Dynamic Safety Stock Calculation

World Academy of Science, Engineering and Technology
International Journal of Mechanical, Aerospace, Industrial, Mechatronic and Manufacturing Engineering Vol:7, No:10, 2013
Dynamic Safety-Stock Calculation

Julian Becker, Wiebke Hartmann, Sebastian Bertsch, Johannes Nywlt, Matthias Schmidt
AbstractIn order to ensure a high service level industrial

enterprises have to maintain safety-stock that directly influences the
economic efficiency at the same time. This paper analyses
established mathematical methods to calculate safety-stock.
Therefore, the performance measured in stock and service level is
appraised and the limits of several methods are depicted. Afterwards,
a new dynamic approach is presented to gain an extensive method to
calculate safety-stock that also takes the knowledge of future
volatility into account.
International Science Index Vol:7, No:10, 2013 waset.org/Publication/17180
KeywordsInventory dimensioning,
planning, safety-stock calculation.
material
requirement
I. INTRODUCTION
HE logistic performance of industrial enterprises has

become more and more important in the last two decades.
Recent surveys show that delivery time and delivery
reliability in particular are the main decision criteria besides
price and quality of the product [1], [2]. However, in times of
increasingly dynamic markets and growing interdependences
of international enterprises a high service level can often just
to be ensured by stocking sufficient inventories. In these cases
the maintained safety-stock depend on volatile factors such as
the demand, replenishment time and replenishment quantities
[3].
On the one hand, industrial enterprises always try to
improve their service level in order to reach customer
requirements. As a result high safety-stock is required. On the
other hand, inventory generates capital commitment that
influences the economic efficiency. These competing logistic
objectives high delivery capability and low stock level
describe the so called dilemma of inventory management [2],
[4]. Within this area of conflict many different methods were
developed to dimension inventory.
Most of the existing approaches to calculate safety-stock
based on statistical parameters (e.g. standard deviation of the
demand, mean of the demand rate) and therefore just take
historical data into consideration [5]-[7]. Some approaches
deal with dynamic inventory control by applying statistical
equations that are embedded in rolling planning [7]-[9].
Hence, they only can be characterized as quasi-dynamic [10].
J. Becker is with the Institute of Production Systems and Logistics, Leibniz

University of Hannover, Germany (phone: 0049-511-762-19812; fax: 0049511-762- 3814; e-mail: becker@ifa.uni-hannover.de).
W. Hartmann was with the Institute of Production Systems and Logistics,
Leibniz University of Hannover, Germany, and works as a production
engineer at Volkswagen AG in Hannover.
S. Bertsch, J. Nywlt and M. Schmidt are with the Institute of Production
Systems and Logistics, Leibniz University of Hannover, Germany.
International Scholarly and Scientific Research & Innovation 7(10) 2013
The goal of the paper is to introduce a dynamic approach to

calculate safety-stock that is easy to implement for
practitioners, as well.
II. COMPARISON OF SAFETY-STOCK CALCULATION METHODS
A. Existing Approaches for Dimensioning Safety-Stock
In publications many different mathematical- stochastic
methods for determining safety-stock are described and
discussed. Previous simulation studies conducted at the IFA
compared and benchmarked the most common basic
approaches in industrial practice [11]. In the following, we
will briefly characterize the approaches that perform best
under the simulated conditions.
One of the frequently referred standard formula of safetystock calculation (see a. o. [5], [12]) multiplies a safety factor
which depends on the service level based on a normal
distributed demand with the standard deviation of the demand
during the replenishment time:
SSL = SF ( SL ) D
(1)
where SSL: safety-stock level [units]; SF: safety factor

depending on service level [-]; SL: service level [-]; D:
standard deviation of demand [units/SCD]; SCD: shop
calendar day.
Where Method (1) the safety-stock is determined as a
function of the service level (SL), which in turn is determined
as the percentage of the total demands served punctually. The
safety factor for the service level results via the inverse of the
standard normal distribution [5], [13].
ALICKE provides a calculation rule for safety-stock that
takes the replenishment time and a forecast error, derived
from forecast data as a stochastic component, into
consideration [5]:
SSL = SF ( SL) F TRP
(2)
where F: standard deviation of the forecast error for the

demand during TRP [units/SCD]. The standard deviation of
the prognosis error is calculated via historical data from the
mean squared deviation of the forecasted demand from the
actual. Thus, Method (2) applies independent of a specific
statistical distribution of the demand [5].
The following concepts take up the preceding concepts and
extend them with a stochastic replenishment time [5], [6]:
2
SSL = SF ( SL) TRP D2 + D 2 TRP
985
(3)

(1 SL) QRP
2
2
2
SSL = SF1
TRP N + D TRP
TRP D
QRP > TRP D
(5)
where QRP: replenishment quantity [units].

