1) The team developed a strategy to minimize total supply chain costs which included determining the optimal number of production resources to activate, capacity levels, and maximum delivery delays.
2) Demand orders were ranked and allocated based on penalty cost, delivery date, and order quantity to reduce delay costs.
3) Three demand orders were produced in advance to release production capacity and deliver other critical orders on time or with reduced delays, further lowering costs.
5. • The available levers can be divided in three operational areas: Heuristic, Demand ordering and Demand
anticipation.
• Even though these areas are mutually interrelated, we decided to follow a reference decisional
sequence.
1. Heuristic: deciding how many resources to activate,
which capacity level to deploy and the maximum
delivery delay.
2. Demand ordering: setting the right priorities to
efficiently allocate demand orders in the available
production capacity for each time bucket.
3. Demand anticipation: deciding 3 orders to produce
in advance in order to further improve the cost
performance.
General strategy
7. Cost per hour
(€/h)
Laser Cutting Cutting Bending Assembly Final
Assembly
Preferred
Resource
50 – 60 - 70 50 – 60 – 70 50 – 60 – 70 50 – 60 – 70 50 – 60 – 120
Internal
Alternative
/ 50 – 60 – 70 50 – 60 – 70 50 – 60 – 70 60 – 75 – 120
Contractor
(open order)
/ / / 100 100
Contractor
(extra)
/ / / 300 300
While making decisions in terms of heuristic management (how many resources to activate
and which capacity level to deploy), we decided to focus on just two activities - assembly and
final assembly – as production costs for laser cutting, cutting and bending are not
not differential.
Heuristic: Production costs Table
8. Option 1: No extra suppliers
Delay cost = 200*7 = 1400 €
Cost of Bianchi’s capacity =
10*1*100 = 1000 €
Total cost = 2400 €
Option 2: Activating extra
suppliers
Cost of Verdi’s capacity = Total
cost = 10*1*300 = 3000 €
Heuristic: How many resources to activate
Lower
cost
We chose not to activate extra suppliers, as relying just on suppliers with long term contracts is always cheaper than
activating extra suppliers.
AN EXAMPLE
• Capacity Level = 2
• We have to perform final assembly (LT=1
hour) on a 10-unit order (the average
order size is 9.8), but Bianchi has reached
full capacity
• With no extra suppliers, the worst case
scenario is that we are forced to delay the
order and use Bianchi’s capacity (the most
expensive) in the following week; the cost
of delay is 200 € per day (max delay cost)
• If we activated extra suppliers, we would
use Verdi’s capacity
Even the worst case scenario with no extra suppliers is
cheaper than activating extra suppliers
10. 63 days of
allowed delay
ensures that each
and every
demand order is
satisfied
11.
12.
13. Option 1: Level 2
Cost of Bianchi’s capacity = Total
cost = 10*1*100 = 1000 €
Option 2: Level 3
Cost of the internal alternative’s
capacity = Total cost =
2*1*75+8*1*120 = 1110 €
Heuristic: Which capacity level to deploy
Lower
cost
We chose not to activate Level 3, considering that most of scenarios where Saturday overtime is deployed (Level 3)
are actually more expensive than scenarios where production takes place 5 days a week, 10 hours a day (Level 2).
AN EXAMPLE
• We activated only suppliers with long term
contracts
• We have to perform final assembly (LT=1 hour) on a
10-unit order
• With no Saturday overtime, Bianchi is the only
resource with enough available capacity
• By activating Level 3, we have available capacity on
the internal alternative: we can produce 2 units in
the last two hours on Friday and the remaining 8 on
Saturday
Activating Level 3 may be convenient when Saturday overtime is used instead of delaying orders,
but this specific situation did not occur frequently in the simulations we ran;
therefore, not activating Level 3 leads to a general cost reduction
16. Penalty
cost
A-Z
from lowest to
highest delay cost
per day
Z-A
from highest to
lowest delay cost
per day
Delivery
date
A-Z
from nearest to
farthest delivery
date
Z-A
from farthest to
nearest delivery
date
Order
quantity
A-Z
from smallest to
biggest order size
Z-A
from biggest to
smallest order size
Our ranking
Penalty Z-A
Date A-Z
Quantity A-Z
1
3
2
Demand ordering: Available criteria and ranking
17. Delay
Penalty
(€/Day)
No. of
orders
No. of
delayed
orders
200 12 0
150 4 0
140 1 0
130 3 0
120 4 0
100 302 23
80 11 2
50 1 1
40 1 0
20 1 1
We selected delay penalty cost as the first criterion in our customer order
ranking. In this way, we were able to reduce the probability that an order
with a high penalty cost were delivered in delay.
The table aside shows the number
of orders for each “delay penalty”
category.
As the last column underlines, in
our solution only low-penalty cost
orders have been delivered in
delay (those with a penalty of
100€/day or below).
