventory Management which also hurt our further decisions. We
4. Clearing Backlog Orders lost valuable time due to this.

During the game:

5. Scheduling at Station Two
6. Contract Decisions
7. Lot Sizing and Lead Times
8. Decisions related to Debt
9. WIP Limit Calculations

Towards the end of the game, following

decisions became important:
10. When to sell and how much to sell?
11. What Capacity and Inventory settings
to leave when the game ends?
Table 2: Predicted versus Actual Demand

I. BEFORE THE START

Inventory Decisions
Demand Prediction
The next decision was to set inventory
The demand was expected to follow a parameters – EOQ and ROP. Once the
trapezoidal shape. We used linear game opened, only ten days of increasing
regression to predict the stable demand demand were left; while our first shipment
level which was expected to be achieved of inventory would have arrived only on
by day 60. However, when the game day 62 (Arrival of previous order – day 55;
started, we only had true data for 17 days, lead time of 7 days after that). Therefore,
since the inventory was exhausted after we decided to calculate EOQ based on the
this period. The result of the analysis gave stable demand prediction.
the regression line to be (See fig. 1):

2∗𝐷∗𝑆
𝐸𝑂𝑄 =
𝐻
𝐷 = 35 ∗ 365 = 12775 𝑢𝑛𝑖𝑡𝑠
𝑆 = $ 50 𝑝𝑒𝑟 𝑜𝑟𝑑𝑒𝑟
𝐻 = 0.12 ∗ $10 ∗ 60 = $ 72 / 𝑢𝑛𝑖𝑡 / 𝑦𝑟

Putting in the values, we got the EOQ as:

𝐸𝑂𝑄 = 133 𝑢𝑛𝑖𝑡𝑠

Also, to calculate ROP,

𝑅𝑂𝑃 = 𝐷𝐷𝐿𝑇 + 𝑆𝑆
Table 1: Demand Forecast based on first 17 days
𝑅𝑂𝑃 = 7 ∗ 35 + 0 = 245
(𝑐ℎ𝑜𝑠𝑒 𝑠𝑠
= 0 𝑠𝑖𝑛𝑐𝑒 𝑑𝑒𝑚𝑎𝑛𝑑 𝑤𝑎𝑠𝑛′ 𝑡 𝑡𝑜𝑜 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒)
2
Since EOQ < ROP, this will mean that our with a purchase of 9, 18 and 19 machines
inventory will get exhausted pretty soon. for station 1, 2 and 3 respectively – to
Minimum we need to order, therefore, is make the total count of machines to reach
ROP, i.e., 245 units per week. 12, 24 and 27 respectively.

Initial Capacity Planning We later increased the machine count to

20, 30 and 37 in order to reduce lead time
This is one area where the group feels it to bid for level 3 contracts. These numbers
did a fatal mistake. Instead of going for were arrived using the 4 step model which
heavy investments at the very beginning is described towards the end of section II.
and switching to highest paying contracts,
we decided to invest in capacity gradually Clearing Backlog Orders
and move from lowest paying to highest
paying contract, after generating money Another Mistake! We missed the fact that
through operations only. In retrospect, this we can clear the backlog orders quickly by
wasn’t a good decision, as the demand setting WIP limit to 0 in the very
period was quite short and it made sense to beginning. This cost us 20 valuable days,
invest as early as possible to reap benefits where new orders piled up as backlog,
for a larger period of time. which we were relying to clear by using
our capacity. It took us some time to
To begin with, station 2 and station 3 realize and correct this mistake.
appeared to be the bottleneck as their
utilization was stuck at 100% whereas II. DURING THE GAME
station 1 was underutilized (Fig 1)
The following five decisions were very
closely related to each other, and were
closely monitored during the entire game:
Contract decision, WIP, Lot Sizing,
scheduling at station two and Debt. We
therefore, first give a brief overview and
rationale for each of them. We’ll then
present a comprehensive mathematical
model that we used to make these
decisions.

