Case Study, Optimizing Order Fulfillment in A Global Retail

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

Int. J.

Production Economics 127 (2010) 278–291

Contents lists available at ScienceDirect

Int. J. Production Economics


journal homepage: www.elsevier.com/locate/ijpe

Case study: Optimizing order fulfillment in a global retail


supply chain
Yousef Amer , Lee Luong, Sang-Heon Lee
School of Advanced Manufacturing and Mechanical Engineering, University of South Australia, Mawson Lakes, SA 5095, Australia

a r t i c l e i n f o abstract

Article history: This paper demonstrates a model which uses an adaption of Design for six sigma and fuzzy
Received 31 December 2008 logic to optimize, monitor and control the order fulfillment process in the supply chain of a
Accepted 12 August 2009 global retail firm. Even though the firm has advanced SCM practices in place retail stores
Available online 27 August 2009
suffer from stock outs of popular items and excessive stock of lower selling items. The
Keywords: order fulfillment optimization model presented enhances supply chain integration and
Supply chain management collaboration across supply chain partners through effective monitoring and controlling of
Performance measurement supply chain variables. It considers the critical to customer requirements at the supply
Process control chain design onset making it a useful model for dealing with customer differentiation and
Order fulfillment
channel separation. With 47% of firms only focusing on developing smooth internal
Design for six sigma
processes this model offers opportunities for external supply chain improvements and
Fuzzy logic
partial integration across the chain to improve business outcomes.
& 2009 Elsevier B.V. All rights reserved.

1. Introduction approaches that encourage greater uptake of supply chain


management, integration and process improvement
A key issue facing companies today is how to monitor (Branch, 2009; Xu et al., 2009).
and control performance across the supply chain. Taking One of the difficulties in managing a supply chain is
incremental steps towards supply chain integration can their complexity, having to deal with multiple market
lead to competitive gains and focusing on key supply segments (McKay, 2003) with each segment a system in
chain processes can be a starting point for improved itself, but having to closely interconnect with dependant
performance and collaboration between chain members subsystems. Supply chain models such as the global
(Aryee and Naim, 2008; Lummus et al., 2008; Bask and supply chain forum (GSCF) model (Croxton et al., 2001;
Juga, 2001; Lambert et al., 1998). However, there is a lack Croxton, 2003), the supply chain operational reference
of managerial techniques and tools for SCM which may be (SCOR) model (Supply-Chain Council, 2009) and the
inhibiting the uptake of SCM amongst firms. Branch Collaborative Supply Chain Framework (CSCF) (Simatupang
(2009) states that most companies, 47% of them, are and Sridharan, 2005) take an holistic approach to SCM but
working to create seamless processes within their own remain generic in nature and one size does often not fit all.
organization only, 34% of companies focus on integration Another consideration is that while these models refer to
with first tier suppliers, 11% of companies focus on the development of supply chain metrics they do not
integration with key customers and only 9% integrating describe how to determine specific metrics for individual
logistics up and down stream between customers and company circumstances. The SCOR, relies on benchmark-
suppliers. Hence there is great scope to develop new ing and has recommended level one metrics but there is
some criticism that the model is too manufacturing
centric (source, make, deliver, return), that there is not
 Corresponding author. enough emphasis on process control (Bolstorff, 2003a–c)
E-mail address: Yousef.amer@unisa.edu.au (Y. Amer). and an overreliance on benchmarking which can inhibit

0925-5273/$ - see front matter & 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.ijpe.2009.08.020
Y. Amer et al. / Int. J. Production Economics 127 (2010) 278–291 279

