“Dealing with Complexity: Improving delivery reliability in a make-to-order environment”
E.W. van Winsum Boermarke 24A
9481 HA Vries 0624208310
e.w.van.winsum@student.rug.nl
student number 1838334
Business Administration Operations and Supply Chains Faculty of Economics and Business
University of Groningen Supervisors:
Dr. M.J. Land Dr. G.D. Soepenberg Alfa Laval Groningen supervisor:
W. van der Steege, MSc.
April 26
th, 2011
TABLE OF CONTENTS
1 INTRODUCTION AND PROBLEM STATEMENT ... 3
1.1 Original management problem ... 3
1.2 Problem statement ... 3
1.3 Methodology ... 3
1.4 Thesis structure ... 4
2 ORDER PROCESS ... 4
2.1 Customer and Sales ... 4
2.2 Engineering and preparation ... 4
2.3 Purchasing and supplier ... 4
2.4 Storage ... 4
2.5 Sheet metal processing and storage of sheet metal ... 5
2.6 Production ... 5
2.7 Storage and distribution ... 5
3 DIAGNOSIS ... 5
3.1 Diagnosis methodology ... 5
3.1.1 Operationalization of variables ... 5
3.1.2 Diagnosis framework ... 7
3.2 Determining the relevant problem area ... 8
3.2.1 Step 1: Analysis of the distribution of lateness ... 8
3.2.2 Step 2: Analysis of differences among order subsets ... 8
3.2.3 Step 3: Analysis of differences over time ... 10
3.2.4 Step 4: Analysis Delivery Time Promising or Realization process ... 11
3.3 Determining PPC causes within the main problem area ... 11
3.4 Managerial indications of possible problem factors ... 12
3.5 Diagnosis outcome ... 13
4 THEORY-BASED DISCUSSION ... 14
5 CONCLUSIONS AND RECOMMENDATIONS ... 15
5.1 Conclusions ... 15
5.2 Recommendations ... 16
ACKNOWLEDGEMENTS ... 16
REFERENCES ... 16
APPENDIX 1 ... 18
APPENDIX 2 ... 19
1 INTRODUCTION AND PROBLEM STATEMENT
Alfa Laval Groningen B.V., previously named Helpman B.V., has been active in producing first- class air-coolers and air-cooled condensers since the beginning of the 20
thcentury. These products are used for refrigerating, freezing and air-
conditioning. Production is mainly concentrated on the industrial market and is approximately 80 % engineer-to-order, whereas the remaining 20 % is make-to-order. An organizational chart of Alfa Laval is attached in appendix 1.
1.1 Original management problem
The original management problem called out a need for transparency in the main order routing.
Recently, it has become organizational policy to exert the principles of Lean Manufacturing (Womack, Jones and Roos, 1990) and therefore try to eliminate all types of waste as mentioned in the Lean paradigm – overproduction, waiting, transportation, processing, inventory, movement and defects (King, 2009). However, applying the tools to eliminate these types of waste proved to be a complicated matter. Unforeseen situations arose, such as faulty information concerning inventory levels, large variance in delivery time of material and components, an irregular production schedule of sheet metal processing, etc. These situations are diametrically opposed to the preconditions for Lean Manufacturing, which provoked the need to increase the transparency of the logistical stream, before fully applying the lean principles in order to eliminate waste.
Exploration of the original management problem (lack of transparency in the logistical stream) led to the conclusion that various performance aspects suffer from the lack of transparency in the logistical stream. One of these performance variables which especially suffer from lack of transparency is delivery reliability. In this thesis, the generally low delivery reliability (est. 40
% on time) is recognized and investigated as the organizational problem. The importance of delivery reliability performance of make-to-order and engineer-to-order organizations has been widely acknowledged in literature (Hendry and Kingsman, 1991; Kingsman et al., 1996; Stefansson et al., 2009; etc.). Soepenberg et al. (2008) state that
“high delivery reliability is one of the main order winning criteria for make-to-order companies”.
Diagnosing the delivery reliability will provide management with conclusions and
recommendations with respect to improving this performance variable. A concurrent development is made during this diagnosis, namely improvement of transparency in the logistical stream. However, the
research focuses on determining factors that negatively influence delivery reliability.
The delivery reliability is calculated by measuring the average percentage of orders delivered on time to customers. Based on the preceding explanation the following organizational problem can be devised: The delivery reliability is
too low (est. 40 % on time). In consultation with theprincipal, the problem statement has followed from this organizational problem.
1.2 Problem statement
This section contains the research objective and the research questions. These subjects were drawn up in consultation with the principal.
Research objective:
The goal of the research is to produce recommendations with respect to improving the delivery reliability.
Research questions:
- Which variables influence the delivery reliability and how do they relate to each other?
- How can these variables and the relations between these variables be influenced in such a way that the delivery reliability will be improved?
1.3 Methodology
The overall methodology is to perform a quantitative diagnosis and a qualitative theory- based discussion in order to achieve the research objective. In order to effectively diagnose the variables which are of influence on delivery reliability and the relations between them, a diagnosis framework (Soepenberg, 2010) plays an important role. This recently developed framework supports the diagnosis, in order to “indicate those production planning and control (PPC) decisions that have a negative influence on the delivery reliability of make-to-order companies”. The diagnosis framework is also applicable to engineer- to-order environments, and essentially consists of two steps. The first step is to determine the relevant problem area, the second is to determine production planning and control causes within the relevant problem area.
