Project Study on Supply Chain Planning in German Automotive Industry

Description
Mass customization that aims at offering customized products in a high variety but for still low prices and within short delivery times gains increasing importance in various branches of business and, in the meantime, also captivates the automotive industry.

DOI: 10.1007/s00291-004-0168-4
OR Spectrum (2004) 26: 447–470
c Springer-Verlag 2004
Supply chain planning
in the German automotive industry
Herbert Meyr
Institute of Transport Economics and Logistics, Vienna University of Economics
and Business Administration, Nordbergstraße 15, 1090 Wien, Austria
(e-mail: [email protected])
Abstract. Following the evolution in the computer industry, quite a lot of car
manufacturers currently intend to move from a built-to-stock oriented production
of standardized cars towards a customized built-to-order (BTO) production. In the
premiumsegment of Germany’s automotive industry, the share of customized BTO
cars traditionally is comparatively high. Nevertheless, German car manufacturers
have spent a lot of efforts in recent years to further increase this share in order
to realize short delivery times, high delivery reliability and a fast responsiveness.
Surprisingly, comprehensive overviews of the short- and mid-term planning land-
scape of car manufacturers cannot be found in the scienti?c literature. Thus, the
?rst part of the paper discusses supply chain planning, as traditionally established
in the premium segment of the German automotive industry, and reviews methods
of Operations Research (OR) that are able to support the various planning tasks
involved. In the second part, the major change in strategy, currently to be observed
in the German automotive industry, is brie?y summarized in order to derive its
impacts for the planning system and for the respective planning methods. In this
way, challenges for a future application of OR methods in the automotive industry
can be identi?ed.
Keywords: Supply chain planning – Operations research – Automotive industry
1 Introduction
Mass customization [42] that aims at offering customized products in a high variety
but for still low prices and within short delivery times gains increasing importance
in various branches of business and, in the meantime, also captivates the automotive
industry. The BMW Group, for example, spent $ 55 million on its new European
online-orderingsystem[24] tocut order-to-deliverytimes by20days onthe average.
At the same time, BMW offers up to 10
32
variants (at least theoretically), several
448 H. Meyr
thousands of them actually being demanded [51, p. 42]. Other manufacturers also
declared their intention to decrease order-to-delivery times from an average of 40
days to about 15 days [22] and try to make the transition from “build-to-stock“
(BTS) to “build-to-order” (BTO) that has successfully been demonstrated by the
computer industry, and ?rst and foremost by its paragon Dell.
The transition to BTO in the computer industry caused a reorganization of
planning processes and led to an increased use of “Advanced Planning Systems”
(APS, [29]), i.e. of computer-based decision support systems, which – at least partly
– rely on sophisticated methods of Operations Research (OR). Thus the questions
arise, whether and how the transition of the automotive industry changes their
respective planning tasks and planning processes, and to what extent planning and
OR methods are and will be affected. Since mutual interrelations are particularly
important for operational planning tasks, the discussion will concentrate on mid-
and short-term supply chain planning, and here especially focusing on the car
manufacturers’ point of view. But before discussing changes it has to be shown
what the planning landscape of automotive industries traditionally looks like. There
are, of course, discussions of various individual planning tasks (see Sect. 3) and
some overviews of the order-to-delivery process (see e.g. [23, 51]). However, to
the author’s knowledge, in scienti?c literature no comprehensive overviews of the
short- and mid-term planning landscape of car manufacturers can be found.
Due to this lack of literature and since common scienti?c approaches like ques-
tionnaires and structured interviews did not seem to be very promising because
quite a lot of con?dence is needed to get such a sensitive information, the following
characterization of the planning system of car manufacturers mainly builds on var-
ious joint projects with German car manufacturers and communication with their
responsible planners and with employees of automotive consultancies. In order to
verify the conclusions drawn, a working paper has been written, sent to skilled
people in these companies and they have been asked for statements about its va-
lidity. The results of this process are presented in the following. To sum up, the
contribution of this paper is
– ?rst, that the planning systems of German car manufacturers are analyzed,
described and thus made available to the academic literature,
– secondly, that OR methods suitable for planning within the automotive indus-
tries are reviewed, categorized with respect to the planning tasks of (German)
car manufacturers and that insuf?ciently supported planning tasks are disclosed,
and
– thirdly, that the challenges of the managerial changes from BTS to BTO are
outlined that arise for the planning tasks, the planning systems and for the OR
models/methods involved.
Due to this broad scope of the paper, a review of OR methods – even though
restricted to short- and mid-term planning – cannot be comprehensive. This paper
rather intends to give an idea where (i.e. at which subsection within the overall
planning system of a car manufacturer) OR methods already contribute or may
contribute in the future.
Supply chain planning in the German automotive industry 449
Long-term, strategic planning provides potentials, which mid-term planning
has to further develop and short-term planning has to implement. Of course, also
long-term planning tasks are supported by OR methods. Concerning the product
design, for example, the optimal commonality of automotive components (e.g. wire
harnesses) is determined [53] or the impact of product variety on the performance of
mixed model assembly lines is analyzed [6, 16]. It is even worth to include assembly
sequencing issues into product design decisions [52]. Analytical and simulation
models provide general hints (“chaining strategies’’) how to assign products to
manufacturing plants so that high process ?exibility is achieved for both single stage
[28] and multi stage [5, 19] automotive supply chains. Linear Programming (LP) or
Mixed-Integer linear Programming (MIP) models are, for instance, used by APS to
design the inbound system of assembly plants [20] or car distribution networks [2]
(see also [36] without use of APS). Concerning the inside of assembly plants, the
planning of the physical layout and of buffer sizes of assembly shops, in general,
and of body shops [41, 49, p. 20 ff. & 73 ff.], in particular, can be supported by
simulative, analytical and combinatorial optimization methods. A comprehensive
overview of OR methods for the well-known assembly line balancing, which is a
rather strategic than mid-term task in the automotive industry, is given by Scholl
[47]. Arecent survey of heuristic methods for cost-oriented assembly line balancing
can be found in [1].
In order to understand why automotive planning systems are organized the way
they are, Section 2 describes the characteristics of automotive supply chains. These
vary substantially for car manufacturers in different parts of the world (North Amer-
ica, Japan/Korea, Europe/Germany), operating on different market segments. This
paper mainly concentrates on premiumbrands (like BMW, Mercedes, Audi) but not
luxury cars (like Rolls-Royce, Maybach, Bentley) and on the German automotive
industry. Nevertheless, quite a lot of the statements and ?ndings of this paper can be
transferred to car manufacturers in other parts of the world tackling related product
segments because (even though beginning with different starting points) many of
them similarly intend to change to a BTO production. Section 3 then presents the
traditional short- and mid-term planning system and – after introducing the respec-
tive planning tasks – points to appropriate planning and OR methods. After brie?y
summarizing the measures to improve BTOassembly currently being implemented
in the German automotive industry (Sect. 4), their impact on the planning system
is discussed in Section 5. Thus changed requirements for planning methods can be
derived and challenges for future research can ?nally be identi?ed (Sect. 6).
2 Automotive supply chains
Cars are sold to ?nal customers either directly via sales subsidiaries of the car man-
ufacturer or indirectly via legally separate retailers. The bill-of-material (BOM) is
strictly convergent, i.e. assembly processes are dominant. Cars often are thought
to be standard products. However, in the premium segment of this line of busi-
ness, there is a high degree of customization. This allows the customer to specify
obligatory features like the color of the car and type of upholstery or optional
features like air conditioning or a navigation system, to name only a few. In the
450 H. Meyr
following both obligatory and optional features are just referred to as “options”.