Figure 3: Utilization status when the game opened Contract Decisions

We started with level 1 contract which

A quick analysis showed that (for a lot size gave us $ 7500 for each completed order.
of 60) for a balanced line, the ratio of Very soon we realized that we are capable
machines required at station 1, 2 and 3 of moving to level 2 which is
should be – 1:2:2.2, which will give a economically better. Keeping this in mind,
streamlined throughput of 3.2 orders per we switched to level 2 contracts on day 82.
day per machine. The move from level 2 to level 3 wasn’t
that easy. A quick analysis showed that
Going by our overall approach of no debt achieving lead time of 0.5 days would not
and gradual investment, we decided to buy be possible with the capacity that we had,
2 more machines each for station 2 and 3. even if we reduce the lot size and the
Considering the need for high capacity, number of accepted orders. We, therefore,
this decision was quickly supplemented invested in capacity – and chose a

3
combination of WIP and lot size that number of lots is shown in the following
would give us a theoretical lead time of 0.6 graph:
days. This switch was made on day 122
and actual lead time achieved was close to
0.59 days. Why didn’t we push for a lead
time less than 0.5 days? It was already day
122, maximum demand was still not
known to us (we were accepting orders
less than the actual demand to meet lead
time requirements), the game stalled twice
for brief period of 4 days each – therefore,
the payback of investment was uncertain.
Following our conservative strategy, we
decided against making any new
investments after this point. In retrospect,
this was again a bad decision. A quick
breakeven analysis (see table 2) showed Figure 4: Analysis for lot size
that the investments would pay back very
quickly. This, however, didn’t become
visible to us since we were not exposed to Decisions Related to Debt
the total demand in the market.
Our initial aversion for raising debt was
Lot Sizing and Lead Times very soon changed, when we realized that
capacity (rather than immediate cash)
The question of lot size came into picture would be the key to success. Moreover, the
when we were trying to push the lead time cost of debt was also much lower than the
below 0.5 days. As per our calculations, benefits achieved from buying the
for a very large number of machines, lot machines. For example, the payback for a
size of 10 would have given best results. debt of $ 750,000 with interest would
break even in less than two days! (See
However, since we had limited number of Table 2)
machines at our disposal, and since we
were constrained by the capacity towards Assumption: Level 2 Contracts
the later part, To calculate this, we first Debt Raised $ 750000
Interest for 2 days 600
calculated data tables with lot size and Total 750600
WIP limit as two parameters, to calculate Payback < 2 Days
Time in System (= Time in Queue + Time Table 2: Illustrative Break Even Analysis
of processing) for each of the stations. The
total lead time for the order would be the
sum of these three times. We then chose Therefore, after day 80, Debt was raised as
the minimum possible lead time that was and when required, and was paid back as
possible with the number of servers that soon as the cash became available.
we had at our disposal. Using this
approach, for (20, 30, 37) machines, WIP
optimal lot size for us came out to be 20
units per lot. The tables are shown in Calculating WIP number was relatively
Appendix 2. For a system with number of easier. We used Little’s Law for this:
servers equal to the machines we had
bought, total processing time versus 𝑊𝐼𝑃 = 𝑅 ∗ 𝑇
𝑅 = 𝑇ℎ𝑟𝑜𝑢𝑔ℎ𝑝𝑢𝑡 𝑅𝑎𝑡𝑒

4
𝑇 = 𝑇𝑖𝑚𝑒 𝐼𝑛 𝑇ℎ𝑒 𝑆𝑦𝑠𝑡𝑒𝑚 = 𝐿𝑒𝑎𝑑 𝑇𝑖𝑚𝑒
Step 3: We then summed up total system
Towards the later part when we had to times of each of the stations to get a
push the lead time below 0.5 days, we ballpark number of lead time (Appendix
chose a different procedure to estimate the 2). Excluding the infeasible solutions, we
WIP that will give the maximum revenue chose the lot size and effective demand
within our capacity constraints. Details of value which gave the least lead time for
the method are provided in next subsection the number of machines that we had.
and Appendix 3.
Step 4: For the chosen lot size, we
Mathematical Model empirically calculated the throughput rate
when no queue was present using 3 day
As mentioned earlier, we were aversive of average after implementing the decision at
losing cash by investing in machines after a very low WIP. We then worked out the
day 122. Therefore, we kept the number of trade-off by increasing WIP, which will
servers constant and decided to calculate lead to higher queue times, thereby
optimum lot size and WIP which will give increasing lead time => lower revenues;
us the maximum revenue/day with the while at the same time, it will allow us to
capacity constraints. We used the process more orders – thereby increasing
following steps to take decisions about the the revenues. This analysis gave us the
variables mentioned in section 2, following graph.
especially WIP and Lot Size.