Triangulation of the outcome of the diagnosis
framework delivered two problem indications that
have been provided by Alfa Laval management,
which possibly influence delivery reliability. An
individual quantitative diagnosis of these
managerial indications has been executed while
using tools from the diagnosis framework of
Soepenberg (2010). An extensive description of the methodology is given in section 3.1.
1.4 Thesis structure
The thesis is structured as follows. In order to offer insight into the main order stream, section 2 provides a description of the order process of Alfa Laval Groningen. Section 3.1 provides an extensive description of the utilized diagnosis framework of Soepenberg (2010). Section 3.2 is based on this framework, and contains the quantitative diagnosis of delivery reliability in order to develop insights in the main problem area. Section 3.3 discusses production planning and control causes in the main problem area. Section 3.4 delivers an individual diagnosis, not related to the diagnosis framework, to investigate managerial indications of possible problem factors influencing delivery reliability.
Section 3 is concluded with a concise description of the outcome of the diagnosis, together with a causal model. This causal model indicates the variables influencing delivery reliability and their relations to each other. Section 4 delivers a theory-based discussion concerning the design that should be provided to the organization in order to create opportunities to influence variables in such a way that delivery reliability will be improved. Section 5 concentrates on conclusions and recommendations.
2 ORDER PROCESS
The goal of this section is to provide a concise description of the order process of Alfa Laval Groningen, in order to facilitate understanding of the diagnosis phase. The main order routing (figure 2.1) is used as a guideline to structure the
description of the order process.
2.1 Customer and Sales
The customer applies for a quotation for a specific product. Specifications of the product are provided by the customer to the sales department of Alfa Laval. The sales department discusses with the engineering department concerning the feasibility of the proposed product, meaning whether or not Alfa Laval has the competencies to produce the product. If the engineering department approves the product, the sales department delivers an initial quotation to the customer, quoting a price and a lead time. The price is based on norm times of engineering, preparation and production for specific product groups. The lead time is based on a
delivery time list. This delivery time list indicates a limited number of different product groups, which all have norm lead times. Following the quotation, three options are possible: the customer sends his/her approval, the customer wishes to change parameters of the quotation or the customer does not wish to order. In the first case, the order is released to engineering. Additionally, a more accurate quotation of lead time is made after engineering and preparation. In the second case, Alfa Laval deliberates with the customer, in order to synchronize all quotation parameters, after which the order is released to engineering. The additional quotation is also conducted. In the third case, the quotation is null and no order is released.
2.2 Engineering and preparation
The engineering department creates technical drawings of the product and determines which components are necessary to meet the needs of the customer. When completing this task, a bill of materials is produced. This bill of materials is processed into the ERP system (IFS) by the preparation department.
2.3 Purchasing and supplier
When a need is ascertained from IFS,
components and/or materials have to be ordered by the purchasing department. Subsequently,
according to the Standard Operating Procedure (SOP), the purchasing terms of at least two suppliers should be reclaimed, from which a supplier is selected on the basis of specifications, price, delivery time and international commercial terms. However, interviews show that in many cases the number of reviewed suppliers is constrained to one. Due to the nature of an engineer/make-to-order environment, the supplier lead times can vary.
2.4 Storage
Purchasing orders are examined by warehouse personnel for any visible defects. When visible defects are detected, the purchasing department is informed and the order is temporarily refused.
When the order is accepted, the waybill is signed by warehouse personnel, after which it is labeled with paper labels, stocked and processed in IFS by hand.
Warehouse personnel bear the responsibility of
storing materials and components at appointed
locations, guaranteeing the quality of the material.
From interviews, observation and review of documents it seems that no standardized warehousing approach is used in the area of warehouse management. Warehouse management is based on experience and know-how of a limited number of workers. Within Alfa Laval, one warehouse location distinction is made: planking, steel profiles and body work may be stored outdoors, while remaining products must be stored indoors.
2.5 Sheet metal processing and storage of sheet metal
Sheet metal processing has three significant functions: inspecting metal sheets for flaws, processing the metal sheets into bodies and storing the finished bodies. When the quality of metal sheets is approved, CNC machine programs are produced in order to process the metal sheets. These metal sheets are used in the production department for the finishing of final products.
A storage point is created between sheet metal processing and the production line. At this storage point, no standardized warehousing approach is used. Storage of material is based on experience.
Most of the produced material has to be coated by an external organization. This organization returns the material assembled in carts, with paper labels attached to these carts. This material is picked when a production order needs sheet metal for the finishing process.
In all cases, individual work orders are made for the sheet metal department. Sheet metal is needed in the final stage of production, and almost every production order needs a different set of sheet metal. Hence, these sheet metal work orders are linked to production work orders, combining to a single sales order.
2.6 Production
Production consists of four generally similar production lines. Initially, an order is released for production, after which the documenting of quality data takes place. The program for the CNC machine is produced and materials are stamped for quality purposes. Production roughly consists of three steps: building the structure (consisting of body work, fins and tubes), soldering/welding connections and finishing.
Throughout production, warehouse personnel are responsible for delivering materials and components to the production lines.
2.7 Storage and distribution
Hypothetically, finished products are to be stored for a short period of time (approx. 1 day).