A car manufacturer usually offers several types of cars (e.g. the E-class or C-class
of DaimlerChrysler), which again differ in several body-in-white variants (coupe,
convertibles, etc.). Not every customer needs his car immediately. According to
Stautner [51, p. 38] the order lead time desired by a ?nal customer is normally
distributed with a mean value of 4 to 6 weeks.
The sales organization and distribution network of a car manufacturer have a
divergent structure, which comprises several stages like the central sales department
of the manufacturer, sales persons responsible for different world regions (also at the
headquarter), sales companies in different countries or local areas and a rather high
number of further retailers and sales subsidiaries. This type of customized premium
cars can only be assembled “to order”, i.e. there has to be an “order” available –
either by the ?nal customer, a retailer or a sales department of the manufacturer
– that speci?es the options of the car. Current SCM initiatives in the automotive
industry try to increase the share of ?nal customers’ orders and to decrease the
share of retailers’ and sales departments’ orders (see Sect. 4.1).
Commonly, manufacturer and retailers communicate in two types of interac-
tion rounds: In the ?rst one, a retailer sends his mid-term requests for cars to the
manufacturer. Both “negotiate” the number of cars (so-called “quota”) the retailer
will get during the next year. Usually, this “negotiation” process is clearly domi-
nated by the manufacturer so that – due to the preferences of the manufacturer –
the agreed quota may be less or even higher than the original requests. Since these
quotas are, for example, de?ned for the next year on a monthly basis, only body-
in-white variants and the type of engines (referred to as “models” in the following)
are considered, but the options are not speci?ed at this point in time.
In a second round, about three to ?ve weeks before planned production, the
retailer has to specify the options for all cars of his quota, which are due and have
not been assigned to ?nal customer orders that had arrived in the meantime. From
a retailer’s point of view, these cars are “built to stock” (BTS-cars), based on a sort
of forecasting process for options. From the manufacturer’s point of view, an order
of the retailer exists, thus justifying the term “built to order”.
Figure 1 illustrates the different states of demand information that are implied
by these two interaction rounds. The curve (I) shows the cumulated share of fully
speci?ed orders of ?nal customers with respect to the overall number of orders
(incl. forecasts) considered by planning. It can be computed by calculating the
distribution function of the order lead times, that are desired by ?nal customers
(see [51, p. 38]). This distribution function is drawn backward in time, starting with
the delivery of cars to ?nal customers.
For the section above the curve, no information about the preferences of ?nal
customers is available. This lack of information has to be replaced with forecasts.
Concerning the options of BTS-cars, this is done by retailers with a lead time of
3–5 weeks before production (II). Beforehand, with a lead time of one year at a
maximum, only the retailers’ requests for models are known (III), which also are
the result of a forecasting process of the retailer. For yet earlier planning tasks, a
car manufacturer has to rely on his own pure forecasts for models (IV).
Supply chain planning in the German automotive industry 451
pure
forecast
for models
(IV)
requests
for models
by retailers
(III)
sending requests
for models
options
specified
by
retailers
(II)
specification of options
by retailers
share of
final customers‘ orders
100 %
time
50 %
delivery to
final customer
options
specified by
final customer
(I)
„built-to-stock“
of retailers
Fig. 1. Demand information available to a car manufacturer
The production systemin a car assembly plant usually comprises the four stages
pressing of metal or aluminiumsheets, welding the body-in-white fromthe moulded
sheets in the body shop, painting it in the paint shop and ?nal assembly, where
painted body, engine, transmission and the further equipment are brought together
or built in. For the ?nal assembly one or several production lines are used. A
productionline consists of quite a lot of seriallyarrangedassemblystations, between
which cars are conveyed with a ?xed belt rate. The processing time at an assembly
station depends on the option chosen for the car to be assembled. Therefore, the
overall utilization of a station is determined by the sequence in which cars/orders
are assembled on a line (the so-called “model mix”). If too many cars requiring the
same options are following one another, some of the stations may be overloaded
whereas others are underloaded. Thus a “balanced” model mix has to be found,
almost equally utilizing the various stations of an assembly line.
Because of the convergent BOM and ten thousands of components to be pur-
chased, a procurement network with several hundreds direct and an enormous num-
ber of indirect suppliers has to be coordinated. For the delivery of incoming goods
normally several transport modes are applied. Voluminous and expensive compo-
nents are – as far as possible – delivered “just in time” (JIT) at the day of assembly,
partlyevendirectlytothe assemblyline andthus arrangedinthe sequence of planned
assembly (“sequence-in-line supply”, SILS). The remaining incoming goods are
collected by regional carriers, consolidated and brought to intermediate warehouses
of the car manufacturers, which are close to their assembly sites.
The structure of an automotive supply chain is characterized by a convergent
?ow of material upstream of the assembly plants of the car manufacturer and a
divergent ?ow of ?nished cars downstream. An automotive SC is dif?cult to co-
ordinate, because not only production capacity and manpower may turn out to be
bottlenecks, but also incoming goods. For reasons of ?exibility, high-volume car
models can sometimes be produced at several assembly plants. Because of these
constraints, the promised delivery dates cannot always satisfy the expectations of
452 H. Meyr
the ?nal customers. Furthermore, ?nal customers need a reliable delivery date be-
cause important further activities (like selling the old car, making money available)
have to be synchronized with the arrival of the new car. Thus, “order promising”
not only has to aim at setting a delivery date close to the customer’s wishes, but
also at promising a reliable date considering as many of the above constraints as
possible.
Besides a signi?cant intra-organizational information ?ow between different
planning units/departments of the car manufacturer itself (as will be discussed in
Sect. 3), there is also a vital inter-organizational exchange of information between
the different members of the SC. Commonly, car manufacturers prepare a rough
mid-term supply plan of the next year for their (?rst-tier) suppliers in order to draw
early attention to potential capacity bottlenecks. In the short term, daily supply
plans are sent to the suppliers. These include binding orders for the next day, but
also quite reliable “forecasts” for the next days/weeks and even rough forecasts for
the next months.
3 Traditional planning processes
To cope with the various planning tasks of automotive supply chains, quite a lot of
planning units/departments have to be involved. These planning tasks and the re-
spective decisions can be assigned to several planning levels (e.g. strategic, tactical,
operational) comprising different planning horizons (e.g. long-, mid-, short-term).
Depending on the planning horizon and the lead time necessary to make a certain
decision – different phases of the time axis of Figure 1 are relevant and thus a
different state of knowledge about actual customer demand is available. Therefore,
from a manufacturer’s point of view, one may distinguish between forecast-driven
long- and mid-term planning (phases (III) and (IV) of Fig. 1) and order-driven
short–term planning (phases (I) and (II)). In Section 3.1 forecast-driven mid-term
planning tasks and their information ?ows, which are more or less common for
the German automotive industry, will be discussed ?rst. Order-driven planning will
then be the concern of Section 3.2. Within both sections organizational issues are
left aside. This will be covered in Section 3.3.
3.1 Forecast–driven planning
Figure 2 summarizes the forecast-driven planning activities. Planning tasks are
marked by rectangles, arcs illustrate the information ?ows in between. From the
bottom to the top, the level of aggregation and the planning horizon are increasing,
the frequency of planning is decreasing, however. The planning tasks are roughly
assigned to the logistical functions procurement, production, distribution and sales,
again. Of course, not all of the mid-term planning tasks of a car manufacturer will
be discussed. Only the most important ones which show a close interrelation have
been selected.