Step 1: We approximated each station

with a queue system and calculated the
queue times as well as processing times for
a given lot size and demand. A screenshot
of the tool is provided below (This excel
tool was provided to us in OM-I by Prof.
Chetan Soman):

Basic Inputs: Number of Servers, S = 50

Arrival Rate, l = 5
Figure 1: Maximizing revenue within capacity
Coefficient of Variation of Inter-arrival time, CV(a) = 1
Service Rate Capacity of each server, m = 0.33333
constraints
Coefficient of Variation of Service time, CV(s) = 0

The Waiting Line: Average Number Waiting in Queue (Nq) = 0.000 <== The Approximation
Average Waiting Time (Tq) = 7.5E-07 Based on this, we chose our WIP to be 28,
Service: Average Utilization of Servers = 30.00% which was supposed to give us revenue of
Average Number of Customers Receiving Service (Ns) = 15 $800,000 per day. The actual average was
The Total System (waiting line plus customers being served): also very close to this number.
Average Number in the System (N) = 15.000
Average Time in System (Tq + Ts) = 3
III. TOWARDS THE END
Figure 5: Snapshot of tool used to approximate system
times
We didn’t want to leave to many assets
and inventory when the game was taken
Step 2: Using this tool, we then created a out of our control, since their salvage
data table which showed lead times for value would be lost. This forced us to take
various combinations of demand and lot two more decisions:
size, for each of the stations. These values
are approximate only and are provided in
appendix 2.

5
Selling the Capacity Back We needed to make the best use of
available resources to get the
On day 217, we sold all our machines maximum benefit out of it.
leaving only 3, 5 and 5 machines at station
1, 2 and 3 respectively. The rationale 6. Utilization and Lead times: There
behind all these decisions was– Not to is always a trade-off. We need to
leave too many assets for the end when choose a value which gives us the
they will give us no salvage value. optimum result, in this case,
maximum revenue.
How much to leave for the last 50 Days

We set back the contract to the very first

level; reduced WIP so that queue times Strategic
don’t become very high. We also switched
back to a lot size of 60. The rationale 1. Go Aggressive. There is no point in
behind all these decisions was again – Not shying away from the investments
to leave too many assets for the end when if you see the worth of it. We could
they will give us no salvage value. have raised the debt in the very
beginning only – which we didn’t
IV. KEY LEARNINGS do.

Operational 2. Do not underestimate the

competition: We assumed that in
1. Forecasting Demand and making the very beginning, all teams will
adjustments on it as and when the take some time to push the
new data becomes available. production forward. This didn’t
happen. Competition was
2. Actual implementation of all the aggressive, better prepared and
inventory theory that we had therefore took an early lead.
studied so far.

3. Predicting queue times and system

times by modelling the systems and
then using the results to predict the
lead times.

4. Tweaking with number of servers

and lot sizes to achieve the desired
lead times. This was probably one
of the major learning for us – how
to balance utilization and lead
times practically using all the
theory that we had learnt.

5. How to achieve maximum revenue

when the capacity is constrained.

6
 APPENDIX 1
DECISION HISTORY

Day Parameter Value 91 Max WIP limit 60

50 Reorder quantity (kits) 2700 91 Decrease Debt 6,499,997
50 Reorder point (kits) 3000 92 Reorder point (kits) 18000
station 2 machine 107 Reorder quantity (kits) 13200
50 5
count 107 Max WIP limit 62
station 3 machine 109 Decrease Debt 5,599,997
50 5
count
109 Decrease Debt 4,599,998
64 Reorder quantity (kits) 6000
109 Decrease Debt 3,599,999
65 Increase Debt 900,000
109 Decrease Debt 2,600,000
65 Increase Debt 1,800,000
109 Decrease Debt 1,700,000
65 Increase Debt 2,700,000
109 Decrease Debt 800,000
65 Increase Debt 3,600,000
109 Decrease Debt 0
station 1 machine
65 9 110 Lots per order 3
count
station 2 machine 110 Increase Debt 900,000
65 18
count 110 Increase Debt 1,800,000
65 Increase Debt 4,000,000 110 Increase Debt 2,700,000
station 3 machine station 1 machine
65 19 110 20
count count
73 Reorder quantity (kits) 2700 station 2 machine
110 30
77 Reorder quantity (kits) 16800 count
77 Increase Debt 4,999,999 station 3 machine
110 37
count
77 Increase Debt 5,999,998
111 Lots per order 1
station 1 machine
77 12 111 Reorder quantity (kits) 12000
count
station 2 machine 111 Decrease Debt 1,800,000
77 24
count 112 Lots per order 3
77 Increase Debt 6,999,997 112 Decrease Debt 1,400,000
station 3 machine 113 Decrease Debt 1,190,867
77 27
count
114 Decrease Debt 390,867
79 Max WIP limit 20
119 Max WIP limit 40
81 Max WIP limit 0
119 Decrease Debt 0
88 Max WIP limit 50
124 Contract number 3
91 Contract number 2
124 Reorder quantity (kits) 15000
91 Reorder point (kits) 13980
124 Reorder point (kits) 16800