Sometimes, finished products are stored longer because orders are produced early due to gaps in
capacity utilization when another order is put in backlog. No standardized approach could be determined in the way finished products are stored, except for a trivial division of the four product groups in the storage lot. Every finished product is manually labeled with paper. Outbound finished products are handled by external distributors, which transport the products to the customers. When a finished product leaves storage, the transaction is manually processed into IFS.
3 DIAGNOSIS
The goal of this section is to perform a diagnosis in order to distinguish the variables that influence delivery reliability and the relations between these variables. This section is organized as follows. Section 3.1 provides a description of the diagnosis methodology. Then, the first stage of the diagnosis is executed in section 3.2, which is the determination of the relevant problem area. Section 3.3 comprises the second stage of the diagnosis, determining PPC (production planning and control) causes in the relevant problem area.
3.1 Diagnosis methodology
3.1.1 Operationalization of variables
Soepenberg et al. (2008) state that in order to achieve high delivery reliability, both the average lateness and the variance of lateness of orders need to be controlled.
The delivery reliability is defined as the conformity of a schedule to a given due date with respect to finished orders. Lateness is defined as the conformity of a schedule for an individual order to a given due date and can be measured by
subtracting the promised delivery time from the realized throughput time (Soepenberg et al., 2008).
The definition of average lateness is bound to be clear. In order to diagnose average lateness, the throughput diagram is a suitable method. Variance of lateness is defined by the deviation of an order set from the schedule across jobs. For diagnosing the variance of lateness of orders, the order progress diagram can be used (Soepenberg et al., 2008). The percentage of late orders can be decreased by reducing the average lateness and/or by reducing the variance of lateness. Thus, the goal of the proposed diagnosis is to reveal variables negatively influencing delivery reliability, by using throughput diagrams and/or order progress diagrams.
The throughput diagram is regarded as a useful
tool for facilitating diagnosis of performance in
terms of delivery reliability (Wiendahl, 1988). It
presents the cumulative input and output of work of
a particular capacity group across time (Soepenberg
et al., 2008). Wiendahl (1988) has proposed
plotting the number of jobs in a general diagram known as the throughput diagram.
Fig. 3.1. Basics of throughput diagrams. (Soepenberg et al., 2008)
Figure 3.1 illustrates the basics of throughput diagrams, as drawn up by Soepenberg et al. (2008).
The horizontal axis shows the cumulative time in working days. The vertical axis shows the cumulative work in hours. However, this measure can also be indicated in number of orders. Two curves are incorporated in the diagram; the cumulative input and output curve. This number can be expanded, facilitating multiple possibilities for diagnosis. The vertical distance between the two curves represents the work-in-process, the
horizontal distance represents average throughput times. In this way, throughput diagrams can facilitate diagnosis concerning the causes of average lateness. For example: a controlled process shows parallel input- and output curves, an uncontrolled process shows input- and output curves which are not parallel. Observed deviations of the controlled situation can have different causes, which are discussed in diagnosing the throughput diagrams. Throughput diagrams are created for the encompassing process, and for the individual workstations of Engineering and Preparation (Preliminary phase), Purchasing, Sheet Metal Production, Production and Outbound. To produce throughput diagrams, several data are needed.
Therefore, a typology of needed data for throughput diagrams has been created (appendix 2).
Order progress diagrams are regarded as “tools which help to explain the variance of lateness by relating the lateness of individual orders to the progress of these orders and to the input and output control decisions affecting this progress”
(Soepenberg et al., 2008). The order progress diagram helps to show how the lateness of orders varies per workstation, and thus helps to reveal variables which negatively influence delivery reliability. While lateness is monitored through each stage of the process, the extent of delay or speeding up of individual orders per workstation can be determined. Figure 3.2 shows an example of an order progress diagram, derived from
Soepenberg et al. (2008). The line indicated with number three is used to illustrate the functionality of the order progress diagram.
Fig. 3.2. Order progress diagram. (Soepenberg et al., 2008)
Each dot represents the completion at a workstation. The first dot represents order
acceptance, the last dot represent order completion.
If an order would be perfectly on time, this line would follow the reference line, which is zero days lateness. However, this is rarely the case. Line number three shows a lateness of -3 after completing order acceptance. This means that in this case, the order is three days early at order acceptance. Furthermore, the order is accelerated in the process, seen that lateness is -6 after completion at the second workstation. This means that at the point of completion at the second workstation, the order is six days early. Then the order is accelerated again, be it not as dramatic as in the previous phase.
After this, something peculiar can be noticed. The order is delayed, indicated by the upward slope of the order progress line between the fourth and the fifth workstation, and at order completion the order is late. This phenomenon has something to do with various production planning and control (PPC) decisions. Diagnosis of order progress diagrams reveals which PPC decisions influence speeding up or delaying orders, thereby determining factors causing both average lateness and variance of lateness.
To determine average lateness (which is the relationship between average throughput times (ATTs) and realized throughput times) per workstation in the situation of Alfa Laval, virtual due dates (VDDs) are set by subtracting average throughput times of the following stage. Figure 3.3 shows how this process is constructed to facilitate the diagnosis of Alfa Laval’s delivery reliability.
Fig. 3.3. Due date calculation.
For example, the VDD of production is calculated by subtracting the ATT of the outbound process from the due date (DD) of an order. Due to complications in the data gathering process, the engineering and preparation processes as mentioned in the main routing order have been aggregated in a
‘preliminary phase’. Furthermore, purchasing orders could not be interlinked with sales orders.