innovation. Gunasekaran et al. (2004) defines supply improve processes with a strong focus on results. To
chain metrics as strategic, tactical, and operational and increase system reliability and reduce failure rates, the
gives generic measures, but the task still remains for components utilized in the complex system and products
managers to determine what the most appropriate have to have individual failure rates approaching zero
measures for their specific circumstances are. This often (Summers, 2007).
depends on the configuration of the chain and the position Six sigma stands for a measure of customer quality as
of the firm within the chain. The model presented well as a philosophy of giving customers what they want
provides a means for creating specific targeted metrics each and every time i.e. zero defects. Anything less than
and a way of isolating where process and quality ideal is an opportunity for improvement; defects costs
improvement efforts should be focused from any position money and understanding processes and improving them
within the chain and is not reliant on any specific supply is the most efficient way to achieve lasting and cost
chain configuration to implement. effective results (Summers, 2007; Brusse, 2006; El Haik
The purpose of this paper is to demonstrate a real case and Roy, 2005). It is also a methodology that can be used
application of the order fulfillment model which uses an to change processes and company culture. It is based on
adaption of design for six sigma (DFSS) (Amer et al., the premise that all actions are a process. Each process has
2007a) and fuzzy logic (Amer et al., 2008). In the case a start, a stop, inputs i.e. from supplier and outputs i.e. to
study the model was applied at a large international retail customers, and actions that happen during the process
firm which has an extensive supply chain. Through the steps, i.e. subprocesses. Each process has measurable
application of the model by a cross functional supply characteristics as each other processes change entities and
chain design team the order fulfillment subprocesses and this change can be measured. Measurements can be of
key performance indicators (KPIs) are identified and the input or output characteristics such as number, weight, or
order fulfillment variables as key process input variables type. They can also be process characteristics at various
(KPIVs) and key process output variable (KPOV) is stages such as count and time taken. Measurements can
identified, showing the ‘‘perfect order’’ as one of the be of continuous data items such as time, money, size or
critical to customer requirements (CCRs). A transfer they can be of discrete data items such as integer counts.
function was then developed for the perfect order and a The process itself will have requirements for the inputs,
membership function assigned to measure the gap and the customer will have requirements of the outputs.
between the current performance and the perfect order. Measures follow a frequency distribution showing how
The presented transfer function of ‘‘perfect order’’ was many measures fall within a given range of data such as in
formulized and refined using fuzzy logic. A greater a frequency histogram. Under DMAIC the problem is first
understanding of the global supply chain was grasped by defined, and quantified, measurement data are collected
participants and a response was triggered across the chain though root cause analysis and solutions to the root causes
which aided further collaboration for process control and are determined. Finally improvements undergo monitor-
improvement in supply chain outcomes. The demon- ing and control to prevent reoccurrence.
strated model can be applied to any supply chain process A key limitation of the six sigma DMAIC methodology is
(Amer et al., 2007b) and from any position within the that it is focused on defect correction and is a reactive
supply chain making it a very flexible model for supply methodology. It is often used when the existing processes
chain management. do not satisfy the customers or are not able to achieve
The paper is organized as follows. A literature review of strategic business objectives (Eckes, 2001). DMAIC is
six sigma and DFSS is given to develop an understanding increasingly being applied in supply chain improvement
of the approach taken. The order fulfillment model is projects for this purpose (Yang et al., 2007; Zhang et al.,
described followed by the research methodology, a brief 2007) and there is interest in combining DMAIC with lean
description of the case firm and the application of the techniques termed ‘‘lean six sigma’’ (Martin, 2007;
model. In conclusion the results and findings are dis- Bolstorff and Rosenbaum, 2007; El Haik and Al Aomar,
cussed and recommendations made for further research. 2006; George, 2003). It has also been incorporated with
SCOR to address the lack of root cause analysis and process
improvement in that model (Bolstorff and Rosenbaum,
2. Literature review 2007; Bolstorff, 2003a–c, 2004). However, these remain as
quality improvement initiatives. The reality is even if zero
There are two major improvement methodologies defects are achieved, it does not mean the product or
under six sigma umbrella. Define, measure, analyse, service will satisfy the customer’s requirements. For
improve and control (DMAIC) is used for continuous example if you are producing the perfect product i.e.
improvement of already existing products/processes and water bottles with pure and high quality water but if the
DFSS is used for designing new product/processes. Six customer perceived a scratch on the water bottle as a
sigma is a highly disciplined, data-oriented, top-down defects then the product is not going to sell well.
approach to quality management that brings together In contrast, the second six sigma methodology, DFSS, is
elements from past quality initiatives and aims to reduce proactive as it focuses on getting things right the first time
variability, focus on what the customer wants and and continuing to do them right all the time. El Haik and
improve core processes (Klefsjö, et al. 2001; Summers, Roy (2005) state DFSS is the same whether it is used to
2007). A key feature of six sigma is the use of statistical accommodate high customer satisfaction externally or
techniques in a systematic way to reduce variation and whether it is focusing on an internal process facing the
280 Y. Amer et al. / Int. J. Production Economics 127 (2010) 278–291

employees, it can proactively produce highly consistent 2005; Brue and Launsby, 2003; Yang and El Haik, 2003), a
processes with very low variation in performance. Given generalized DFSS flow follows the steps of identify, define,
the limited uptake of supply chain frameworks and design, optimize and verify or validate (IDDOV). The
models amongst firms the design of supply chains in inputs can be customer needs, business needs, raw
many industries exhibit deficiencies like modest levels of materials, constraints and so on. The outputs can be
quality, unawareness of what the customer really wants, quality products, processes or services designed to reach
and too much complexity. Such shortcomings add a six sigma levels. Going through these phases each CCR is
considerable amount of non-value added activities to presented as a function (transfer function) of a number of
SCM and can mainly be attributed to the lack of a systems KPIVs, which can be used to predict the performance level
design methodology. of the CCR. The following sections present a brief
Implementation of DFSS has successfully delivered introduction to the sequence of major activities covered
quality products to customer satisfaction for many indus- under IDOV phases of DFSS in the order fulfillment model
tries. DFSS has successfully been implemented for product as shown in Fig. 1.
design in a number of leading manufacturing firms such as
Motorola, GE, Allied Signals, Honeywell, and Seagate, and
now there is increasing interest in designing a service such 3.1. Identify
as a supply chain through DFSS (Amer et al., 2007a, 2008;
El Haik and Al Aomar, 2006; El Haik and Roy, 2005). The major activity in this phase is the collection and
translation of often vague and abstract customer require-
ments also known as voice of the customer (VOC). The
3. Order fulfillment model using DFSS and fuzzy logic supply chain design team identifies the customers
potential needs, enlists and prioritizes their expectations
While DFSS is an open methodology with numerous through data collected from customer interviews, focus
interpretations suiting varying situations (El Haik and Roy, groups, surveys, complaint data and enhancement request

Fig. 1. Order fulfillment model showing IDOV phases of DFSS.