Demand ordering: Penalty cost
18. In choosing between Date A-Z and Date Z-A, we tried to figure out how this decision affected delay costs.
Therefore, we analysed how orders are differently allocated according to these two criteria.
Demand ordering: Delivery date (1/3)
By allocating orders with nearest delivery date first (Date A-
Z), we had several delayed orders with a reduction in the
average number of delay weeks per order. On the contrary,
by allocating orders with farthest delivery date first (Date Z-
A), we had fewer delayed orders but an increase in the
average number of delay weeks. In the two scenarios, these
effects tend to neutralize each other, causing nearly the
same delay costs.
After running a few simulations, we saw that, by choosing
Date A-Z, we could achieve a slight decrease of delay costs.
19. In an “infinite capacity” simulation, 2 orders are in overflow on TB 2. Hence, these orders
exceeding capacity have to be relocated on the following time buckets, causing a delay.
AN EXAMPLE
• Only 1 resource
• Capacity = 30 units in every time
bucket
• Orders to be processed have the
same quantity (10 units)
• Delivery dates
• 1 order on Time Bucket (TB) 1
• 5 orders on TB 2
• 3 orders on TB 3
• 2 orders on TB 4
• 1 order on TB 5
Demand ordering: Delivery date (2/3)
20. Date Z-A Date A-Z
Number of delayed orders = 2
Average delay weeks per order = 2.5
Number of delayed orders = 5
Average delay weeks per order = 1
Demand ordering: Delivery date (3/3)
21. At this stage of allocating demand orders, weekly delay cost and delivery date are the same for each
order we consider, as these two criteria come first in our customer order ranking.
Demand ordering: Order quantity (1/2)
Therefore, in choosing between Quantity A-Z and Quantity
Z-A, we tried to minimize the number of delayed orders as
we knew that it would lead for sure to a decrease in delay
costs (as said, at this stage weekly delay cost is a fixed
parameter).
We decided to allocate orders with fewest quantity first
(Quantity A-Z), as this option allows a smarter distribution
over the time buckets, characterized by a lower number of
delayed orders.
22. AN EXAMPLE
• Only 1 resource
• Capacity = 50 units in
every time bucket
• Orders to be processed
• 1 order : 20 units
• 4 orders: 10 units
• 2 orders: 5 units
• Delay cost = 700 €/week
for every order
• Delivery date on Time
bucket 1
Number of delayed orders = 1
Total delay cost = 700 €
Number of delayed orders = 3
Total delay cost = 2100 €
Quantity A-Z Quantity Z-A
Demand ordering: Order quantity (2/2)
26. Plan cost
minimization
Production
costs
Moving orders from
expensive resources (e.g.
external contractors) to
more efficient ones (e.g.
preferred resource)
Delay costs
Producing on time (or
reducing delay time) orders
in delay
Demand anticipation: Strategy
Low
potential
gains
Better
option
How to reduce
delay costs?
1. Anticipate orders with the highest total delay cost (= daily penalty * days of delay)
2. Anticipate orders to release production capacity and deliver other critical orders on
time or at least with reduced delay
3. A mix of the two! Our strategy!
27. Demand anticipation: In detail
• First we analysed the “infinite capacity” scenario, which showed that the most critical time bucket is
week 25 (20/06): 290 items to deliver.
In our “finite capacity” solution, this situation resulted in a huge number of delayed orders from week
26 to week 30.
• Second, we isolated the orders with the highest total delay cost: order 204 (delay cost = 1750€) was
produced in week 30 (5 weeks in delay), whereas orders 210, 217 and 219 (delay cost = 1400€ each) in
week 27.
Order Delivery
Date
Planned
Date
Delay
(days)
Delay
Penalty
(€/day)
Total Delay
Cost (€)
OV20160204 20/06/2016 25/07/2016 35 50 1750
OV20160210 20/06/2016 04/07/2016 14 100 1400
0V20160217 20/06/2016 04/07/2016 14 100 1400
0V20160219 20/06/2016 04/07/2016 14 100 1400
28. Demand anticipation: Final decision
• If we had anticipated order 204, we would have set some production capacity free in week 30, but,
evidently, this capacity could not have been used to produce any other order on time, since no delayed
orders were scheduled in week 31, 32 and 33.
• After running some simulations, we verified that, even though order 204 had the highest total delay
cost, by anticipating orders 210, 217 and 219 we could release capacity in week 27 and produce on time
(or with lower delay) some delayed orders that were previously allocated in week 28, 29 and 30.
Orders we decided
to anticipate
OV20160210
OV20160217
0V20160219
29. BEFORE . . .
Thanks to the anticipation, the number of delayed orders decreased by
13. This is attributable to two reasons:
• the 3 orders we anticipated are no longer in delay;
• other 10 delayed orders have been delivered on time thanks to the
released capacity that was previously used to produce the 3 orders
anticipated.
Accordingly, delay penalty
cost decreased by 35%.
Since load cost remained
approximately steady, the
plan cost is lower.
. . . AFTER
Demand anticipation: Result