Customer Demand

Customer demand varies in both timing and quantity:

Time
Quantity
Individual Customer Order
Customer Demand
If demand for a product comes from many, independent
customers, then we dont need to be concerned about
individual customer orders, but rather cumulative
demand over a period of time.
Customer Demand
Demand
Day 1 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Day 9
Individual Customer Demand
Cumulative Demand for Period
Customer Demand
In statistics, when there is reason to suspect the
presence of a large number of small effects acting
additively and independently, it is reasonable to assume
that the observations will be normally distributed.
Therefore, if demand for a product comes from many,
independent customers, we can assume that the
variability in cumulative demand over a period of time
can be described by the normal distribution.
EOQ and Reorder Point Systems
Q
ROP
LT LT
Using the EOQ model, we developed a reorder point
(ROP) inventory management system:








In the EOQ model, demand is assumed to be constant
When demand is not constant, the reorder point
calculation should consider demand variability. If the
reorder point is only based on average demand,
stockouts will occur:
ROP with Variable Demand
Q
ROP
LT LT
Q
Q
DDLT*
*(average) Demand During Lead Time
Safety Stock
To avoid stockouts, the reorder point should include
additional inventory, safety stock, to reduce the
probability of a stockout.
ROP
LT LT
Safety Stock
Q
DDLT*
*(average) Demand During Lead Time
Safety Stock
DDLT

DDLT
DDLT SL
Z SS=
Probability of a
Stockout
Using the standard deviation of
the DDLT, we can set an
a safety stock level based on
the probability of a stockout
Safety Stock
Cumulative
Probability Z
0.50 0.0000
0.55 0.1257
0.60 0.2533
0.65 0.3853
0.70 0.5244
0.75 0.6745
0.80 0.8416
0.85 1.0364
0.90 1.2816
0.95 1.6449
0.98 2.0537
0.99 2.3263
0.995 2.5758
0.998 2.8782
Z
Cumulative
Probability
For a given service level (cumulative probability),
the safety stock is calculated as:

DDLT SL
Z SS=
Safety Stock Example
Suppose we have the following
weekly demand (consumption)
data for a product:
Week Demand
1 98
2 92
3 111
4 88
5 124
6 94
7 86
8 109
9 97
10 76
Average Demand 97.5
13.9
Standard Deviation
of Demand
If the lead time is one week,
then we have:


If we want a 95% service level,
then the safety stock should be:


So the reorder point should be:


So a ROP of 120 should be used
DDLT = 97.5
SS = (1.6449)(13.9) = 22.86
ROP = 97.5 + 22.86 = 120.36
Safety Stock using MAD
Many times, Safety Stock levels are calculated using the
Mean Absolute Deviation as a measure of variability
rather than the Standard Deviation. There are two
reasons for this:
Historical: Before calculators, the calculation of a
standard deviation was not a trivial task, while the
calculation of the Mean Absolute Deviation is fairly
simple to perform by hand
Robustness: The Mean Absolute Deviation measure
is not as easily affected by outlier points as it is using
the absolute value of the deviation rather than the
squared deviation
MAD Calculation
x
i
x
97.5
10
975
X = =
Week Demand
1 98 0.5
2 92 5.5
3 111 13.5
4 88 9.5
5 124 26.5
6 94 3.5
7 86 11.5
8 109 11.5
9 97 0.5
10 76 21.5
975 104
10.4
10
104
MAD = =
Standard Deviation Calculation
( )
2
x
i
x
97.5
10
975
X = =
Week Demand
( )
13.9
9
1744.5
1 n
2
X
i
X
SD
= =

=

1 98 0.25
2 92 30.25
3 111 182.25
4 88 90.25
5 124 702.25
6 94 12.25
7 86 132.25
8 109 132.25
9 97 0.25
10 76 462.25
975 1744.5
Safety Stock using MAD
The standard deviation can be
estimated from the MAD using:


As a result, we can define a
safety factor R which can be
used to determine the safety
stock based on the MAD and
the desired service level:
Cumulative
Probability R
0.50 0.0000
0.55 0.1571
0.60 0.3167
0.65 0.4817
0.70 0.6555
0.75 0.8431
0.80 1.0520
0.85 1.2955
0.90 1.6019
0.95 2.0561
0.98 2.5672
0.99 2.9079
0.995 3.2198
0.998 3.5977
SD = 1.25 MAD
SS = (R)(MAD)
Safety Stock Example Revisited
The following weekly demand
(consumption) data for the
product was:
The Demand During Lead Time is:


For a 95% service level,
the safety stock should be:


So the reorder point should be:


So a ROP of 119 should be used
(vs. 120 calculated using the SD)
DDLT = 97.5
SS = (2.0561)(10.4) = 21.38
ROP = 97.5 + 21.38 = 118.88
Week Demand
1 98
2 92
3 111
4 88
5 124
6 94
7 86
8 109
9 97
10 76
Average Demand 97.5
10.4 MAD
Demand Period vs. Lead Time Period
In the previous example, the demand period (the period
of time used to accumulate customer demand) was one
week, which was the same as the lead time.
Suppose the lead time was two weeks. Then the
variability of the demand for a two week period would be
greater than the MAD calculated from demand data
aggregated weekly.
We have assumed that customer demand is
independent, i.e. that the demand for the product comes
from a number of unrelated customers. In that case,
then we can use a theorem from statistics to determine
the appropriate variability of demand during lead time
when the demand period is different from the lead time
period
Demand Period vs. Lead Time Period
Suppose we have two
independent, normally
distributed random variables:
X: mean
X
, standard
deviation o
X

Y: mean
Y
, standard
deviation o
Y


Then the sum of these
variables, Z = X + Y has mean:

Z
=
X
+
Y

and standard deviation

2
Y
2
X
Z
+ =
X

DP
DP DP
LT
2 = = +
DP

DP

LT

DP = 1 week
LT = 2 weeks
Demand Period vs. Lead Time Period
Suppose that the demand period is 1 week (customer
demand is measured on a weekly basis) and the lead
time is two weeks. Then the standard deviation for the
lead time can be calculated as:
DP LT
LT
2
LT
2
LT
DP

2
1

=
+ = =
DP

LT

DP = 2 weeks
LT = 1 weeks
Demand Period vs. Lead Time Period
Suppose that the demand period is 2 weeks (customer
demand is accumulated in 2 week intervals) and the lead
time is one week. Then the standard deviation for the
lead time can be calculated as:
LT

In general terms, the standard deviation of the demand

for the lead time is:





where the lead time and demand period are measured in
the same time units (typically days). The demand period is
level of aggregation used for determining demand.

Safety Stock
DP
period demand
time lead
LT
=
So the safety stock level can be calculated as:



using the standard deviation of demand and:



using the MAD, where:
Safety Stock
Demand

SL
W Z SS=
MAD W R SS =
Period Demand
Time Lead
W=
Note that if the Demand Period does not equal the Lead
Time, then the DDLT is calculated as:
DDLT
|
|
.
|

\
|
|
|
.
|

\
|
= Demand W DDLT
Demand data for a material has been collected on a weekly
basis for 6 months. Demand appears level, with:

Mean: 270 units/week
Standard deviation: 40 units/week

The lead time is 10 days. Calculate the safety stock
required for a 99% customer service level.
Safety Stock: Example 1
Safety Stock: Example 1
The formula for safety stock using the standard deviation
is:


so for this example we have:
Demand

SL
W Z SS=
111 111.2
40
7
10
2.3263 SS
~ =
=
Demand data for product has been collected on a weekly
basis with the following results:

Mean: 109 units/week
MAD: 20 units/week

The lead time is 4 days. Calculate the safety stock
required for a 95% customer service level.
Safety Stock: Example 2
Safety Stock: Example 2
The formula for safety stock using the standard deviation
is:


so for this example we have:
31 31.08
20
7
4
2.0561 SS
~ =
=
MAD W R SS =
Demand Period and Lead Time in SAP
Demand period is set by the Period Indicator on the Forecasting View
of the Material Master

The applicable periods are:
M Monthly
W Weekly
T Daily
Demand Period and Lead Time in SAP
In-house production is used for
Lead Time for products made in-house
Plnd delivery time + GR processing time +
Purchasing proc. time is used for
Lead Time for externally procured materials
Exposure to Stockout
Stockouts usually occur when stock gets lowfor
example, during the lead time period before a new order
arrives:
Periods of maximum exposure to stockout
LT LT LT
Exposure to Stockout
The more frequently we order, the more chances there
are of stocking out.

LT LT LT
LT
Twice as many
opportunities for
stockout
LT LT LT LT LT LT
Exposure to Stockout
To fully evaluate the customer service level, we should
calculate the customer service level on an annual basis:







where D is annual demand and Q is the order quantity.
|
|
|
|
.
|

\
|
|
|
.
|

\
|
=
Q
D
Order
Annual
SL SL
Exposure to Stockout
For example, if we used a service level of 95% in
calculating the safety stock, the annual demand D is
12,000 units and the order quantity Q is 800 units, then
we have:






So there is only a 46.3% chance of going a year without
a stockout
463 . 0
15
95 . 0 95 . 0 = = =
=
|
|
.
|

\
|
|
|
|
|
.
|

\
|
|
|
.
|

\
|
|
|
.
|

\
|
|
|
|
|
.
|

\
|
800
12,000
Order
Q
D
Annual
SL SL
Demand Patterns
Day 1 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Day 9
Week 1 Week 2
Regular Demand
Day 1 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Day 9
Week 1
Week 2
Sparse Demand
Demand Patterns
In developing the safety stock calculations, it was
assumed that demand was generated from a large
number of independent sources, and
The individual demands are aggregated over a time
period sufficiently long so that there are a number of
individual demands contributing to each period demand.
If these conditions are not met, then the safety stock
values may not perform as expected.
Demand Patterns
If demand is sparse, then a more detailed approach to
inventory planning that considers the expected time
between orders as well as the expected order quantity
Time
Quantity
Expected time
between orders
Expected
order
quantity