Purchasing orders are naturally related to sales orders. However, purchasing orders do not have a systematical relation with sales orders which makes it impossible to relate individual purchasing and sales orders with each other. Therefore, ATT’s of purchasing orders have been combined with respectively the ATT of production and sheet metal production.
Calculation of the ATT may be influenced by outliers in data sets. For diagnosis in order progress diagrams this is not relevant, since every order is treated individually. However, this does deliver problems when a ‘promised output curve’ is drawn in a throughput diagram. A promised output curve is a curve which shows how much work should be completed at a certain moment in time to meet promised due dates. In that case, cumulative data are used, which are influenced by outliers in data sets because they are treated as a group. Therefore, either the mode, median or average is used to make an estimate of the ATT value, depending on which indicator is most reliable when drawing a ‘promised output curve’. Montgomery and Runger (2007) state that a logical principle of estimation, when selecting among several estimators, is to choose the estimator that has minimum variance (minimum variance unbiased estimator). In order to produce accurate estimates of VDDs, these minimum variance unbiased estimators are calculated per workstation for four separate periods.
3.1.2 Diagnosis framework
To achieve improvements, the importance of the diagnosis phase cannot be overstated. The thesis work of Soepenberg (2010) has partly concentrated on producing a framework for diagnosing delivery reliability performance of make-to-order
companies. This framework is used as a solid foundation for achieving results in the diagnosis phase, from which conclusions can be derived.
The initial stage of the diagnosis framework - determining relevant problem areas - consists of four steps. Figure 3.4 illustrates the initial stage of the diagnosis phase. Four results are possible with respect to the area in which the diagnosis will continue following this initial stage: average- oriented delivery time promising, average-oriented realization process, variance-oriented delivery time promising and variance-oriented realization process.
The first three steps primarily help to focus the diagnosis on relevant subsets and to build a
decision whether the diagnosis should be directed towards average lateness or towards variance of lateness. Subsequent to completing the third step, a decision is made concerning whether to focus on an average-oriented or a variance-oriented diagnosis.
Land (2004) argues that in job shops “the delivery reliability performance must result from a combination of (1) well estimating the throughput times for a job when promising a delivery time to the customer, and (2) controlling job progress such that the promised delivery time is met”. In relation to that, the fourth step concerns the analysis of the process of (1) delivery time promising (DP) and (2) the realization process (RP), where it is determined whether diagnosis should continue in the DP or in the RP. The process of DP produces promised delivery times in response to customer’s enquiries.
The main point of interest in DP is whether the made decisions result in accurate promised delivery times, when compared to realized throughput times.
Fig. 3.4. Initial stage of diagnosis phase.
Achieving high delivery reliability performance requires controlled input and output moments throughout the RP. The RP can be subdivided into pre-shop-floor processes and shop-floor processes.
Pre-shop-floor processes are triggered by the order
acceptance decision. The pre-shop-floor processes
and shop-floor processes are separated by the
release decision, which enables controlled progress
of orders on the shop floor (Soepenberg, 2010). The
main point of interest in the RP is whether the
utilized input- and output control measures are
sufficient promised The g framewor delivery r delivery t This stag of the dia 3.2 Dete
3.2.1 S lateness
Quan times and orders co until 06-1 substantia most imp Orders w complete this data The f the percen relevant p sales orde
Fig
However whether t or varianc distributio figure 3.6 whether t lateness.
dimensio percentag late deliv delivery, delivery.
t to control job d delivery time goal of the sec rk is to determ reliability with time promisin
e is based on agnosis.
ermining the
Step 1: Analysntitative data c d realized thro mpleted in the 12-2010. This al period of tim portant season which are accep
d after 01-05- set.
first action in ntage of order period. Figure ers, 73,24 % h
g. 3.5. Percentag
r, these numbe to focus the di ce of lateness on of lateness 6, giving more to focus on av This distribut ns: the latenes ge of orders. A very, a negativ
and zero laten
b progress suc e is met.
cond stage of mine PPC caus
hin the proble ng or the realiz the results of
relevant prob
sis of the distrconcerning pro oughput times
e period from data set repre me, which inc al fluctuations pted before 01 -2010 are also
the diagnosis rs delivered la e 3.5 illustrate has been deliv
ge of orders del
ers are not suf iagnosis on av . A histogram s has been con e support for t verage lateness
ion is based o ss in weeks an A positive late ve lateness ind ness represent
ch that the the diagnosis ses of low em areas of
zation process the first stage
blem area
ribution ofomised delive are gathered f m 01-05-2010
esents a corporates the
s in demand.
1-05-2010, bu o incorporated is to determin ate in the
s that, for all vered late.
livered late.
fficient to deci verage latenes m showing the nstructed in
the decision s or variance o on two
nd the eness indicates dicates early
ts on time s.
e
ery for
ut d in
ne
ide s
of
s
at sta pe eq tim 4 la pe m nu m su Re pe co sh
re th va A w th pr as co di
3.or
se ge
Fig.
The averag t 16 days (or o andard deviat ercentage of o quivalent to th me. The exces weeks late co teness and var ercentage of o most of the ord umbers clearly maintaining del
ubstantial num educing varian ercentage of o onceptualized hows the expe
Fig. 3.7. Conce
educing varian he distribution ariance, averag
dditionally, on ill cause all or
These resul hat reducing av riority. Howev ssumption, thi ompletion of e iagnosis.