7
124 Max WIP limit 28 station 2 machine
217 5
count
134 Max WIP limit 25
station 3 machine
140 Max WIP limit 28 217 5
count
151 Reorder quantity (kits) 16800
217 Lots per order 1
196 Reorder quantity (kits) 12000
217 Reorder quantity (kits) 7200
206 Reorder quantity (kits) 8400
217 Reorder point (kits) 8400
station 1 machine
217 3
count

APPENDIX 2
TIME IN SYSTEM FOR VARIOUS LOT SIZE AND DEMANDS

Machine 1 System Times

3.355219 60 30 20 15 12 10
20 7.54 4.57 3.65 3.36 4.08 -0.46
21 7.56 4.60 3.71 3.58 9.45 0.90
22 7.57 4.63 3.80 4.05 -1.38 1.25
23 7.60 4.67 3.94 5.52 0.74 1.40
24 7.62 4.73 4.17 #DIV/0! 1.24 1.46
25 7.65 4.81 4.63 -0.55 1.45 1.48
26 7.70 4.93 5.75 0.92 1.54 1.48
27 7.75 5.10 12.25 1.38 1.59 1.46
28 7.82 5.37 -4.12 1.59 1.60 1.43
29 7.90 5.84 0.30 1.70 1.60 1.39
30 8.02 6.80 1.25 1.75 1.58 1.35
31 8.17 9.77 1.65 1.78 1.55 1.29
32 8.39 #DIV/0! 1.85 1.78 1.51 1.23
33 8.69 -2.31 1.97 1.77 1.47 1.17
34 9.16 0.65 2.03 1.75 1.42 1.09
35 9.91 1.61 2.07 1.72 1.36 1.02
36 11.34 2.07 2.08 1.68 1.30 0.93
37 14.86 2.32 2.07 1.64 1.24 0.85
38 36.24 2.48 2.06 1.59 1.17 0.75
39 -13.81 2.57 2.04 1.54 1.09 0.66
40 -1.37 2.63 2.01 1.48 1.01 0.55

Machine 2 System Times

10.28932 60 30 20 15 12 10
20 7.56 3.78 2.52 1.89 1.51 1.26
21 7.58 3.79 2.53 1.89 1.52 1.26
22 7.61 3.81 2.54 1.90 1.52 1.27

8
 23 7.66 3.83 2.55 1.91 1.53 1.28
24 7.72 3.86 2.57 1.93 1.54 1.29
25 7.80 3.90 2.60 1.95 1.56 1.30
26 7.91 3.95 2.64 1.98 1.58 1.32
27 8.06 4.03 2.69 2.01 1.61 1.34
28 8.26 4.13 2.75 2.07 1.65 1.38
29 8.54 4.27 2.85 2.14 1.71 1.42
30 8.94 4.47 2.98 2.23 1.79 1.49
31 9.49 4.74 3.16 2.37 1.90 1.58
32 10.29 5.14 3.43 2.57 2.06 1.71
33 11.48 5.74 3.83 2.87 2.30 1.91
34 13.36 6.68 4.45 3.34 2.67 2.23
35 16.55 8.27 5.52 4.14 3.31 2.76
36 22.71 11.35 7.57 5.68 4.54 3.78
37 38.29 19.14 12.76 9.57 7.66 6.38
38 134.18 67.09 44.73 33.55 26.84 22.36
39 -91.38 -45.69 -30.46 -22.85 -18.28 -15.23
40 -35.74 -17.87 -11.91 -8.94 -7.15 -5.96