Y. Amer et al. / Int. J. Production Economics 127 (2010) 278–291 281

or through market research. Using tools like quality necessary to achieve a robust product/service design. A
function deployment (QFD), VOC is translated into design transfer function is a mathematical representation of the
specifications termed as CCRs, comprehendible to the relationship between the input and output of a system or
supply chain design team (Buss and Ivey, 2001). While a process. It facilitates the DFSS optimization of process
QFD is the major tool in identify phase, there are other output by defining the true relationship between input
tools that may be used, such as kano model, pareto variables and the output. Optimization in this context
analysis, affinity diagram, brainstorming and benchmark- means minimizing the requirement variability and shift-
ing (Rath and Strong, 2002). QFD is a structured approach ing its mean to some desired target value specified by the
to defining customer needs or requirements and translat- customer. The transfer function is formulized and refined
ing them into specific plans to produce products/services with the use of fuzzy logic. Usually transfer functions
to meet those needs. This is accomplished by a cross resulting from DFSS implementation are expected to have
functional or multidisciplinary ‘‘supply chain design’’ a mathematical representation. However, in this model
team who can use a series of approaches to deploy the the refined transfer function is presented as sets (se-
CCRs throughout the design development (Yang, 2008; quences) of fuzzy logic rules evaluated using the fuzzy
Taguchi et al., 2005). Customer voices are diverse. Hence, logic toolbox in MATLAB (Amer et al., 2008).
there may be a variety of different needs and there can
even be multiple customer voices within an organization. 3.4. Validate
This demonstrates the ability of six sigma tools and QFD
to address the needs of complex and variable supply chain At the final stage validations are performed to check
environments. A supplier may be supplying to multiple that the process is complete, valid and will meet
firms, each with their own requirements. By applying QFD requirements in practice. It involves verification of the
to each channel, specific customer needs can be met with design to ensure that it meets the set requirements;
the aim of achieving zero defects through DFSS. Where assessment of performance, reliability, capability, etc. At
benchmarking occurs as in the SCOR model, for example, a this stage simulation may be used to test the order
company may benchmark their delivery time as matching fulfillment design. If this stage suggests that the design
their competitors results of ‘‘zero defect’’. However, their does not or may not meet the required capability, then it is
customers may have different requirements to the necessary to retract back through the earlier stages of
customers of the firm they are benching marking design or optimize. If the validation results are satisfac-
themselves against. Where benchmarking is used in Six tory, development of a process control plan for the mean
Sigma, it is combined with QFD (Summers, 2007) which and variance of CCRs (Rath and Strong, 2002) are initiated.
allows for innovation to occur along with comparison to
competitors.
4. Case study methodology

3.2. Design A case study application of the model was conducted


over a period of four months at the retail store of Company
After having translated the VOC into engineering terms A. This involved an initial survey of functional managers to
as CCRs, the supply chain design team has the task of obtain a picture of the current performance measurement
proposing and evaluating conceptual solutions to address system and supply chain relationships. Secondly, the stock
the above CCRs. Use of the ‘‘functional design’’ approach of control (SC) and goods administration manager, logistics
systems engineering and multiple functional system manager, finance manager and operations manager were
models (or conceptual solutions) are proposed in this then interviewed to obtain a picture of the supply chain
phase. Each solution is presented as a hierarchy of structure and order fulfillment process. The DFSS and
functional subsystems. The functional concept designs fuzzy logic model was applied by a small supply chain
are evaluated to narrow down solution options. Measure- design team consisting of the logistics manager, stock
ment metrics and methods to inspect the design are control manager and replenishment manager. The infor-
verified prior to detail designing. Having narrowed down mation gathered through this process, company docu-
the number of conceptual solutions, requirements are ments and historical data are now presented.
followed down from top to lowest hierarchical level of the
functional system models. As a result of this exercise, the
4.1. Description of case Company A
team is able to translate the high level system CCRs to
lower level KPIVs.
As an international retailer the firm, Company A, in the
fiscal year of 2005, had sales of $17.4 billion. It employs
3.3. Optimize over 90,000 people in some 44 countries. Company A
owns or franchises approximately 220 stores in 24
Performance of real world products/services is always countries in Europe, North America, Asia and Australia.
shadowed with variation. DFSS therefore, stresses pre- The case retail store is a franchise and employs over 400
dicting and optimizing the probability of the design to people with an annual turnover of approx. $14,500,000.
meet the required targets (CCRs) given environmental The company must secure that the customer promises
variation, manufacturing variation, and usage variation. of lowest price, availability and quality are fulfilled.
Statistical analysis and optimization of the design is Company A is divided into several different companies
282 Y. Amer et al. / Int. J. Production Economics 127 (2010) 278–291