2.2 Step 2:
rder subsets
Two differe elected to anal
eneral, a sales
3.6. Distributio
ge lateness of a over 2 weeks),
ion of 25 days orders delivere he percentage ssive percenta ontributes to a riance of laten orders is delive ers is delivere y indicate seve livery reliabili mber of orders nce will have orders delivere in figure 3.7.
cted effect of
eptualized distr reducing var
nce in this spe
will compres ge lateness wi ne could state rders to be del lts directly lea verage latenes ver, to build m s decision wil either the seco
: Analysis of d
ent subsets of lyze the differ
order is divid
on of lateness.
all orders is de , with a relativ s (3,5 weeks).
ed 1 week late of orders deliv age of orders d
higher averag ness. A small ered early, of ed 1 week earl ere problems ity. Moreover are delivered minimal impa ed late, as
This conceptu f only
ribution of laten riance.
cific case. Alt ss, causing low
ill still be posi e that reducing livered late.
ad to the assum ss has the high more support f
ll be delayed u ond or third ste
differences am
f work orders a rences between ded into two w
etermined vely high . The e is almost
vered on delivered ge
which ly. These
in r, a d late.
act on the ualization
ness after
though wer
itive.
g variance mption hest for this
until the ep of the
mong
are
n them. In
work
orders: on which pa Thus, the subsets or the sheet routing fr productio That is, s metal dep operation system. T are used i these item departme difficultie average l influence departme high aver it may be diagnosis The f complete subset, la promised by the she the actual Hence, th represent these valu reliability The secon through th is calcula of the ord order. He percentag meaning reliability Small dif and figur which are departme percentag 3.9 illustr
Fig. 3.8. P
ne for the shee sses through t e decision to sp
riginates from metal departm rom the order on department eparate work partment. Afte ns, these work The items prod in the finishin ms are produce ent, the produc
es finishing an ateness and va ed by the outpu ent. If one of th rage lateness o e possible to n s to this subset
first subset co d by the sheet ateness is calcu d due dates of
eet metal depa l completion d he percentages t intermediate ues represent y of the sheet m
nd subset com he production ated by relatin der to the prom ence, in this si ges represent o
that these valu y of Alfa Lava fferences are n e 3.9. This is e not handled ent. The first a ge of orders de
rate this for bo
Percentage of o Metal
et metal depar the production pecifically an m the order set ment having a set passing th t, as illustrated orders are ma er completing orders are rem duced by these ng stage of pro ed late by the ction departme n order on tim
ariance of late ut of the sheet he subsets is r or a high varia arrow down th t.
mprises all or t metal depart ulated by relat work orders t artment for pr dates of these s calculated in lateness value the internal de metal product mprises all ord n department.
g the actual co mised complet
tuation the ca overall latenes ues represent al towards the noticed betwee
due to sales o by the produc action is to det elivered late. F oth subsets.
orders delivered l Department.
rtment and on n department.
alyze these t delivered by a different hrough the
d in figure 2.1 ade for the she the sheet met moved from th e work orders oduction. In ca
sheet metal ent has me. Thus,
eness may be t metal responsible fo ance of latenes
he focus of th rders that are tment. In this
ting the o be complete roduction, with
work orders.
n the first subs es, meaning th elivery tion process.
ders passing Here, lateness ompletion dat tion date of th alculated
ss values, the delivery customer.
en figure 3.5 f spare parts ction termine the Figure 3.8 and
d late by the She
ne
. eet tal he s
ase
or a ss, he
ed h set hat
s te he
d
eet
O de th
% of di su in la m de di to de do va ill
F
su or
Fig. 3.9. Perce passing th
f the orders th epartment, 27, hat pass throug
% is delivered l f orders delive ifferent outcom ubset is -5 day n this subset ar teness in the s meaning that av elivered 14 da
The second istributions of o determine wh eviates in such own or directe ariance-oriente lustrate the dis
Fig. 3.10. Distrib S
Fig. 3.11. Dis throug
ubsets. It is no
rders delivered
ntage of orders hrough the Prod
hat are deliver ,33 % is deliv gh the product late. It is noted ered late in bo me. Also, aver ys, meaning th
re delivered 5 second subset veragely, the o ays late.
d action in this f lateness of bo hether either o h a way that fo ed towards an ed diagnosis.
stributions of
bution of latene Sheet Metal De
stribution of late gh the Producti
oted that the la d by the sheet
s delivered late duction departm
red by the shee vered late. Of t
tion departme d that the perc oth subsets hav rage lateness i hat averagely, days early. A is around 14 orders in this s s step is to com oth subsets. T one of the dist ocus can be na
average-orien Figure 3.10 an lateness for b
ess of orders de epartment.
eness of orders on department.
ateness percen t metal departm
of orders ment.
et metal the orders
nt, 73,29 centages ve a very
in the first the orders Average
days, subset are mpare the he goal is tributions arrowed nted or a nd 3.11 both
elivered by
passing
ntages of
ment have
less varia departme deviation is determ lateness i at 26 day first subs lateness i the distrib distributio weeks an is concen large part sense, the respect to
To na
significan percentag departme through p figure 3.1 the sheet reducing on time, c reliability the effect overall de Data anal been deliv 62 % is d orders (99 the sheet productio Large subsets o departme productio determine of latenes of orders pertains p factors, p orders pa reducing effect on delay tak productio subset of metal dep diagnosis productio to improv reduction departme
3.2.3 SFour whether o role in av so, maxim
ance than thos ent. When ana n of lateness in mined at 15 day in the producti ys. However, th
et is -5 days. I is around 14 d butions of late on of lateness nd 0 weeks. In ntrated on the p
t of the orders ey are relative o each other.