All of the formulas described thus far have been developed
unless otherwise noted under the condition of normally
distributed parameters. The following Method (6) is a function
for determining the safety-stock for a target service level that
also takes into account extreme values beyond the mean and
standard deviation and that refrains from a specific statistical
distribution (for a detailed derivation cf. [15]):
SSL = LSL0 SL2 1 + SSL100% C 1 (1 SL)C

with LSL0 =
SSL100% =
QRP
and
2
(DV
(6)
) + ((D
2
+
d , max D
) TRP)2 + (DVQRP
, max )
max D
where LSL0: lot stock level [units]; C: C-Norm parameter [-];

DVd,max+: max. positive deviation from due date [SCD]; Dmax:
maximum demand per period [units/SCD]; DVQRP,max-: max.
negative deviation in replenishment quantity [units].
Method (6) is based on calculating a safety-stock for a
target service level of 100% (SSL100%). This safety-stock can
be adjusted via the C-Norm Function to a service level lower
than 100% [15].
B. Underlying Structure of Simulation Study
The so called inventory replenishment policies provide the
procurement quantities and time-points in order to control the
stock in a storage echelon. According to approaches mainly
used in industrial practice, the s, q- policy serves as a basis for
the simulation of the stock [16]. When the order point s is
reached an order of quantity q is triggered that restock the
store after the replenishment time. For this purpose, the
C. Results of Simulation Study

The conducted simulation study illustrates that there is no
superior approach among the various methods of safety stock
calculation. Each of the presented methods has its respective
strength depending on the particular conditions [11]. Fig. 1
depicts an overview of the gained results of the simulation
study.
high
Gudehus amplifies Method (4) with an adaptive service

level factor that takes into consideration that only disruptions
during the replenishment cycle can lead to a lack of delivery
capability (Method 5). If a mean delivery capability is to be
attained over the period between the input of two orders, a
value that is smaller than the required delivery capability
suffices for the service level during the critical replenishment
time phase [14]:
2, 6*
2, 6*
3, 4, 5
medium
(4)
2, 6*
2, 6*
3, 4, 5
low
2
SSL = SF ( SL) Var (U ) + TRP D2 + D 2 TRP
Economic Order Quantity approach according to Harris [17] is

used for the optimal order quantity [11].
To simulate the necessity of safety-stock there are
deviations from the planned values implemented. First,
differences in the required quantities and the time intervals
between the two demands constitute demand fluctuations from
the output- side. Second, deviations are schedule or quantity
related from the input- side [11].
For this simulation three different continual distribution
forms are taken as a basis for the distribution of the demand
since they can be seen in industrial practice. The normal
distribution is suitable for displaying a large number of
random variables [18]. The gamma distribution is particularly
suited for depicting sporadic demands and does not take any
negative value into account [6], [19]. The log-normal
distribution is suitable for representing skewed distribution
and also does not take any negative values into account [11].
Since three divisions are used for the classification of the
demand, three classes are also used for the replenishment
time. Though, these classes have stricter restrictions with
regards to possible distribution parameters. First, an almost
constant replenishment time is illustrated by a mean value of 3
SCD and a mean standard deviation of 1 SCD. Second, for an
irregular replenishment time a mean of 5 SCD and a mean
standard deviation of 3 SCD are assumed. Third, an uncertain
replenishment time is characterized by a mean value of 20
SCD and a standard deviation of 16 SCD. Deviations from the
planned supply quantities are not relevant from practical
perspective. Thus, they are not taken into account [11].
Based on the total number of methods and parameter
variations a total number of 3645 simulation runs were
conducted. The resulting logistic performance (target service
level 95%) and costs for a method are used as assessment
factors [11].
Variance of
Replenishment
Time
where D: mean demand per period [units/SCD]; TRP:

standard deviation of replenishment time [SCD].
Method (4) by Herrmann also takes into account the so
called undershoot that refers to the problem in which the
stock might have already fallen below the order point
immediately before an order is triggered [13]:
*for a normal distribution

of the demand
1
low
medium
high
Variance of Demand
Fig. 1 Preferable application area Cluster of the calculation methods
It shows the application areas divided by the main

influencing factors (variation of demand and replenishment
986
deviations; f(t): forecast information on a daily basis [units];