The annual budget planning determines the overall monetary budgets of the
car manufacturer’s departments and assembly plants for the next year. For this,
Supply chain planning in the German automotive industry 453
retailers
suppliers
plants
demand
planning MRP
allocation
planning
take rates
forecasts,
take rates
requests
for models
requests of
regions
volume goals,
earning goals,
annual
working time
production
plans
(weekly)
production
plans,
overtime
(weekly)
aggregate quotas
detailed quotas
supply
plans
procurement production distribution sales
budget planning
pro-
duction
sales
master production pl.
pro-
duction
sales
(monthly)
production
& sales
plans
(monthly)
aggregate
quotas
Fig. 2. Overview of (mainly) forecast-driven planning
production plans for the respective plants and the sales plans for the respective
sales regions have to be calculated, too. This is done once per year, for the next
year, by deciding about production and sales quantities of car models (per plant
and world region, for example) on a monthly basis. The overall yearly quantities
can be considered as “volume goals” of the next year for both sales and production.
From these, the expected production costs and earnings can be derived (“earning
goals”).
A further result of the annual budget planning is the usage or reservation of
additional capacities, as far as these can still be in?uenced on a mid-term basis.
Because of the long lead times (e.g. two years or more to install an assembly line
or a plant), usually capacities of production resources are adapted to customer
demand in the long term and thus are a concern of strategic planning. However,
agreements about the extent and ?exibility of the yearly working time, for example,
are also a task of mid-termplanning. Alot of further constraints have to be respected
like potential bottlenecks of suppliers, model mix restrictions (capacities of crucial
options, minimum utilization) and upper or lower bounds of the sales in certain
markets. Lower bounds, for example, result from strategic directives about the
presence in important markets, upper bounds may be due to marketing analyses
about ?nal customer demand.
Input data for the annual budget planning mainly are forecasts for ?nal cus-
tomers’ demand (see also Fig. 1), which result fromthe demand planning. These are
made on basis of historical sales data of the few already known and fully speci?ed
orders from ?nal customers (e.g. car rentals), of the retailers’ annual requests for
models, of the sales companies’ decentral knowledge about the local preferences of
their customers (“requests of regions”) and on basis of information about marketing
capabilities to in?uence ?nal customer demand.
454 H. Meyr
Since budget planning has to decide about car models on the one hand and
to anticipate potential bottlenecks of suppliers on the other hand, the component
demand needs to be estimated, too. One way to do this is to forecast take rates
directly, i.e. to calculate the probability that a certain option or even component is
demanded (in a speci?c customer region) and to multiply it with the total number
of car models planned (for this region).
The task master production planning is similar to the annual budget planning.
Again, production and sales plans have to be determined and coordinated. However,
both nowrequire a higher level of detail (e.g. weekly instead of monthly quantities)
and they are not used to derive budget goals any further. The planning horizon
of a monthly rolling horizon planning varies between three months and one year.
Nevertheless, only the weekly quantities of the ?rst month or the ?rst two months
(depending on the lead times of planning) are put into practice.
Input data (see Fig. 2) are the already mentioned sales forecasts for models and
forecasts for take rates. Because of the high share of ?nal customers’ orders, that
is available for this shorter planning period (see Fig. 1), these monthly forecasts
are more reliable than the annual forecasts used for budget planning. Further input
data are the production and sales quantities per month that have been agreed upon
in the budget planning, or the respective volume and earning goals (e.g. per year).
One objective of the master production planning is to meet these targets as close
as possible in the short term. Constraints to be respected are quite the same as
were relevant for the budget planning. However, again a higher level of detail is
necessary.
Results of the master production planning are the updated and more detailed
(e.g. weekly) productionplans of the assemblyplants andsales plans. The latter ones
include the quotas for the different sales regions. Because of the above mentioned
constraints, these quotas may exceed or fall below the requests for car models,
originallydemandedbythe regions. Asimilar settingof (monthlyinsteadof weekly)
quotas for sales regions may possibly also be part of the annual budget planning.
For both budget and master production planning LP or MIP models seem to be
appropriate. However, for reasons to be explained in Section 3.3, they are not used
in practice at the moment. Planning usually is only supported by simple spreadsheet
modeling.
The production plans for car models, which are a result of the annual budget
and master production planning, are the basis to derive the component demand in a
further material requirements planning (MRP) procedure. The component demand
is communicated to the ?rst-tier suppliers as a preview of the quantities to be
delivered within the next months. As the options of the cars are just speci?ed for
the 3–5 weeks before production (see Fig. 1, phases (I) and (II)), and since the share
of ?nal customers’ orders decreases rapidly for longer lead times (phases (III) and
(IV)), this component demand becomes more and more unreliable, the longer the
forecast horizon is.
On the sales side, the allocation planning has to allocate the aggregate quotas,
which are known as a result of the budget planning on a monthly basis and as a
result of the master production planning on a weekly basis, to the lower levels of the
sales system. Depending on the organizational structure of the car manufacturer,
Supply chain planning in the German automotive industry 455
this planning task may occur on several hierarchical levels, e.g. ?rst an allocation of
quotas of world regions to different countries, and afterwards an allocation of these
more detailed quotas to the countries’ respective retailers and sales subsidiaries.
As an example, in the following only the relation “world region ? countries” is
considered: After the annual budget planning, the respective monthly quotas (sales
plans) of the world regions have to be allocated to the countries with respect to
their original requests. If the requests cannot all be satis?ed, it has to be decided,
whose demand will only be ful?lled partly. This “shortage planning” may follow
some prede?ned rules (so-called “fair share rules”, see e.g. [30, p. 169 f.]), which,
for example, might re?ect the purchase behavior of a country in the past, or more
or less be based on “negotiations” between representatives of the world regions and
of the respective countries. Furthermore, a region has to balance the deviations of
the countries’ actual demands from their former requests between all the different
countries assigned to the region. For this purpose, the region may also (call for and)
hold a “regional” pool of cars, originally not having been requested by one of the
countries.
3.2 Order-driven planning
Until now planning tasks have been discussed, which mainly build on forecasts for
options. In other words, only a few fully speci?ed orders are known at the time of
planning. In this section planning tasks will be considered which are exclusively
triggered by fully speci?ed orders, either of ?nal customers or of sales subsidiaries
and retailers. Figure 3 gives an overview of these order-driven planning tasks and
their interrelations.
final
customers
suppliers
allocation
planning
order promising
line assignment &
model mix planning
sequencing
plant
assignment
MRP &
lot-sizing
MRP distribution
customer
orders
customer &
retailer orders
specified orders
with due dates
(weekly)
production orders
per plant
daily buckets
daily buckets
procurement
lot-sizes
JIT-
calls
SILS
car
sequence cars
specification
changes
(weekly)
production plans,
overtime
aggregate quotas
promised
dates
detailed
quotas
procurement production distribution sales
Fig. 3. Overview of order-driven planning
456 H. Meyr
Direct buying of cars via the Internet is not (yet) worth mentioning. Normally
private customers order their cars via the sales subsidiaries or retailers of the car
manufacturer. The respective sales personnel tells the ?nal customers the expected
delivery dates of their desired cars. Usually, a granularity of weeks is suf?cient
for the customer, who e.g. has to provide the money on time and to synchronize
the delivery with the selling of his used car. Thus order promising, i.e. promising
reliable delivery dates to the customer, is an important task. If a free quota of the
sales subsidiary or retailer is available, the ?nal customer gets his desired delivery
date promised. Otherwise, the next free quota is recommendedor a standarddelivery
time is proposed (if quotas are not available in suf?cient detail). The customer may
accept the promised date, change the options of his desired car or even the model
type (in order to get an earlier delivery date), or try his luck with another retailer.