and divisions worldwide. Fig. 2 shows the firms supply strategy to provide a high service to the stores simulta-
chain structure with distribution methods. neously with low transportation costs. The DCs work with
The Head Company is the core of the organization and the retail stores to shorten lead times and make them
is situated in Europe. The supply process commences here stable. The DC is responsible for keeping track of the
consisting of forecasting future demand, ordering, produ- inventory level at the DC therefore their inventory level is
cing and delivering the offer to the sale floor (retail) and sent to the database. Good communication is essential to
ultimately to the customer. The head company has the top enable the distribution of the required volume of products
responsibility of sourcing and contracting suppliers. It is at the right time to the right store. The DC for the case
also responsible for setting correct maximum and mini- retail store is in Singapore. Company A is using other
mum level of inventory of the distribution centres (DCs). different types of distribution systems; order point
The Trading Company is responsible for all the trading distribution centres (OPDC) and direct delivery (DD) as
done by the Head Company. The trading company is well as VMI. The case retail store is focused on selling
divided into 16 different trading areas worldwide. The products. The information that is sent by the store to the
trading company works with over 1500 suppliers in over Head Company database is the inventory level and sales.
55 countries. Each trading area is responsible for a
geographic area and the suppliers within that area. They
are in daily contact with the suppliers. They ensure that 4.2. Current KPIs of Company A
the supplier is able to produce the products in terms of
best quality and function at the lowest price. The Trading The current KPIs are indicated in Fig. 3. Once the
Company office for the case retail store is in Singapore. delivery is received order quality, quantity are checked;
The DCs stock the entire product on the stock list with the however, bad quality product are still accepted and
transferred to the recovery department where the team
tries to assemble the product for it to be sent to the
Headquarters, discount department to be sold as a defect product.
design and
marketing Whether the order quantity is over or under the requested
Trading order the delivery will still be accepted and no action is
Company
VMI taken. The delivery time is often late and even if on the
other hand it arrives earlier than requested, it is still
Supplier accepted and no action is taken. Finally most of the orders
received are missing high demand requested product
whilst carrying excess of the low demand products.
Direct Delivery (DD) The lead time goal is that 95% of all orders are
delivered within the agreed lead time of six weeks with
a window of 74 days. The total lead time is the lead-time
Customer Retailer Distribution
measured from the date the order is created until date it is
Centre received in the store. Lead-time takes into consideration
of retailer conditions, handling time at wholesaler, filling
Fig. 2. Company A’s supply chain structure with distribution methods. rate of transports and transport structure. Manifest

KPI’s Goals Responsible Freq.

1 Service levels in S1 min 99% Sales weekly


retail store S2 min 95% Manager
S3 min 90%

2 Lead times 95% Business weekly


(6 week lead time) 95% Support weekly
-DC delivery +- 4 days Manager
-DD deliveries

3 Load quality <1% Logistic monthly


Managers

4 Load quantity 5% either way Logistics monthly


of order Manager
quantity

5 Manifest accuracy ≤1% Logistics Each


Manager order

Fig. 3. Current KPIs Company A.


Y. Amer et al. / Int. J. Production Economics 127 (2010) 278–291 283

accuracy consists of delivered quantities to be equal to the external aspects of order fulfillment. Internally, the store
notified quantities on the delivery note (manifest) r1%. customer wants to be able to walk into the store and
Exceptions to the order quantity are the ‘‘rounding up’’ purchase their chosen product, take it home on the day in
parameter. If the order quantity is more or equal to 70% of a perfect condition at the right price. The store wants to receive,
full pallet, a full pallet will be delivered. The minimum order from an external source, in this case from the DC, a perfect
quantity is always a multi-pack. If the ordered quantity is order, arriving on time, in good quality and in the correct
less than a multi-pack a full pallet will be delivered. quantity. The team proceeded to develop a QFD. Also known
as ‘‘house of quality’’, QFD is a matrix based tool that takes
5. Implementing the order fulfillment model customer ‘‘whats’’ as input and transforms them into design
team’s ‘‘hows’’. The prioritized list of ‘‘whats’’ is correlated
5.1. Identify with the proposed list of ‘‘hows’’, while each relation is scored
on the basis of relation strength. The cumulative score of
‘‘hows’’ covers their mutual interactions and delivers a
The supply chain design team identified the current
prioritized list of ‘‘hows’’, which can be used as the input
problems at the retail store that need to be considered as
for a next QFD iteration. The final outcome of a QFD exercise
should be a prioritized list of CCRs as mentioned above (Rath
1. Excess inventory of low demand product.
and Strong, 2002). Indeed the technical knowledge, experi-
2. Stock outs occur on many top selling product lines.
ence and innovativeness of the design team cannot be
3. Whilst KPIs are in place there are not acted upon.
overlooked as they would form the basis of ‘‘hows’’. Fig. 4
4. There is no continuous improvement focus.
depicts a first level QFD output translating the subjective VOC
into meaningful ‘‘hows’’ for supply chain designers. In the
5.2. Translation of VOC QFD matrix, the relation intensity (Rij) of ‘‘whats’’ with
potential ‘‘hows’’ is indicated with numbers 1, 3 and 9, where
The goal of Company A is that the ‘‘range offer’’ has to be 1 shows a very low, 3 a moderate and 9 indicates a significant
available at all times in all stores and to all customers, with relation. A blank relation entry depicts no existing relation of
the lowest total supply cost at point of sale. The primary corresponding ‘‘whats’’ and ‘‘hows’’. The cumulative weight
concern of the team is to meet their customer needs by (CWj) for each ‘‘how’’ is indicated as ‘‘importance of hows’’ in
reducing/eliminating the stock outs. The KPIs for customer Fig. 4 and is calculated using the relation depicted in Eq. (1).
satisfaction for product availability have been averaging 66%, Where Wi indicate the importance assigned to each ‘‘what’’
with product quality at 70%. The store manager has requested on the scale of 5–1 (5 being highest):
everyone work towards improving these figures but has not
given any direction as to how this can occur. The other X
n
concern is that there is an over abundance of low demand CWj ¼ Wi Rij ; j ¼ 1; . . . ; m ð1Þ
product. The team chose to initially focus on the internal and i¼1