arrow the focu nt to determin ges of orders d ent relate to th
production. As 10, around 15 metal departm variance, this causing a dram y of the sheet m tiveness of thi elivery reliabi lysis shows th vered early by delivered late a 9 %) which ar metal departm on.
e differentiatio f orders delive ent and orders on department ed concerning ss and percent delivered by positive outco particularly wh assing through
variance of th overall delive es place in the on department f work orders p
partment is ex s. However, w on order proce ve delivery rel n of variance o ent.
Step 3: Analys
different subs or not time-de verage lateness mum support i
e in the produ lyzing the dat n the sheet me ys. The standa ion departmen he average lat In the second days. This is al eness. In figur s is concentrat figure 3.11, t positive side, s is delivered l ely heterogene us of the diagn ne how the late
delivered by th ose of the ord s can be determ
% of the orde ment is 1 week s 15 % may we matic increase
metal departm is measure wit ility must be d hat of all order y the sheet me after productio re delivered on ment, are deliv on is determin ered by the sh passing throu t. Large differ g average laten
tages of latene the sheet meta mes with resp hen compared h production. I
he first subset ery reliability, e order proces t is part of. Th passing throug xcluded from f when the large ess is removed liability could of lateness in t
sis of differenc
sets are select ependency pla s and variance is given to the
uction ta, the standar etal departmen
ard deviation o nt is determine teness in the
subset, averag lso depicted b re 3.10, the ted around -1 the distribution
meaning that late. In that eous with
nosis, it is eness
he sheet metal der set passing
mined from ers delivered b
k late. By ell be delivere e in delivery ment. Howeve
th respect to determined.
rs that have etal departmen
on. Almost all n time/late by vered late afte ned between th heet metal ugh the
entiation is ness, variance ess. The subse al department pect to these d to the subset
Initially, will have littl since a large ss which the herefore, the gh the sheet further
delay in the d, future action d include the
the sheet meta
ces over time
ted to analyze ays an importa e of lateness. I e assumption
d nt of ed ge by
n a
l g by ed r,
nt, l y er
he
e et t
of le
ns al
ant If
th di nu pe in sig pe
D va sh of
Ta va va la su in de su de de su de de hi an th su th im av ta va lit de ha di av pr
hat diagnosis m irection. Utiliz umber of orde eriods through nfluences in de gnificant devi eriods. The fo 1 All ord
and 30 2 All ord
and 31 3 All ord
and 31 4 All ord
and 06 ata analysis sh ariance of late hows the avera
f the four subs
Tab. 3.1. Avera
able 3.1 demo alues for subse aries over time teness remain ubset. The stan n subsets 1, 2 a eviation of sub ubsets. A devi
etermined. In elivered on tim ummer period
ecreased accor emand. Howev igh outliers, ca nd a high stand Resulting f hat despite the
ubsets must be his situation, ti mportant role.
verage latenes rdy. As conclu ariance with a ttle effect on a elivery reliabi as been provid iagnosis on an verage latenes roduction depa
must focus on zed subsets ar ers which are h hout the order emand. Past sa iations in dem
llowing subse ders complete 0-06-2010
ders complete 1-08-2010
ders complete 1-10-2010
ders complete 6-12-2010
hows differen eness figures o age lateness an sets.
age lateness an four subs
onstrates relati et 1, 2 and 3. T e, while the st ns fairly equal, ndard deviatio and 3. Howev bset 4 does dif
ating characte subset 4, mos me. This is bec
has ended an rding to seaso ver, subset 4 c ausing a posit dard deviation from this step
deviating cha e investigated ime-dependen All subsets pe ss, meaning th
uded in parag a positive aver average latene lity. Therefore ded to the assu n average-orie ss of all orders artment is the
an average-or e based on the handled in cer set, following ales results sh mand between ets are used:
ed between 01 ed between 01 ed between 01 ed between 01
nt average late over time. Tab
nd standard d
d standard devi sets.
ively high late The average l tandard deviat , except for th on does not di ver, the standar
ffer from the o er of subset 4
t of the orders cause the dem d demand has onal fluctuatio contains a num tive average la n.
it can be conc aracter of subs in further dia ncy does not p ertain a positiv hat average lat graph 3.2.1, re rage lateness w ess and thus no e, sufficient su umption to foc nted direction s passing throu
focal point to riented e average rtain g seasonal how
these
-05-2010 -07-2010 -09-2010 -11-2010
ness and ble 3.1
eviation
iations of
eness ateness tion of he fourth
ffer much rd other is s are manding
s on of
mber of ateness cluded set 4, all
gnosis. In play an
ve eness is ducing will have
ot on upport cus the n. The
ugh the
o follow.