Furthermore, retailers and sales subsidiaries have to specify the options for
that part of their quotas that has not been ?lled up with ?nal customers’ orders
until the agreed date of speci?cation (see phase (II) of Fig. 1). In order to reduce
inventories at the retailers’ sites, the desired options of potential customers have
to be anticipated as precisely as possible. Because of the rather small number of
customers and large number of options, this is an almost unsolvable problem for
a single retailer. Thus, Stautner [51] suggests central support of the manufacturer
for these decentral forecasts of the retailers (see Sect. 5.1) and Holweg and Pil [24]
even propose a central pool of BTS cars.
Traditionally, these fully speci?ed orders are collected by the respective sales
organization, responsible for a certain retailer, and sent in bulk (e.g. all orders of
a week) to the next higher level of the sales hierarchy. A central order manage-
ment department of the car manufacturer ?nally has to select an assembly plant,
able to produce the car model requested by a certain order. This plant assign-
ment has to consider the production quantities and capacities per plant, that have
been agreed upon in the master production planning. If the actually requested car
options signi?cantly deviate fromthe ones assumed within master production plan-
ning (e.g. when anticipating bottlenecks of components or model mix constraints),
some orders have to be ful?lled earlier and others have to be delayed, thus resulting
in a re-assignment of orders to weeks.
The selection of an assembly plant was not a big problem so far because tradi-
tionally car manufacturers had little ?exibility in assigning cars to plants and thus
this task has (up to the author’s knowledge) not directly been addressed in the OR
literature. However, recently body shop and assembly have become ?exible enough
to allow model swap and thus the degrees of freedom and the need for intelligent
planning methods grow. In [17], the more important assignment of customer orders
to discrete time buckets with respect to promised due dates and to material/capacity
constraints is introduced as a planning task called “demand supply matching” and
corresponding LP/MIP models are formulated. However, the speci?c requirements
of the automotive industry (e.g. several assembly plants, model mix constraints)
are not considered. Lovgren and Racer [33] make a ?rst step towards mixed model
assembly line sequencing with respect to given due dates of orders. They calculate
detailed sequences of cars for a single assembly line. Thus, their model is rather
designed for the short-term line sequencing (see below) than the more aggregate
Supply chain planning in the German automotive industry 457
plant assignment. However, the problem of early or late demand ful?llment in the
automotive industry is at least generally addressed.
After this assignment, the decentral short-termproductionplanningdepartments
of the assembly plants have production orders available, that ought to be assembled
within (or up to) their pre-de?ned week of production (ideally still the promised
week minus a standard lead time for delivery to the respective customer). The
shorter the planning horizon is, the more restrictive the model mix constraints are.
Thus, the line assignment & model mix planning have to distribute the production
orders among the possibly parallel assembly lines and to assign days of production
to the orders. Doing this, the most important model mix constraints (e.g. “at a
maximum 300 air conditionings per day”) have to be considered, but the assembly
sequence of a day is not yet determined. Scholl [47, p. 108f.] denotes this task as
“Master Sequencing” and suggests, for reasons of complexity, a further aggregation
of individual orders to families of cars. Again, this planning task has not adequately
been tackled in the literature. Only Mergenthaler et al. [35] and Ding and Tulani
[11] address the single line (sub-)problem directly. The former ones try to smooth
the daily workload of a week by modifying a bin packing algorithm in order to
minimize the quadratic model mix deviation in a greedy manner, whereas the latter
ones apply simple neighborhood operations like “switching models of differently
utilized days” in a two-phase greedy algorithm.
As compared to mid-term planning, car options are now known with a high
reliability. Since the daily assembly buckets are also known as a result of the line
assignment & model mix planning, the daily demand of components can directly
be derived. For components and material, that are collected by regional carriers and
temporarily stored in an intermediate warehouse (see p. 451), an MRP & lot-sizing
procedure is appropriate that balances the trade off between inventory holding costs
and degressive transportation costs of the regional carriers and determines adequate
supply frequencies.
The daily buckets of the line assignment & model mix planning are also guide-
lines for the daily sequencing of the assembly lines. Here, the sequences of the
production orders on the ?nal assembly lines are determined on a rolling horizon
basis with a planning horizon of one to two weeks. The level of detail again is higher
than in model mix planning. Now all potential bottlenecks have to be considered,
for example, the availability of all of the components and “distance” restrictions of
the lines like “no two cars with air-conditioning are allowed to follow each other ”.
For this reason, sometimes the earlier assignment to days of production cannot be
maintained. However, it should be avoided to postpone an order to another week
than the planned (and promised) one. To use ?exible workforce or to work during
lunch breaks are short-term measures to extend capacity.
Undoubtedly, most scienti?c research on planning aspects of automotive supply
chains has been done in the ?elds of balancing and sequencing mixed-model assem-
bly lines. In the sequencing literature, usually it is assumed that orders have already
been assigned to a certain period (e.g. a day) of production, so that promised due
dates need not to be considered any further. The various sequencing approaches
differ with respect to their different objectives. Besides cost-oriented objectives,
mainly time related or JIT-objectives and combinations thereof are pursued (see
458 H. Meyr
e.g. [32, p. 44 ff.] and [47, p. 98 ff.]). A comprehensive literature review of models
and exact/heuristic solution methods is given by Scholl [47]. Summing up, priority
based (greedy) heuristics [47, p. 205 ff.] are – for reasons of complexity – clearly
favored over exact (mainly branch and bound) methods [47, p. 199 ff.]. Newer
heuristic approaches also apply multi agent systems [9] or local search methods
like simulated annealing or genetic algorithms (see e.g. [35],[25], [43],[46, p. 40]).
For a review of models and methods with respect to the different objectives,
the reader is referred to Lochmann [32]. Models with time related objectives [32,
p. 58 ff.] try to smooth the work load and minimize the overload of the various sta-
tions of a line. For this, usually MIP models are formulated. JIT-objectives attempt
to smooth the material supply at the stations in order to keep the inventory of com-
ponents constantly low. The usage rates of components are either leveled directly
[32, p. 81 ff.] or, in case the cars require a similar number and mix of components,
the mix of cars is leveled instead [32, p. 86 ff.]. The latter “level scheduling” was
introduced by Miltenburg [37] and commonly pursues nonlinear goals. Thus both
time relatedobjectives andJIT-objectives directlyaddress the model mixconstraints
discussed so far.
The car sequencing problem(CSP), originally introduced by Parretto et al. [40],
allows to model the above mentioned minimum distances between orders with the
same options and further separation rules like a “maximum number of identical
options within a car sequence of prede?ned length”. The CSP in not widely known
within the OR community, but one of the classical problems in the literature on
constraint satisfaction problems [7]. Brailsford et al. [7] review this kind of litera-
ture, showing that these “soft” constraints also pursue time-related objectives and
that JIT- and some further objectives can also be modeled as soft constraints of a
CSP. They report that – by using a hybrid approach combining simulated anneal-
ing and constraint logic programming – David and Chew [10] are able to obtain
good solutions for a practical problem at Renault involving 7500 cars with 50–100
options each.
Recent approaches of Drexl et al. combine the classical CSP with level schedul-
ing [12] and solve it in a two stage approach [13]. Also Monden [38, Chapt. 17]
extends his JIT-oriented “goal-chasing” heuristics in order to respect CSP distance
objectives (denoted as “continuation control” and “interval control”). Zeramdini et
al. [55] propose a two-stage approach, smoothing the components’ usage ?rst and
the workload secondly, to optimize the bicriteria sequencing problem. Kormazel
and Meral [31] reformulate the same combined problem as an assignment problem
with weighted objectives and develop heuristics for it. Hyun et al. [25] and later on
Ponnambalam et al. [43] consider “minimization of setup costs” as a third striving
objective and ?nd (near-) Pareto optimal solutions for the multi-objective prob-
lem by using a genetic algorithm. A further overview of models and methods for
combined objectives is given in [32, p. 92 ff.]. Concluding this brief discussion of
sequencing, it can be stated that there is a trend in recent literature on mixed-model
assembly line sequencing to consider several objectives, simultaneously.