HOWs / Outputs
Undamaged deliveries

Meet demand quantity


Meet urgent demands
Kano Classification

Accurate invoices
Importance (1-5)

Delivery on time

Order flexibility
No extra costs

Transport

WHATs / Inputs

Direction of Improvement
Order Fulfillment 5 9 9 3 1 3 9 3
Cost 4 3 3 9 1 3 3
Product Quality 3 1 9 1 3
Quality Service 2 3 3 9 3 1
Environment friendly 1 9

Importance of the HOWs 66 63 51 36 18 21 62 45


% Weight 18 17 14 10 5 6 17 12
Importance in next level 5 5 4 3 1 2 5 3

Fig. 4. Supply chain design teams QFD.


284 Y. Amer et al. / Int. J. Production Economics 127 (2010) 278–291

For DD
supplier
fills order

Purchase Authorize Place Order


TC receive TC process
suggestion order order received
order order
order warehouse

DC fills
order

Fig. 5. Order fulfillment process flow, moving externally.

Fishbone analysis of making order

Purchase
suggestion order Direct Delivery (DD)
(PSO) from supplier

Place order
SC places PSO 2
Fulfill vol.,
weight, order Authorize order
window
Navision
parameters Create order
Check Navision
Order ordering
quantities parameters
alignment
Perfect order
Stock control/admin
CFO authorizes Sends order SC orders using
1
order ordering routine
SC obtains
with wish list
LM reviews and confirmation
signs off order from CRS
SC notifies Logistics SC evaluates the forecasted
manager (LM) and Chief quantity increase
Stock controller (SC) financial Officer (CFO),
Creates order, SC Receives and completes
internal transaction
form from Sales
Placing order Promotional 3
product ordering
(PPO)

Fig. 6. Fishbone of making an order with actions noted.

Through the QFD the team determined order fulfillment as order. The order fulfillment process which moves exter-
CCR and delivery time, product quality, quantity of delivered nally was then mapped see Fig. 5.
goods and manifest (invoice) accuracy as the key aspects as to As the major involvement internally in this process
how this will be achieved. flow was making the order it was further analysed
through a fishbone structure, see Fig. 6. The areas for
5.3. Design order fulfillment system structure action were identified as

The team then considered process flow maps and


fishbone arrangements to determine where improvement 1. placing order;
efforts should be focused and what are the KPIVs. The 2. placing direct delivery (DD) order and
initial flow was of the order fulfillment system segment 3. promotional product ordering.
Y. Amer et al. / Int. J. Production Economics 127 (2010) 278–291 285

Action Responsible Timeframe KPIV

1 Review placing order Logistics 2 weeks Order dwell


procedures to ascertain if Manager time
procedure time could be Order
reduced processing
time
2 Navision parameters to be Operations 1 month %Navision
reviewed with Head office Manager parameter
and Trading centre accuracy
%Order
Quantity
accuracy
3 Evaluation of forecasted Stock controller 4 weeks % order
quantity increase, review and Sales and quantity
current practise Marketing accuracy
Manager

Fig. 7. Actions determined by design team for making order.

Scrap

Defect product Repair ASIS Customer


picks order

Goods Replenishment Pick location


Buffer
Receiving Orders replenishment

Customer pick
order

Fig. 8. Internal order fulfillment process flow.

Order accuracy on low demand product was ques- delivery/distribution centre, receiving and de-stuffing, AQIS
tioned. Given the ‘‘rounding up’’ parameter the design and stock control, buffer and DR, pick up and returns. DFSS
team considered if this was what was leading to increased was used to develop performance targets related to CCRs of
inventory of these products. Given the stock outs of high these subprocesses. Through brainstorming, the supply chain
demand products forecast accuracy from the retail end design team listed the KPIVs shown under each process step.
was in question, along with internal order fulfillment and KPOVs for each subprocess are also determined and listed on
the supplier’s ability to meet demand. Actions to be taken top of the process flow. This process enabled a merging of
to further investigate these areas were agreed. KPIVs were both the internal and external aspects of the order fulfillment
determined and goals set as seen in Fig. 7. process.
The team then reviewed the internal order fulfillment The overall KPOV for the order fulfillment process was
process flow, see Fig. 8, identifying that considerable best described as the ‘‘perfect order’’. Its elements, on-
internal problems were occurring at goods receiving time delivery, quantity of delivered order, quality of
which could be affecting access to received stock. delivered order, and manifest accuracy as identified in
Further fishbone analysis highlighted several areas that the KPIVs, make the ‘‘perfect order’’ the key metric for
were targeted for action and KPIVs were noted. monitoring the order fulfillment variables.
The perfect order for Company A is represented as
5.4. CCR flow down of perfect order
Perfect order
As depicted in Fig. 9 the order fulfillment process was
¼ f ½delivery time ðDTÞ; quality ðQualÞ;
described by a sequence of subprocesses, such as purchase
suggestion order, authorize/place order, trading centre, direct quantity delivered; ðQtyÞ; manifest accuracy ðMAÞ ð2Þ
286 Y. Amer et al. / Int. J. Production Economics 127 (2010) 278–291

Fig. 9. Order fulfillment process flow with KPIVs/KPOV.