3.2.4 Step 4: Analysis Delivery Time Promising or Realization process
Based on the results from the preceding diagnosis steps, the diagnosis focuses on an average-oriented direction. Continuing with the diagnosis model, it is necessary to determine whether the diagnosis will focus on delivery time promising or the realization process. This decision is supported by a throughput diagram of the entire process (figure 3.12). In this figure, four curves are depicted. The order acceptance curve increases with the number of orders that are accepted on a specific day. The order release curve increases with the number of orders released on a specific day. The order completion curve increases with the number of orders completed on a specific day, and the promised output curve increases with the number of
Fig. 3.12. Throughput diagram of entire process.
orders that are promised to be delivered to a customer on a specific day.
The initial finding is that the promised output curve constantly shows a higher value than the order completion curve. For example, on 01-08- 2010 the promised output was 400 orders, while only 350 orders were completed. This confirms the previously measured average positive lateness throughout the whole data set in table 3.1.
Second, the promised output curve significantly pertains an irregular character as opposed to the order completion curve, indicating uncontrolled delivery time promising. This irregularity is determined throughout the whole data set.
However, the sudden shift upwards in the period 1- 8-2010 until 1-9-2010 causes increased average lateness in the period to follow. This is a typical example of uncontrolled delivery time promising, increasing average lateness. For this period, the sales department promised a large number of orders to one customer, without taking capacity into
account. Other orders were put on hold because of the importance of this customer, ultimately resulting in increased average lateness. The irregularities in the order completion curve arise in the period following the sudden shift in the promised output curve, indicating that the
realization process is trying to satisfy the irregularly promised delivery dates. This would lead one to think there is some flexibility in the production process because capacity is not fully utilized.
However, in the period following the sudden shift high overtime labor costs were made. Weekend shifts were put in to try to satisfy the irregularly promised delivery dates. Hence, the realization process, which is related to
the order completion process, has a stable output when overtime labor is not incorporated.
These findings indicate that further analysis must mainly focus on delivery time promising.
3.3 Determining PPC causes within the main problem area
In the previous sub-section, delivery time promising is defined as the main problem area with respect to the low delivery reliability.
A delivery time list, externally imposed from the larger business unit, is used to promise initial delivery times. This delivery time list indicates similar delivery times for Alfa Laval Groningen and other Alfa Laval factories. However, about 80
% of Alfa Laval Groningen’s products are custom- built, while in other Alfa Laval factories this percentage is not higher than 20 %. Since the greater part of Alfa Laval Groningen’s orders is customer-specific, a standard delivery time list merely based on product group characteristics does not seem to be a sensible tool to promise delivery times. Furthermore, when using a standard delivery time list, shop-floor status is not taken into account.
According to Soepenberg (2010), shop-floor status (e.g. the work-in-process level) should be taken into account when promising delivery times, in order to promise accurate delivery times. Using a standard delivery time list when promising delivery times results in a workload which is more or less equally distributed to demand. Since demand is volatile and customer specific in a make-to-order environment, workload will also become volatile. This is shown by the promised output curve in figure 3.12.
After engineering and preparation, a more
accurate delivery time can be estimated. However,
for many orders this actual estimated delivery time
is much longer than the initial promised delivery
time. Then, the organization is hesitant to correct
delivery times towards the customer. Hence, it
starts to put pressure on capacity groups to finish
these orders on time. This influences the planning
of other orders, causing them to be delivered tardy,
setting off a chain reaction.
With several op managem cause actu longer tha Engineeri anticipate a prolong be tempo work-in-p which pro character which are time list.
3.4 Man facto The p through t managem potential indicated prelimina Second, m originatin goal of th these ind the low d unrealisti The f lateness o there is a the prelim diagram investiga In fig curve (ac of the op continuou station cu throughp perspecti near to th constant prelimina of arrivin processin output cu should be (engineer meet the based on by calcul depreciat workstati
h respect to del pen interview ment have indi ual estimated an the initially ing and prepa ed, special com ged material d
rarily unavail process levels oduces capaci ristics of an en e not congruou
nagerial indic ors
previous diagn triangulation.
ment delivered problem facto d lateness poss
ary phase (eng management i ng from the pr his section is t dications and t delivery reliab ic delivery tim first indication occurs at the p a sense that ca
minary phase.
of the prelimi ate this indicat gure 3.13, a ve ctual input) is peration compl usly follows t urve, indicatin put time. Furth ive the operati he arrival at st work-in-proc ary workstatio ng orders and ng them. In fig urve represent
e completed b ring and prepa
promised del n virtual due da
lating backwa ted by the ave ion (see figure
livery time pr ws with Alfa La
cated various delivery time y promised de aration can tak mponents can elivery time, c able from sup s can be taken ity restrictions ngineer-to-ord us with a stan
cations of pos
nosis results h Interviews wi d two indicatio ors. First, man sibly originati gineering and indicated laten roduction pha to quantitative to determine f bility, in additi me promising.
n comprises th preliminary ph apacity is not f
An individua inary phase is tion.
ery irregular a detected. How letion curve (a the slope of th ng a controlled hermore, from
ion completio tation curve, in
ess level. This on can capacit
has little prob gure 3.13, the ts the numbers by the prelimin
aration) at a g ivery date of t ates, which ar ards from the o erage throughp
e 3.3).
romising, aval
factors that e being much elivery time.
ke longer than be ordered w components c ppliers and
into account s. These are al der environmen ndard delivery
ssible problem
have been test ith Alfa Laval ons of other nagement ing from the
preparation).
ness possibly ase. Hence, the
ely investigate factors causing
ion to
he opinion tha hase, because fully utilized i al throughput
used to arrival at statio wever, the slop actual output) he arrival at
d average m a vertical
on curve remai ndicating a s means that t tate the numbe blems
promised s of orders tha nary phase given time, to
the order. It is re determined order due date put times per
with can
ll nt,
m
ed l
e e g
at in
on pe
ins he er
at
s e
W (a fig pr ab at or ar du tim nu or 30 co th w pr th ca de
Fig. 3.13. Th
When comparin
actual output) gure 3.13 show romised outpu bove the order t a given mom rders have bee re have to be c ue dates (on th mes). Howeve umber of arriv rders. The num 00 orders. Sin omplete more he number of p hen orders arr romised delive he preliminary
apacity manag elivery times.