The frozen car sequence is then the basis to derive the component demand for
JIT calls and SIL supply. This short-term material requirements planning (MRP) is
not a “real” planning task because there is nothing left to be decided about. Just the
Supply chain planning in the German automotive industry 459
BOM has to be exploded as late as possible before the scheduled delivery (usually
several times per day). It is just mentioned to provide a complete picture of supplier
relationships.
If ?nal customers do not pick up their cars at the assembly sites directly, the ?n-
ished cars have to be brought to the customers or their respective retailers and sales
subsidiaries. There again are some decisions to be made concerning the distribution
of the ?nished cars. For example, the actual carrier has to be chosen, and transport
frequencies (how often to deliver to a retailer) and vehicle routes (sequence of re-
tailers within a tour) have to be determined. Some of these tasks are in the planning
domain of logistic service providers [8].
3.3 Organizational issues
One has to be aware that in the preceding sections only “abstract” planning tasks
of car manufacturers have been described, but organizational issues have been left
aside. In reality, often several different planning departments are involved in a single
planning task. Then there are several “coordination rounds“ whose result is a com-
mon plan. Within each coordination round, a single department has to contribute its
own (locally “optimal”) partial plan until some prede?ned date. Such a (temporarily
valid) partial plan is a sort of self-commitment of the respective department and
provides input for the next planning activity of another department. This procedure
iterates until the common plan hopefully respects all relevant constraints and ful?lls
the various and sometimes con?icting objectives of the different departments to an
acceptable level.
The mutual arcs between production and sales in the budget planning and
master production planning boxes of Figure 2 ought to indicate that in practice the
respective planning task usually is not tackled in a single, simultaneous planning
procedure, but in the above mentioned coordination rounds. This is one reason
why LP and MIP models are not used for a simultaneous budget planning or a
simultaneous master production planning as it is common practice in other types
of industries like consumer goods manufacturing, for example [45]. Wahl [54]
proposes appropriate models to – at least individually – optimize the planning
decisions of the sales department in this way. But even such a local application of
LP and MIP has not been implemented in practice for reasons like missing (IT)
infrastructures, inappropriate forms of organization or mostly a lack of acceptance
and understanding of OR methods.
4 Current trends in the German automotive industry
As Figure 1 shows, “to move from BTS to BTO” is a somewhat imprecise formu-
lation. The task is rather to increase the share of ?nal customers’ orders. Further
strategic goals, currently pursued in the German automotive industry, are to shorten
customer order delivery times of customized cars, to keep promised delivery dates
with a high reliability and to allow customers to change their car options also in the
very short term [51, p. 31 ff.]. In order to reach these goals in addition to supply
460 H. Meyr
chain collaboration (see e.g. [18]) two major bundles of measures, online ordering
and late order assignment, have been and are still being implemented.
4.1 Online ordering
The total order-to-delivery lead time (OTD) can be shortened by reducing the lead
times of all individual processes (like order entry and processing, manufacturing,
distribution) involved. Since manufacturing and distribution only comprise a very
small percentage of the OTD (about 16 % according to [22, Figure 3]), the highest
potential can be found in order entry and processing. “Online ordering” initiatives
aim at simplifying and accelerating the circumstantial and timely collecting and
(weekly) bulk processing of orders within the multi-stage sales hierarchy. Thus
retailers send fully speci?ed ordering requests of ?nal customers via the Extranet
or Internet directly to a central order processing system, where the requests are
online (i.e. within seconds or minutes) checked for technical feasibility and pro-
vided with a promised delivery date. In case of ?nal customer’s acceptance of the
promised date, the ?nal order is processed with the same speed on the same route.
By implementing such a system, the car manufacturer BMW tries to reduce the
lead time of order entry from 13–17 days to a single day [44], for instance. Figure
4 graphically illustrates how online ordering reduces demand uncertainty. In this
(?ctitious) example, cutting the lead times of order entry in half triples the share of
?nal customers’ orders known. Thus the forecast-based BTS inventory of retailers
(phase (II), see also Fig. 1) can be reduced signi?cantly.
Fig. 4. Example of lead time reduction by online ordering
4.2 Late order assignment
Traditionally each body-in-white, physically processed within the body shop, is
already assigned to a customer order (“order assignment”) and a re-assignment to
Supply chain planning in the German automotive industry 461
another order is only rarely practicable. Following the pull-principles of the just-in-
time philosophy the ?nal assembly as the last production stage has to be planned ?rst
and synchronizes all direct suppliers and upstreamproduction stages, especially the
paint shop and the body shop. In the light of “lean thinking” the work-in-process
buffers (body store and painted body store) should be small and thus body and paint
shop ideally produce in the same sequence of customer orders as is planned for the
?nal assembly. However, these buffers are still necessary because process failures
in the body and paint shops occur frequently [49, p. 29 f.]. According to Holweg
[21] the rework rate is even up to 40–50 %. For this reason, a planned assembly
sequence can only be considered to be reliable, when the respective orders’ painted
bodies have left the paint shop. Thus the sequence can only be transmitted to the
SIL suppliers a few hours before planned assembly, depending on the assembly
station and the respective component.
In order to guarantee more reliable assembly plans, which can be ?xed for a
longer time interval (about 4–6 days), the order assignment nowadays is postponed
to the ?nal assembly stage (“late order assignment” or “late order tagging”, see
[21]), i.e. the bodies in the body and paint shops are no longer identi?ed by customer
orders. Body and paint shops still get the information about the customer orders to
be assembled, but are free to deviate fromthe planned assembly sequence. Although
there is no demand uncertainty, safety stocks have to be installed for each body-
in-white variant and paint color. These safety stocks exclusively hedge against
the process failures in the body and paint shops. In order to limit the total amount
of safety stock required and to restrict buffer sizes, the number of body-in-white
variations and paints (the so-called “internal complexity”, [21]) should be low. For
this reason, BMW reduced the number of body-in-white variations from 40 000 to
16 for their new 3 series when introducing late order assignment [21]. The higher
stability of assembly plans is expected to increase the radius of JIT/SIL delivery
and the share of JIT/SIL-suppliers signi?cantly.
5 Impacts on planning
Online ordering and late order assignment have been and still are being introduced
by BMW (project title “Kundenorientierter Vertriebs- und Produktionsprozess”
[44]) and DaimlerChrysler (project titles “Global Ordering” and “Perlenkette”
[18]). Further car manufacturers intend to follow. These two types of measures
considerably in?uence the traditional planning landscape as discussed in Section 3.
Thus it is necessary to check how planning requirements and information ?ows
change (some planning tasks may loose importance whereas others win) and which
new planning tasks arise.
5.1 Impacts of online ordering
Online ordering and online order promising require extremely short response times
for incoming customer requests. If highly reliable promised delivery dates shall
462 H. Meyr
be achieved, the capacities of all potential bottlenecks (material or production re-
sources) have to be checked. Thus the formerly decentral order promising has to
be automated and centralized. The changes in the planning landscape depend on
the level of delivery reliability aspired. In the following only two extreme scenar-
ios, denoted as quota-available-to-promise (QATP) and capable-to-promise (CTP)
scenario, are discussed as examples. Of course, there are various intermediates
conceivable between these extremes.