Machine 3 System Times

5.283225 60 30 20 15 12 10
20 16.52 9.02 6.52 5.28 4.55 4.08
21 16.52 9.02 6.53 5.30 4.58 4.13
22 16.53 9.04 6.55 5.33 4.62 4.21
23 16.55 9.05 6.57 5.36 4.69 4.35
24 16.56 9.07 6.61 5.42 4.81 4.62
25 16.59 9.10 6.65 5.51 5.00 5.21
26 16.62 9.15 6.73 5.66 5.38 7.13
27 16.67 9.21 6.83 5.90 6.25 #DIV/0!
28 16.74 9.29 7.00 6.37 9.50 -1.06
29 16.83 9.42 7.28 7.41 -20.60 0.82
30 16.95 9.61 7.78 11.08 -0.71 1.36
31 17.13 9.91 8.79 -38.03 0.98 1.55
32 17.37 10.40 11.59 -1.40 1.53 1.59
33 17.74 11.28 39.64 0.92 1.74 1.55
34 18.29 13.14 -5.56 1.67 1.79 1.45
35 19.16 19.00 0.17 1.97 1.76 1.31
36 20.67 #DIV/0! 1.65 2.08 1.67 1.13
37 23.67 -5.33 2.24 2.08 1.53 0.92
38 31.73 0.48 2.50 2.01 1.35 0.67
39 100.61 2.28 2.59 1.89 1.14 0.38
40 -20.67 3.06 2.57 1.72 0.89 0.07

Total System Times

9
 60 30 20 15 12 10
20 31.62 17.37 12.69 10.53 10.14 4.89
21 31.66 17.41 12.77 10.77 15.55 6.29
22 31.72 17.47 12.88 11.28 4.76 6.73
23 31.80 17.55 13.06 12.80 6.96 7.03
24 31.90 17.66 13.35 #DIV/0! 7.59 7.37
25 32.04 17.81 13.88 6.91 8.01 8.00
26 32.23 18.03 15.11 8.55 8.51 9.93
27 32.48 18.33 21.77 9.30 9.45 #DIV/0!
28 32.82 18.79 5.64 10.02 12.75 1.75
29 33.28 19.53 10.43 11.24 -17.29 3.64
30 33.91 20.88 12.01 15.07 2.66 4.20
31 34.79 24.42 13.60 -33.88 4.43 4.42
32 36.05 #DIV/0! 16.87 2.96 5.10 4.54
33 37.92 14.71 45.43 5.56 5.50 4.63
34 40.81 20.48 0.92 6.76 5.88 4.77
35 45.62 28.88 7.75 7.83 6.43 5.08
36 54.72 #DIV/0! 11.29 9.44 7.51 5.85
37 76.82 16.13 17.08 13.29 10.42 8.14
38 202.16 70.05 49.29 37.15 29.35 23.78
39 -4.58 -40.84 -25.84 -19.42 -16.05 -14.19
40 -57.78 -12.18 -7.34 -5.74 -5.25 -5.34

10
 APPENDIX 3

WIP CALCULATION TO MAXIMIZE TOTAL REVENUE WITHIN CAPACITY

CONSTRAINTS

Total Revenue vs WIP for chosen Lot Size

Total Revenue Lead Time

900000 1.000
800000 0.900 L
R 700000 0.800 e
e 600000 0.700 a
v 0.600 d
500000
e 0.500
400000
n 0.400 T
300000 0.300
u i
e 200000 0.200 m
100000 0.100 e
0 0.000
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
WIP

Average Total
I R T Revenue Revenue
20 55 0.364 30000 600000
21 55 0.382 30000 630000
22 55 0.400 30000 660000
23 55 0.418 30000 690000
24 55 0.436 30000 720000
25 55 0.455 30000 750000
26 55 0.473 30000 780000
27 55 0.491 30000 810000
28 55 0.509 29455 824727.2727
29 55 0.527 28364 822545.4545
30 55 0.545 27273 818181.8182
31 55 0.564 26182 811636.3636
32 55 0.582 25091 802909.0909
33 55 0.600 24000 792000
34 55 0.618 22909 778909.0909
35 55 0.636 21818 763636.3636
36 55 0.655 20727 746181.8182
37 55 0.673 19636 726545.4545
38 55 0.691 18545 704727.2727
39 55 0.709 17455 680727.2727

11
 40 55 0.727 16364 654545.4545
41 55 0.745 15273 626181.8182
42 55 0.764 14182 595636.3636
43 55 0.782 13091 562909.0909
44 55 0.800 12000 528000
45 55 0.818 10909 490909.0909
46 55 0.836 9818 451636.3636
47 55 0.855 8727 410181.8182
48 55 0.873 7636 366545.4545
49 55 0.891 6545 320727.2727
50 55 0.909 5455 272727.2727

12