A marginally short or over order or a slight decline in the Table 1


delivered quantity, or slight delayed/early delivery may Ranges for ‘‘order’’ output and linguistic value.
raise alarm if the factors are measured in specific metrics.
Linguistic value Scores
In real world metrics to measure the performance or
customer’s level of satisfaction from the ‘‘perfect order’’, Optimum 0.90–1.00
DT, Qty, Qual and MA is a subjective matter. Therefore the Satisfied 0.70–0.89
above equation (transfer function relating perfect order to Acceptable 0.50–0.69
Poor 0.30–0.49
DT, Qty, Qual, MA) is optimized with use of fuzzy logic. Reject 0.00–0.29

5.5. Transfer function


them via the rule base in order to obtain a fuzzy value for
The transfer function presented in Eq. (2) of ‘‘perfect the output. The rule evaluation followed the form of if
order’’ was formulized and refined with the use of fuzzy logic. (condition x) and (condition y) then (result z) rules are
The refined transfer function is presented as sets (sequences) applied. Basically the use of linguistic variables and fuzzy
of fuzzy logic rules evaluated using the fuzzy logic toolbox in IF–THEN-rules utilize the tolerance for imprecision and
MATLAB (Amer et al., 2008). The steps for assessing the uncertainty mimicking the ability of the human mind to
‘‘perfect order’’ using the fuzzy inference system were summarize data and focus on decision-relevant informa-
fuzzification, rule evaluation and defuzzification. tion and are generated from expert knowledge.

5.5.1. Fuzzification 5.5.3. Defuzzification


The fuzzification process was performed during run The fuzzy inference system using Mamdani’s fuzzy
time and consists of assigning membership degrees implication rule determined the appropriate fuzzy member-
between 0 and 1 to the crisp inputs of delivery time, ship value. The defuzzification process consisted of combin-
quantity and quality and manifest accuracy. ing the fuzzy values obtained from the rule evaluation step
and calculating the reciprocal in order to get one and only
5.5.2. Rule evaluation one crisp value that the output should be equal to. The
The rule evaluation process consisted of using the output ‘‘order’’ was evaluated in relation to the crisp value
fuzzy value obtained during fuzzification and evaluating and translated into linguistic terms (see Table 1).
Y. Amer et al. / Int. J. Production Economics 127 (2010) 278–291 287

Fig. 10. Graphic layout of transfer function in fuzzy format.

Table 2 gap from the optimum and the processes and perfor-
Manifest accuracy. mance needs significant improvement. Order value be-
tween 0.70 and 0.89 indicates a low level of satisfaction
Fuzzy set Linguistic term Range (% of defects)
aiming to close the gap towards optimum. An order
1 Poor Below 4% output value between 0.90 and 1 indicates that the order
2 Acceptable From 1.5% to 3.5% is fulfilling the requirements and requires the processes to
3 Optimum Above 1% be monitored and maintained as shown in Table 1.

5.6.3. Fuzzy evaluation rules


Table 3 demonstrates examples from the 441 rules
5.6. Optimize—fuzzification of perfect order transfer
which follow the format ‘‘if (condition x) and (condition y)
function
then (result z)’’ corresponding to the combinations of
input conditions. For example: if manifest accuracy is
Fig. 10 depicts the empirical transfer function from
‘‘poor’’, if quality is ‘‘poor’’, if delivery time is ‘‘very early’’
Eq. (2) as a fuzzy logic system with inputs and output
and if quantity is ‘‘low’’ then order is ‘‘reject’’. The rules
being fuzzified using appropriate membership functions.
are determined through the expert knowledge of the
The following sections narrate the component of manifest
design team and are further refined following real life
accuracy as an example.
application and reappraisal which will either confirm
them or require them to be modified.
5.6.1. Manifest accuracy
Manifest accuracy value was measured based on the
5.6.4. Fuzzy solution results
delivered quantities being equal to the notified quantities on
A continuum of fuzzy solutions for Eq. (1) is presented
the delivery note (manifest). There are three categories of the
in Fig. 11 using the fuzzy tool box of Matlab. The four
manifest accuracy, namely, ‘‘poor’’, ‘‘acceptable’’, and ‘‘opti-
inputs can be set within the upper and lower specification
mum’’ as shown in Table 2. Any defect rate equal or below
limits and the output response is calculated as a score that
4% is considered ‘‘poor’’ and rejected. Any order defect rate
can be translated into linguistic terms as per Table 1 i.e. an
from 1.5% to 3.5% is considered acceptable. Any defect rate
order output of 0.818 indicates ‘‘satisfied’’ linguistically.
above 1% to 0 is optimum. The range between ‘‘optimum’’
A further graphical representation example as re-
and ‘‘poor’’ is considered ‘‘acceptable’’. ‘‘Acceptable’’, shows
sponse surface for the order is presented in Fig. 12.
that the manifest accuracy needs to be improved aiming
However, given the limitation of Cartesian axes only two
towards achieving ‘‘optimum’’. For the ‘‘poor’’ standard, the
of the four input variables can be selected to map the
degree of membership is zero, while the ‘‘optimum’’ standard
resulting variations in the output (order).
has a degree of membership of one. The rest of the standard is
Fig. 12 illustrates the order resulting from the interac-
presented within the range 0–1 degree of membership. The
tion of delivery time and quality. Again an early delivery
boundaries between the standards are graded which provides
time is better than late, however, the best score is
the fuzzy set sensitivity to the membership function by
corresponding to on target delivery. Moreover, it can be
providing degrees of closeness to the required value of the
seen that quality represents a much higher influence on
perfect order.
deciding a score for the order.
The above example surface was developed based on the
5.6.2. Order output set of 441 fuzzy relational rules corresponding to the setting
Any value for order output between the values of 0 and of the fuzzy membership function of each input/output
0.29 will cause the order to be rejected, between 0.30 and variable. As described earlier the rules can be adjusted in
0.49 indicates that the order is poor and needs major accordance to their suitability or appropriateness. The
attention and action. Although the order value from 0.50 plotted surfaces are intended to give an overall picture of
to 0.69 indicates acceptance of the order it shows a high the order performance rather than discrete values.
288 Y. Amer et al. / Int. J. Production Economics 127 (2010) 278–291