Fig. 3.14. Th
hroughput diagr
ng the order c with the prom ws different p ut curve predo r completion c ment in time, e
en completed, completed in o he basis of ave er, at that mom vals is slightly mber of order ce the worksta
orders than th promised outp rive according ery times are u y phase does n gement but fro
hroughput diagr
am preliminary
ompletion cur mised output cu
patterns, with t ominantly pres curve. This me .g. 15-06-201
whereas 400 order to meet erage through ment in time, t y higher than 3 completions i ation can neve he number of a put cannot be p g to plan. Hen
unrealistic. La not originate fr om unrealistic
ram production
y phase.
rve urve, the siding
eans that 0, 300
orders promised hput
the 300 is around er arrivals, produced
ce, ateness in rom poor
promised
phase.
The second indication comprises the opinion that an increase in lateness is caused by insufficient capacity utilization in the production phase. Figure 3.14 shows a smooth arrival at station pattern and a smooth operation completion pattern for the production department. It is remarkable to see the difference between the operation completion curve of figure 3.13 and the arrival at station curve of figure 3.14. In theory, the arrival at station curve in figure 3.14 should follow the pattern of the operation completion curve in figure 3.13.
However, the arrival at station curve in figure 3.14 is much smoother, hence more controlled. This controlling occurs when the purchasing department (the intermediate department between the
preliminary phase and the production phase) plans its purchasing orders to the capacity of the production department. The minimizing of inventory costs is the main reason for this
smoothing. Controlling the input for production has positive effects on the output of production. The operation completion curve shows a stable, predictable output, with a minimal increase in work-in-process (vertical distance between arrival at station and operation completion curves) and average throughput time (horizontal distance between the same curves) in the eventful summer period. Also, output of the production department is usually at maximum capacity, provided they are supplied with the right materials on the right time.
However, the promised output curve constantly presides above the operation completion curve.
This means that throughout the measured time, more orders should have been completed in order to satisfy promised due dates.
The reason for this is two-fold. First, as shown in figure 3.13, a large proportion of orders is completed late by the preliminary phase, originating from unrealistic delivery time promising. This lateness causes late arrivals for production. Second, 19 % of purchasing orders are delivered late. A distribution of lateness is shown in figure 3.15 in order to visualize this percentage.
The purchasing orders which are delivered late cause increased variance of lateness. Planning adjustments are made to allow other orders to come first in case of tardy purchasing order deliveries.
However, since the output of the production
Fig. 3.15. Distribution of lateness Purchasing orders.
department is usually at maximum capacity, overall average lateness, and thus delivery reliability, is not influenced by the lateness of purchasing orders.
3.5 Diagnosis outcome
In section 3.3, the inaccuracy of the externally imposed delivery time list came forward as the main factor causing the low delivery reliability of Alfa Laval Groningen. Delivery times are promised too tight in comparison with average throughput times. The list with standard delivery times based on product groups does not coincide with the engineer-to-order character of 80 % of the products of Alfa Laval. Furthermore, the remainder of 20 % is make-to-order, which also does not coincide with a standard delivery time list. No integration takes place between sales and operations during the quotation process of delivery times. Often, complications occur because of the nature of a product, causing longer lead times. Despite this, the sales department uses the standard delivery time list, which contains delivery times that are too short in comparison with average throughput times. This causes orders to be late, even before they have started their process through the main routing order.
Processes are put under pressure to deliver late orders faster, which influences the planning of other orders, causing them to be delivered tardy, which increases variance of lateness.
Also, the diagnosis shows that when work orders are completed on time/late by the sheet metal department, 99 % of these orders are eventually delivered late as a sales order. This is in sharp contrast with the work orders which are completed early by the sheet metal department. Of those work orders, 62 % is eventually delivered late as a sales order. Because of the maximum capacity utilization in the production department, this is currently not of influence on delivery reliability. However, it does influence the variance of lateness. As demonstrated in figure 3.7, reducing variance of lateness with a positive (tardy) average lateness has little influence on delivery reliability. However, in the situation of a negative (early) average lateness, delivery reliability can be improved by reducing variance of lateness. Hence, when average lateness should be decreased in the future, this variance of lateness will have more and more influence on delivery reliability.
Furthermore, a managerial indication has
pointed out a factor in the realization process
causing increased variance of lateness. In the
purchasing process, 19 % of the purchasing orders
are delivered late. At the moment, this does not
influence delivery reliability because production is
operating at maximum capacity. Again, should
average lateness be decreased in the future, this
variance of lateness will have increasing influence
on delivery reliability.
Figure 3.16 shows a causal model, which depicts the factors causing low delivery reliability and the relations between each other.
Fig. 3.16. Causal model.