DVd,max+: max. pos. deviation from due date [SCD]. Fig. 2
depicts both the static and dynamic safety stock level for duedate fluctuations. The spotted area illustrates the static
calculation whereas the sum of the spotted and the striped area
represent the dynamic approach.
demand/u
time) and the approach which should be preferably used for

each area.
Method (1) works properly as long as the replenishment
time only has little variations. Although it shows high
performance especially for large lot sizes with low variance of
the replenishment time, it should not be used if the variance of
the replenishment time is medium or high. Methods (2) and
(6) provide excellent results concerning service and stock
level for a medium or high variance of the replenishment time
and a low or medium variance of the demand. For the cases of
a high variance of the replenishment time and a high variance
of the demand Methods (3), (4) or (5) achieve the best results
[11].
Obviously, Method (6) should be preferred in four of nine
cases. Therefore, it presents the basis for further research to
develop a dynamic method to calculate safety-stock. Because
of the presented lack of strengths in cases with a high variance
of the demand the focus lies on an improvement of the
approach for such circumstances.
f(t)
D
t1
t2 DV+d,max t3 time/t
TRP
Fig. 2 Safety-stock level for due-date deviations
III. DYNAMIC SAFETY- INVENTORY CALCULATION

With Method (6) LUTZ provides a calculation rule for the
safety stock that performs better than other concepts.
However, this static approach based on the statistical analysis
of date and does not take any knowledge of future volatility
into account. Therefore, the paper presents a future- oriented
dynamic approach to dimensioning stock.
In order to generate such a dynamic approach, forecast
information is assumed to calculate an accurate demand for
each shop calendar day. Previous research conducted at the
IFA shows that cubic spline function are well suited to
transform a monthly given forecast to a corresponding
forecast quantity for each shop calendar day [10].
Moreover, since deviations from the planned supply
quantities can be counter-balanced by pulling forward the next
delivery quantity deviations are not taken into consideration.
Thus, two deviation types remain which have to be
considered, the due- date deviations and the demand
fluctuations.
B. Dynamic Safety-Stock Level for Demand Fluctuation

Second, demand fluctuations come into play when the
actual demand (Dactual) during the replenishment time is higher
than the forecasted demand. In order to avoid lasting effects
on the delivery capability a safety-stock level for demand
fluctuations need to be maintained. Accordingly, a forecast
error in regards to quantity effects should be utilized. This
factor multiplied with the forecasted demand during the
replenishment time can be determined as the required safety
stock level.
t2
SSL2,min = f e f (t )dt
where SSL2,min: Safety stock level for demand fluctuations

[unit]; fe: forecast error of the demand [%].
A. Dynamic Safety-Stock Level for Due Date Deviations

First, due-date deviations are balanced in Method (6) by the
product of the maximal positive deviation from due date and
the mean demand. Is an accurate daily forecast given the
integral of demand with respect to time can be used to
calculate the required dynamic safety-stock. The lower limit
of integration is the end of the replenishment time (t2) because
at this point the delay of the demand begins. The upper limit
of integration is represented by the maximal positive deviation
from due date (t3).
SSL1,min =
Dmax
Dactual
f(t)
D
t1
TRP
t2
time/t
Fig. 3 Safety stock level for demand fluctuations
t3
f (t )dt
(8)
t1
demand/u

(7)
t2
where SSL1,min: Dynamic safety-stock level for due-date
Fig. 3 illustrates the static calculation approach (spotted

area) and the dynamic calculation approach (striped area) for
demand fluctuations.
987