5.1.1 QATP scenario. The QATP scenario is more or less an automation of al-
ready existing processes. As Figure 5 shows, the general logic of planning stays
the same. The quotas for retailers and sales subsidiaries, which have been deter-
mined on a weekly basis anyway and have been synchronized with capacities in
the medium term, are (as far as they have not yet been assigned to ?nal customers’
orders) considered to be “available to be promised”. Incoming customer requests
and customer orders, respectively, are checked for technical feasibility [26], ?rst,
and according to simple precedence rules [30] for free quotas, secondly. Such a
proceeding is known from material constrained industries like the computer indus-
try and successfully applied there [29]. In contrast, however, material availability
is not yet checked in the simple QATP scenario. The installation of an online order-
ing system (OOS) is technically lavish and costly, but hardly changes the planning
logic. When comparing Figure 5 with Figure 3, the major differences are that spec-
i?ed orders (and their due dates) are directly transmitted to the plant assignment
instead of using the multi-stage sales hierarchy and that speci?cation changes can
be sent faster (and thus later) to the model mix planning.
promised
dates
line assignment &
model mix planning
plant
assignment
final
customers
allocation
planning
online
order promising
specified orders
with due dates
aggregate quotas
detailed
quotas
(QATP)
production orders
per plant
specification
changes
daily buckets
daily
buckets
retailers
requests,
orders,
specification
changes
promised
dates
promised
dates
production sales
(weekly)
production plans,
overtime
Fig. 5. QATP scenario
However, because the mid-term capacity check, on which free quotas (QATP)
are based, had no detailed information about the customers’ choice of car options,
there is a high probability that the promised delivery dates do not ?t the model mix
constraints and thus cannot be kept on the short-term.
Supply chain planning in the German automotive industry 463
5.1.2 CTP scenario. In order to achieve a higher delivery reliability, a shorter-
term and more detailed capacity check is necessary, which motivates the other,
more challenging extreme, the capable-to-promise (CTP) scenario. When accept-
ing orders and con?rming delivery dates, the customer orders are directly booked
[22] to a day of production or week of production (if a late delivery is desired by the
?nal customer) of an adequate assembly plant. In contrary to the QATP scenario,
all or at least the most crucial constraints, relevant for model mix planning (like
options of the orders, material required, production capacity, quotas of the respec-
tive sales hierarchy), are considered. The order promising is extended such that
production orders can automatically be generated. Thus the online order promising
takes on planning tasks of the short-term production planning or – at least – limits
its scope. Furthermore, also the plant assignment has to be integrated into such a
comprehensive online ordering.
Questions, which have to be answered online, are for example: Is a BTS car
physically available somewhere in the supply chain, which ?ts the requirements of
the new customer order to a very high degree? Is a similar BTS car planned and can
its options be changed so that the order still can be assigned to it? Which plant has
to be chosen if a new production order has to be generated? Should be produced
earlier or later than the desired date, if this is already (over)booked? If model mix
constraints are limiting, which car speci?cations should a customer change in order
to still get his desired delivery date promised?
However, one has to keep in mind that the computational burden to update all
the necessary data and the desired response times of the OOS are con?icting. The
major problem is to ?nd the right trade off between modeling capacities as detailed
as necessary (increases delivery reliability) and updating as few data as possible
(in order to guarantee short response times).
Figure 6 shows the embedding of a CTP online order promising into the plan-
ning landscape. The online order promising needs free quotas (QATP) and not
yet assigned net capacities of material (“material-available-to-promise“, MATP)
and assembly resources (CTP), e.g. expressed by a maximum number of cars with
a speci?c (combination of) option(s) per day, as inputs. The results of the order
promising are weekly delivery dates, which are promised to the customers, and
“production orders” with a promised delivery date and a planned day (or week)
of production, which are sent to the line assignment & model mix planning of the
respective plants.
A decentral model mix planning is still necessary for several reasons. For ex-
ample, the preliminary production plans of order promising have to be updated
with respect to (for complexity reasons) still unconsidered capacity constraints and
a line assignment has to be made. Production orders, which have only been allo-
cated to a week of production because of rather long customer order lead times
being desired, have to be assigned to a day of production. The more detailed the
capacity constraints of order promising are, the less changes of its plans should be
necessary later on in the model mix planning because the most crucial potential
bottlenecks have already been anticipated. However, short-term failures in supply
and production can never be avoided and thus make a re-planning necessary.
464 H. Meyr
line assignment &
model mix planning
netting
final
customers
allocation
planning
sequencing
production orders
per plant
and day / week
specification
changes
aggregate quotas
detailed quotas
(QATP)
retailers
online
order promising
(incl. plant assignm.)
promised
dates
promised
dates
promised
dates
MATP,
CTP
daily
buckets
daily
buckets
production sales
(weekly) production
plans, overtime
requests,
orders,
specification
changes
Fig. 6. CTP scenario
The results of the model mix planning are daily buckets, which again are sent to
the sequencing, but are – in a further netting procedure – also used to calculate the
(net) MATP and (net) CTP for the online order promising. Further input for the net-
ting is up-to-date information about plant capacities and projected material supply,
which have been synchronized in the master production planning in the medium
term (see Fig. 2). Fleischmann and Meyr [17] illustrate the interaction between
“order entry” (online order promising) and “MATP/CTP (re)calculation” (netting
and model mix planning) by means of two more detailed work?ows and discuss the
planning tasks of demand ful?llment for various positions of decoupling points.
They also propose LP and MIP models for order (re-) promising, which are useful
if several customer requests/orders can be processed in a batch. However, if each
customer request has to be answered immediately, the degree of freedom is rather
low. Consequently, the importance and impact of previous planning tasks, like mas-
ter production and allocation planning, grows. The APS vendor SAP [46] offers a
software module called Realtime-Positioning, which has especially been designed
for the online order promising in the CTP-scenario, and Ohl [39, p. 207 ff.] dis-
cusses the advantages of “code rules”, describing the interrelations between various
car options, for a capacitated BOM-explosion of online queries. However, formal
models for the MATP/CTP calculation and search are not presented. Similarly to
the approach of Ohl, Bertrand et al. [4] propose to use a hierarchical pseudobill of
material for the MATP check in case of strong interdependencies between different
options (so-called “non-modular products”).
Besides newly arriving customer requests/orders also changes of the speci?ca-
tion of already accepted orders can be processed and checked for capacity using
the online ordering system. Furthermore, the speci?cation of not yet ful?lled quo-
tas by retailers (see phase (II) of Fig. 1) can be checked with respect to model
mix constraints. Online ordering accelerates order processing and increases the
share of ?nal customers’ orders (see Fig. 4), but BTS cars cannot completely be
Supply chain planning in the German automotive industry 465
avoided [51, p. 6]. In order to decrease the times in inventory of the remaining BTS
cars, ?nal customers’ desired options should be anticipated more precisely. Central
statistics about the ?nal customers’ preferences and about frequently purchased
options can comfortably be made available to retailers by means of the OOS. They
widen the local view of the retailers and promise a higher quality of forecasts for
BTS speci?cations [51, p. 176 ff.]. These proposals for BTS options and the more
detailed MATP/CTP capacity check can be seen as new potentials that arise due to
the centralization of order promising and the online connection to retailers.
5.2 Impacts of late order assignment
Late order assignment undoubtedly has its major impacts on strategic planning.
Products have to be re-designed so that a high number of options (high external
variety) can be kept up while simultaneously reducing the number of body-in-white
variations (low internal complexity [21]). There is a rich OR literature on design
for postponement and modularization (see Sect. 1 and [3], for instance), which tries
to support such issues. Furthermore, the re-dimensioning of the (body store and)
painted body store is a strategic planning task.