Table 3
Fuzzy evaluation rules for manifest accuracy ‘‘poor’’.

Manifest accuracy ‘‘optimum’’

Quantity Delivery time

Very early Early Acceptable early Optimum Acceptable late Late Very late

Quality ‘‘poor’’
Very low Poor Poor Poor Poor Poor Reject Reject
Low Poor Poor Poor Poor Poor Reject Reject
Acceptable low Poor Poor Poor Poor Poor Reject Reject
Optimum Poor Poor Poor Poor Poor Reject Reject
Acceptable high Poor Poor Poor Poor Poor Reject Reject
High Poor Poor Poor Poor Poor Reject Reject
Very high Reject Reject Reject Reject Reject Reject Reject

Quality ‘‘good’’
Very low Poor Poor Poor Poor Poor Poor Reject
Low Poor Acceptable Acceptable Acceptable Acceptable Poor Poor
Acceptable low Poor Acceptable Acceptable Acceptable Acceptable Poor Reject
Optimum Poor Poor Acceptable Satisfied Acceptable Poor Reject
Acceptable high Poor Acceptable Acceptable Acceptable Acceptable Poor Reject
High Poor Poor Acceptable Satisfied Acceptable Poor Reject
Very high Poor Poor Poor Acceptable Poor Poor Reject

Quality ‘‘excellent’’
Very low Reject Poor Poor Acceptable Poor Poor Reject
Low Poor Poor Acceptable Acceptable Acceptable Poor Poor
Acceptable low Poor Poor Acceptable Acceptable Acceptable Poor Poor
Optimum Acceptable Acceptable Acceptable Optimum Acceptable Acceptable Poor
Acceptable high Acceptable Acceptable Acceptable Acceptable Acceptable Poor Poor
High Poor Poor Poor Acceptable Poor Poor Poor
Very high Poor Poor Poor Poor Poor Poor Poor

Fig. 11. Shows the rule viewer of the fuzzy inference system for an order output=0.818 indicating ‘‘satisfied’’ from Table 1.
Y. Amer et al. / Int. J. Production Economics 127 (2010) 278–291 289

Fig. 12. Output surface of the fuzzy inference system for delivery time and quality.

6. Validate mance measurement. It provides useful managerial


techniques for global SCM and is a practical quantitative
Due to time constraints and staff turnover the validate model for operations managers to redesign/design, moni-
phase was not able to be undertaken in the cases study. tor and control the supply chain order fulfillment process
Simulation may be a good option to consider. A pilot trial is using DFSS and fuzzy logic. There is potential for the
run in as realistic circumstances as possible being subjected model to design, monitor and control other supply chain
to as much ‘‘noise’’ as possible to determine if any processes with Amer et al. (2007b) approaching the
adjustments need to be made. If this stage suggests that the demand management process. Hence there are wider
supply chain design does not or may not meet the required implications for the model to provide further innovative
capability, then it is necessary to retract back through to approaches to supply chain management.
earlier stages of design or optimize (Peplinski, 2004). This paper has highlighted fundamental problems with
current supply chain models which refer to the impor-
tance of a performance measurement systems to achieve
7. Conclusion and discussion supply chain strategy and business outcomes but do not
elaborate on how to determine specific targeted measures
The case application of the order fulfillment model for individual circumstances. The DFSS approach meant
presented in this paper contributes to the qualitative that the case company’s strategy of satisfying their
understanding of supply chain management and perfor- customer was at the forefront of order fulfillment design
mance measurement. By developing a greater under- due to the identification of CCRs through the VOC. By
standing of their own internal processes through a cross considering this at the design onset the focus on designing
functional approach the design team were able to see to meet those requirements meant that they were
where their processes intercept with the processes of their optimized and the appropriate performance measures
supply chain partners’, the trading company and distribu- were matched to the supply chain strategy and the
tion centre. The case application highlighted areas for particular order fulfillment process structure.
linkages across the chain in forecast evaluation and the As 47% of firms are currently only focusing on
‘‘rounding up’’ parameter which called for a mutual developing a seamless flow of internal processes the order
review and closer monitoring to determine if these are fulfillment model presented holds great potential for
the areas impacting on stock outs of top selling items and these firms to develop SCM and extend their focus across
excessive inventory of low selling items. the chain. Partial integration was advocated as an
The order fulfillment model enabled the development alternative to the more holistic approaches of the SCOR
of KPIVs and KPOVs of order fulfillment to be determined and GSCF models in the literature. One of the inhibiting
which added depth and focus to Company A’s perfor- factors to such holistic approaches is the cost, and time to
mance measurement system building supply chain inte- implement such organizational changes which many
gration through focusing on customer requirements, SMEs may not be able to bare. This model provides a
process monitoring and control and the application of good alternative. It could be used to build a new supply
internal and external order fulfillment process perfor- chain as in the case of axiomatic design, to redesign an
290 Y. Amer et al. / Int. J. Production Economics 127 (2010) 278–291