C. Dynamic Safety-Stock Level for a Target Service Level

According to the static method by Lutz both deviation types
can occur individually or in combination. Moreover, a
stochastic independence of due- date deviations and demand
fluctuations are assumed [15]. Hence, both formulas combined
constitute a safety-stock level for a target service level of
100% (SSL100%,dyn):
SSL100%,dyn =
(SSL1,min )2 + (SSL2,min )2
(9)
REFERENCES
[1]
[2]
[3]
[4]
[5]
After substituting (6), (7) and (8):
[6]
[7]
SSL = LSL 0 SL2 1 + SSL100%, dyn C 1 (1 SL )C
where
LSL 0 =
QRP
2
(10)
and
2
SSL100%, dyn
t2
t3
= f (t ) dt + f e f (t ) dt
t
t1
2
[8]
[9]
[10]
[11]
Finally, the safety-stock (10) can be adjusted via the CNorm Function to a target service level lower than 100%.
IV. CONCLUSION
The paper introduces a new dynamic approach to calculate
safety-stock. Based on previous simulation studies conducted
by the IFA an appropriate static approach was selected to
serve as the basis for the new method. Because the new
approach includes forecast information regarding to quantity it
should performer better than the presented static methods. In
particular, a lower level of stock in times with less demand
and a higher service level in boom times are expected.
However, further research is required to verify the
advantages of the presented approach. Moreover, a suitable
forecast error is to be examined. In order to consider high
variations of the demand different kinds of calculations are
possible, e.g. the absolute deviation, mean square deviation,
variance of the demand, to name a few. All calculation should
be based on the replenishment time, because only in this
period deviations can cause lasting effects on the delivery
capability.
After verifying the advantages of the presented dynamic
method to calculate safety-stock and its final devising a
demonstrator program is to be developed in order to bring the
new approach into industrial practice.
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
ACKNOWLEDGMENT
This paper presents results of the project Developing a
Method for Dynamic, Future-Oriented Dimensioning of
Stock, funded by the German Research Foundation (DFG)
and currently being conducted at the Institute of Production
Systems and Logistics.
988
M. A. Karim, P. Samaranayake, A. J. R.Smith, and S. K. Halgamugec,

An on-time delivery improvement model for manufacturing
organisations, International Journal of Production Research, vol. 48, no.
8, 2010, pp. 23732394.
K. K. B. Hon, 2005, Performance and Evaluation of Manufacturing
Systems, Annals of the CIRP 54/2: 139-154.
A. J. Ruiz-Torres, F. Mahmoodi, 2010, Safety stock determination based
on parametric lead time and demand information, International Journal
of Production Research, 48/10: 28412857.
W. J. Hopp and M. L. Spearman, Factory physics, 3rd ed., Boston et al.:
McGraw-Hill/Irwin, 2009.
K. Alicke, 2005, Planung und Betrieb von Logistiknetzwerken, 2nd ed.,
Springer
S. Axster, 2006, Inventory Control, 2nd ed., Springer.
M. Z. Babai, A. A. Syntetos, Y. Dallery, K. Nikolopoulos, 2009,
Dynamic re-order point inventory control with lead-time uncertainty:
analysis and empirical investigation, International Journal of Production
Research, 47/9: 2461-2483.
J. J. Kanet, M. F. Gorman, M. Stlein, 2010, Dynamic planned safety
stocks in supply networks, International Journal of Production Research,
48/22: 68596880.
T. Gudehus, H. Kotzab, 2008, Comprehensive Logistics, Springer.
W. Hartmann, M. Schmidt, P. Nyhuis, Dynamic Inventory
Dimensioning, 2011 African Conference on Software Engineering and
Applied Computing, 18.-23. September 2011, Cape Town.
M. Schmidt, W. Hartmann, P. Nyhuis, Simulation Bbased comparison
of safety-stock calculation methods, CIRP Annals - Manufacturing
Technology 61 (1), pp. 403406.
K. Inderfurth, T. Jensen, 2008, Lagerbestandsmanagement, in: Arnold,
D. (Ed.), Handbuch Logistik, Springer, 153-167.
F. Herrmann, 2011, Operative Planung in IT-Systemen fr die
Produktionsplanung und-steuerung, Vieweg + Teubner.
T. Gudehus, 2006, Dynamische Disposition, 2nd ed., Springer.
S. Lutz, H. Ldding,H.-P. Wiendahl, 2003, Logistics-oriented inventory
analysis, International Journal of Production Economics, 85: 217231.
H. Tempelmeier, L. Fischer, 2010, Approximation of the probability
distribution of the customer waiting time under an (r, s, q) inventory
policy in discrete time, International Journal of Production Research,
48/21: 62756291.
F. W. Harris, 1913, How Many Parts to Make at Once, The Magazine of
Management, 10/2: 135-136.
B. Stausberg, 1992, Entwicklung eines Verfahrens zur Berechnung
notwendiger Sicherheitsbestnde unter Bercksichtigung realer
Verteilungen fr Lagerabgang und Wiederbeschaffungszeit mit Hilfe
von Konfidenzbereichen, Schlubericht zum DFG-Forschungsvorhaben
Ha 531/65-1, FIR e.V., RWTH Aachen.
H. Tempelmeier, 2011, Inventory management in supply networks, 2nd
ed., Books on Demand.

Dynamic Safety Stock Calculation

Uploaded by

Copyright:

Available Formats

Dynamic Safety Stock Calculation

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Dynamic Safety Stock Calculation

Uploaded by

Copyright:

Available Formats

World Academy of Science, Engineering and Technology