But also for the operational planning of the body and paint shops and their re-
spective stores new challenges arise because of the higher degrees of freedom. The
safety stocks of the body store and the painted body store have to be re?lled with
respect to the failure probability of the respective production processes. Because
of the rather loose coupling to the assembly sequence and because of the increased
buffer sizes, lot-sizing issues can now be considered easier in the paint shop. Al-
though changeover times are negligible, batching lots is economically desirable
because a change of the paints incurs costs between 10 and 30 [49, p. 30].
Taking cars out of the body store is a Sequential Ordering Problem [15], a special
variant of the Traveling Salesman Problem. For paint shops as a practical applica-
tion, Spieckermann [49, p. 126 ff.] proposes a branch-and-bound approach which
takes advantage of special knowledge about common structures of body stores in
the automotive industry (see also [50] for earlier approaches to the same problem).
Engel et al. [14] propose a heuristic for workload leveling which can be extended
for the batch sequencing of paints of the same color.
Inman and Schmeling [27] prove the operational advantages of late order assign-
ment by means of simulation. They compare the traditional irreversible coupling of
orders and physical vehicles at the body shop with a ?exible assignment procedure
(“Agile Assemble-to-Order” (AAO) system) that is able to assign and re-assign or-
ders to vehicles before the body shop, paint shop and the ?nal assembly are entered.
The objective of the AAO system is a weighted function comprising penalty terms
for violating lead time, paint colour, spacing and levelness constraints. Orders are
selected by the AAOsystemin a greedy manner with the weights varying according
to preferences of the production stage under consideration.
466 H. Meyr
6 Conclusions and outlook
Concerning forecast-driven planning it can be stated that the quotas of the tra-
ditional master production and allocation planning had a detrimental effect on
meeting ?nal customers’ demand on time. This gets even worse if the same quotas
are directly taken over to an OOS with automatic booking and without the pos-
sibility of human intervention (see Sect. 5.1.1). Thus, if still necessary to smooth
the workload in the medium term, one has to think about more ?exible allocation
mechanisms incorporating the increased knowledge (see Sect. 4.1) about ?nal cus-
tomers’ demand. Virtual, central car pools, accessible for several retailers, are a
?rst step in this direction. The choice of adequate aggregation levels, allowing to
postpone decisions as long as possible, is crucial.
In other lines of business it has been shown that LP and MIP models can support
planning tasks like budget, master production and allocation planning. Wahl [54]
has proven that this would also be true for (at least the sales side of) automotive
industries. The reasons, why the proposals of Wahl have not been put into practice,
should have diminished or even vanished in the meantime. Information technology
has improved dramatically in recent years and there seems to be a higher willingness
to make use of ORtools. APS, for example, are a comfortable and user friendly way
to apply LP and MIP methods in practice. In addition, simultaneous optimization
covering several departments like production, procurement and sales in a single
model could exploit further potentials and – at least simulatively – support and
accelerate the lengthy coordination rounds (see Sect. 3.3).
Regarding the traditional order-driven planning it has been shown that OR
support for the planning tasks plant assignment and line assignment & model mix
planning was very poor. However, these tasks will change their character anyway
when online ordering and the CTP scenario are installed. On the other hand, there is
rich literature on assembly line sequencing and research in this ?eld is an ongoing
process. Recent OR-related papers tend to pursue several objectives simultaneously,
thus becoming more attractive for practical application in the automotive industry.
However, scalability of sophisticated methods is still a problem and should be a
topic of future research.
As we have seen, the measures to move from BTS to BTO also have signi?-
cant impact on planning. The consequences for forecast-driven planning have been
sketched above. Further challenges can be identi?ed for the future order-driven
planning. Due to late order assignment the close coupling of body, paint and as-
sembly shops has been decreased now. Thus there remains supplementary freedom
for paint shop sequencing and batching of paints of the same color. However, be-
cause of still limited buffer sizes, OR models have to take care that paint shop
sequences may not deviate too far from assembly sequences.
Online ordering is most challenging in the CTP scenario when incoming orders
have tobe bookeddirectlyintoa (capacitated) productionplanof a plant. Inthis case,
online order promising takes over functionalities of the traditional plant assignment
and the traditional line assignment & model mix planning. The three most crucial
problems are
Supply chain planning in the German automotive industry 467
– how to model quotas and model mix restrictions as constraints for the online
order promising (within the netting procedure, respecting the results of the
previous master production and allocation planning),
– which fast algorithms or search rules to use for allocating free QATP, MATP
and CTP (within online order promising) and
– how to revise the resulting preliminary production plans in case of still uncon-
sidered constraints and unforeseen short-term events (new line assignment &
model mix planning, respecting the already promised due dates).
Research has to be done on both ORmodels/methods for the different planning tasks
involved and – since responsibilities change – also on the (hierarchical) interrelation
of these planning tasks within the overall planning framework. If, above all, car
manufacturers think about customized sales prices, which may vary according to the
delivery times desired by ?nal customers, the relationship to revenue management
(see e.g. [34, 48]), as common in airline industries, has to be further investigated.
References
1. Amen M (2000) Heuristic methods for cost-oriented assembly line balancing: a survey.
International Journal of Production Economics 68: 1–14
2. Ashayeri J, van Eekelen B, Vossebeld R (2001) Lean and clean distribution in Europe:
how to get an ef?cient European distribution network. Technical Report, Department of
Econometrics and Operations Research, Tilburg University, The Netherlands; presented
at EURO 2001 Conference, Rotterdam, The Netherlands
3. Aviv Y, Federgruen A (1999) The bene?ts of design for postponement. In: Tayur SR,
Ganeshan R, Magazine MJ (eds) Quantitative models for supply chain management,
pp 553–584. Kluwer, Amsterdam
4. Bertrand JWM, Zuijderwijk M, Hegge HMH (2000) Using hierarchical pseudo bills of
material for customer order acceptance and optimal material replenishment in assemble
to order manufacturing of non-modular products. International Journal of Production
Economics 66: 171–184
5. Biller S, Bish BK, Muriel A (2002) Impact of manufacturing ?exibility on supply
chain performance in the automotive industry. In: Song J-S, Yao DY (eds) Supply chain
structures: coordination, information and optimization, pp 73–118. Kluwer, Boston,
MA
6. Bradley JR, Blossom AP (2002) Using product-mix ?exibility to implement a made-to-
order assembly line. Technical Report, Cornell University, Ithaca, NY; under revision
for: IIE Transactions
7. Brailsford SC, Potts CN, Smith BM(1999) Constraint satisfaction problems: algorithms
and applications. European Journal of Operational Research 119: 557–581
8. Carbone V, De Martino M(2003) LSPs in the automotive supply chain: a case study. In:
Juga J (ed) NOFOMA2003: Strivingfor leadingedge logistics, pp416–431. Proceedings
of the 15th Annual Conference for Nordic Researchers in Logistics, Oulu, Finland
9. Caridi M, Sianesi A (2000) Multi-agent systems in production planning and control: an
application to the scheduling of mixed-model assembly lines. International Journal of
Production Economics 68: 29–42
10. David J-M, Chew TL (1995) Constraint-based applications in production planning:
examples from the automotive industry. In: Proceedings of Practical Applications of
Constraint Technology (PACT’95), pp 37–51. The Practical Applications Company,
Blackpool, UK
468 H. Meyr
11. Ding F-Y, Tolani R (2003) Production planning to support mixed-model assembly.
Computers & Industrial Engineering 45: 375–392
12. Drexl A, Kimms A (2001) Sequencing JIT mixed-model assembly lines under station-
load and part-usage constraints. Management Science 47(3): 480–491
13. Drexl A, Kimms A, Matthießen L (2002) Algorithms for the car sequencing and level
scheduling problem. Technical Report, University of Kiel, Germany
14. Engel C, Zimmermann J, Steinhoff A (1998) Objectives for order-sequencing in au-
tomobile production. In: Drexl A, Kimms A (eds) Beyond manufacturing resource
planning (MRP II), pp 307–331. Springer, Berlin Heidelberg New York
15. Escudero LF(1988) An inexact algorithmfor the sequential ordering problem. European
Journal of Operational Research 37: 236–253
16. Fisher ML, Ittner CD(1999) The impact of product variety on automobile assembly op-
erations: empirical evidence and simulation analysis. Management Science 45(6): 771–
786
17. Fleischmann B, Meyr H(2003) Customer orientation in advanced planning systems. In:
Dyckhoff H, Lackes R, Reese J (eds) Supply chain management and reverse logistics,
pp 297–321. Springer, Berlin Heidelberg New York
18. Graf H, Putzlocher S (2002) DaimlerChrysler: Integrierte Beschaffungsnetzwerke.
In: Corsten D, Gabriel C (eds) Supply chain management, pp 47–61. Springer, Berlin
Heidelberg New York
19. Graves SC, Tomlin BT(2003) Process ?exibility in supply chains. Management Science
49(7): 907–919
20. Hoffman KC(2000) Ford develops different kind of engine – one that powers the supply
chain. Global Logistics & Supply Chain Strategies Nov 2000: 42–50
21. Holweg M (2002) Product variety – friend or foe? Technical Report, Lean Enterprise
Research Centre, Cardiff Business School, United Kingdom. International Journal of
Production Economics (to appear)
22. Holweg M, Jones DT (2001) The build-to-order challenge. Automotive World Jan–Feb
2001: 40–45
23. Holweg M, Jones DT ( 2001) The challenge: building cars to order – can current
automotive supply systems cope? In: Tyler D, Brunt D (eds) Manufacturing operations
and supply chain management – The lean approach, pp 362–374. Thomson Learning,
London
24. Holweg M, Pil FK(2001) Successful build-to-order strategies – start with the customer.
Sloan Management Review Fall: 74–83
25. Hyun CJ, Kim Y, Kim YK (1998) A genetic algorithm for multiple objective se-
quencing problems in mixed model assembly lines. Computers & Operations Research
25(7/8): 675–690
26. ILOG (2001) ILOG con?gurator. Technical Report, ILOG Worldwide Information
Center,http://www.ilog.com/products/con?gurator/whitepaper/index.cfm??lename=
wp con?gurator.pdf – last check of address: March 31, 2004
27. Inman RR, Schmeling DM (2003) Algorithm for agile assembling-to-order in the
automotive industry. International Journal of Production Research 41(16): 3831–3848
28. Jordan WC, Graves SC (1995) Principles on the bene?ts of manufacturing process
?exibility. Management Science 41(4): 577–594
29. Kilger C, Schneeweiss L (2002) Computer assembly. In: Stadtler H, Kilger C (eds)
Supply chain management and advanced planning, 2nd edn, pp 335–352. Springer,
Berlin Heidelberg New York
30. Kilger C, Schneeweiss L (2002) Demand ful?lment and ATP. In: Stadtler H, Kilger C
(eds) Supply chain management and advanced planning, 2nd edn, pp 161–175. Springer,
Berlin Heidelberg New York
Supply chain planning in the German automotive industry 469
31. Korkmazel T, Meral S (2001) Bicriteria sequencing methods for the mixed-model
assembly line in just-in-time production systems. European Journal of Operational
Research 131: 188–207
32. Lochmann M (1999) Operative Planung und Steuerung von Mehrprodukt-Fließlinien.
Deutscher Universit¨ ats-Verlag, Wiesbaden
33. Lovgren RH, Racer MJ (2000) Algorithms for mixed-model sequencing with due date
restrictions. European Journal of Operational Research 120: 408–422
34. McGill JI, Van Ryzin GJ (1999) Revenue management: research overview and pros-
pects. Transportation Science 33(2): 233–256
35. Mergenthaler W, Stadler W, Wilbertz H, Zimmer N (1995) Optimizing automotive
manufacturing sequences using simulated annealing and genetic algorithms. Control
Engineering Practice 3(4): 569–573
36. Miller T, Wise D, Clair L (1996) Transport network design and mode choice modeling
for automobile distribution: a case study. Location Science 4(3/4): 37–48
37. Miltenburg GJ (1989) Level schedules for mixed-model assembly lines in just-in-time
production systems. Management Science 35: 192–207
38. Monden Y (1998) Toyota production system, 3rd edn. Engineering & Management
Press, Norcross
39. Ohl S (2000) Prognose und Planung variantenreicher Produkte am Beispiel der Auto-
mobilindustrie. Fortschritt-Berichte VDI 16/120. VDI Verlag, D¨ usseldorf
40. Parretto BD, Kabat W, Wos L (1986) Jobshop scheduling using automated reasoning:
a case study of the car sequencing problem. Journal of Automated Reasoning 2: 1–42
41. Patchong A, Lemoine T, Kern G(2003) Improving car body production at PSAPeugeot
Citroen. Interfaces 33(1): 36–49
42. Pine BJ (1993) Mass customization: the new frontier in business competition. Harvard
Business School Press, Boston
43. PonnambalamSG, AravindanP, Subba RaoM(2003) Genetic algorithms for sequencing
problems in mixed model assembly lines. Computers &Industrial Engineering 45: 669–
690
44. Renner P (2000) Kundenorientierung in der Automobilindustrie. In: Innovatio-
nen in Logistikstrukturen der Automobilindustrie, Tagung 28.9.00 / VDI-Gesellschaft
F¨ ordertechnik, Material?uß, Logistik, VDI Bericht 1571, pp 21–33. VDI Verlag,
D¨ usseldorf
45. Rohde J, Wagner M(2002) Master planning. In: Stadtler H, Kilger C(eds) Supply chain
management and advanced planning, 2nd edn, pp 143–160. Springer, Berlin Heidelberg
New York
46. SAP AG (2001) “Managing the e-supply chain” in der Automobilindustrie mit mySAP
automotive. SAP White Paper (50 047 263), 05/2001
47. Scholl A (1995) Balancing and sequencing of assembly lines. Physica-Verlag, Heidel-
berg
48. Secomandi N, Abbott K, Atan T, Boyd EA (2002) From revenue management concepts
to software systems. Interfaces 32(2): 1–11
49. Spieckermann S (2002) Neue L¨ osungsans¨ atze f¨ ur ausgew¨ ahlte Planungsprobleme in
Automobilrohbau und -lackiererei. Shaker Verlag, Aachen
50. Spieckermann S, Voß S (1996) Paint shop simulation in the automotive industry. ASIM
Mitteilungen 54: 367–380
51. Stautner U (2001) Kundenorientierte Lagerfertigung im Automobilvertrieb. Deutscher
Universit¨ ats-Verlag, Wiesbaden
52. Swaminathan JM, Tayur SR(1999) Managing design of assembly sequences for product
lines that delay product differentiation. IIE Transactions 31: 1015–1026
470 H. Meyr
53. Thonemann UW, Brandeau ML (2000) Optimal commonality in component design.
Operations Research 48(1): 1–19
54. Wahl AH(1994) Simultane Produktions- und Distributionsplanung. Peter Lang, Frank-
furt a.M. et al.
55. Zeramdini W, Aigbedo H, Monden Y (2000) Bicriteria sequencing for just-in-
time mixed-model assembly lines. International Journal of Production Research
38(15): 3451–3470

doc_516676122.pdf