existing one, or just to analyse, understand and improve platforms for the methodology which could make it
the existing chain as was demonstrated in the case more accessible for commercial application.
application and the partial illustration of the model being  Consideration of the use of DOE to optimize process
adapted to the demand management process. parameters and process structures in the optimize phase.
The SCOR model analysed the ‘‘as is’’ state, and then  Using discrete event analysis and simulation in the
through the application of the SCOR template, the ‘‘to be’’ verification phase to verify the final design.
state was benchmarked with other firms. The GSCF
model is a comprehensive model but has limited uptake
amongst firms probably due to its comprehensiveness References
and depth making its implementation time consuming
and expensive. The model presented in this thesis has Amer, Y., Luong, L., Lee, S.H., Ashraf, M.A., Wang, Y.C., 2007a. A systems
approach to order fulfillment using design for six sigma methodol-
the potential for supply chain design innovation that is ogy. International Journal of Business and Systems Research 1 (3),
adaptive to changing circumstances. It can be applied to 302–316.
different channels, addressing issues raised by Bask and Amer, Y., Lee, L., Sang-Heon, L., Wang, W.Y.C., Ashraf, M.A., Qureshi, Z.,
2007b. Implementing design for six sigma to supply chain design. In:
Juga (2001) that supply chains can be obsolete by the time Proceedings of IEEE International Conference on Industrial Engineer-
holistic frameworks are in place and they therefore need ing and Engineering Management, Singapore, 2–4 December 2007,
to deal with multiple goals and strategies within a supply pp. 1517–1521 /http://ieeexplore.ieee.org/xpl/freeabs_all.jsp?arnumber=
4419446S.
chain. The order fulfillment model presented makes a Amer, Y., Lee, L., Sang-Heon, L., Ashraf, M.A., 2008. Optimizing order
considerable contribution to improving supply chain fulfillment using design for six sigma and fuzzy logic. International
integration through an appropriate performance measure- Journal of Management Science and Engineering Management 3 (2),
83–99.
ment system. The approach of DFSS allows for targeted
Aryee, G., Naim, M.M., 2008. Supply chain integration using a maturity
metrics that are relevant to the process performed and the scale. Journal of Manufacturing Technology Management 19 (5),
order fulfillment model can accommodate varying order 559–575.
Bask, A., Juga, J., 2001. Semi-integrated supply chains: Towards the new
fulfillment structures thereby making the metrics relevant
era of supply chain management. International Journal of Logistics 4
to that structure. (2), 137–152.
The application of fuzzy logic to the transfer function Bolstorff, P., 2003a. Putting it all together Chief Logistics Officer 10/2003
also addresses the variability of the four elements of the /www.logisticstoday.comS.
Bolstorff, P., 2003b. Measuring the impact of supply chain performance:
perfect order namely delivery time, quantity, quality and a step-by-step guide to using the supply-chain council’s SCOR model
manifest accuracy. The input of the expert design team /www.logisitcstoday.comS.
makes the measures relevant to their specific circum- Bolstorff, P., 2003c. How to integrate six sigma quality objectives with
the SCOR model /www.logisitcstoday.comS.
stances as the rules are determined from the common Bolstorff, P., 2004. Supply chain by numbers 24th July /www.logisitcsto
sense, human thinking and judgment of the design team day.comS.
members. The fuzzy transfer function enables a clearer Bolstorff, P., Rosenbaum, R., 2007. Supply Chain Excellence A Handbook
for Dramatic Improvement Using the SCOR Model, second ed
look at the relationships between the variables which can Amacom, USA.
provide more robust information for determining the Branch, A.E., 2009. Global Supply Chain Management and International
supply chain improvement focus. Logistics. Routledge, New York, London.
Brue, G., Launsby, R., 2003. Design for Six Sigma. McGraw Hill, USA.
The flexibility of the model lies in that it can
Brusse, W., 2006. All About Six Sigma. McGraw-Hill, USA.
accommodate the many varieties and degrees of supply Buss, P., Ivey, N., 2001. Dow chemical design for six sigma rail delivery
chain integration by providing a means to target some project. In: Proceedings of the 2001 Winter Simulation Conference.
Croxton, K., 2003. The order fulfillment process. The International
aspect of processes relationships or channels that is being
Journal of Logistics Management 14 (1), 19–32.
examined and therefore provides a means for incremental Croxton, K., Garcia-Dastugue, S., Lambert, D., Rogers, D., 2001. The supply
steps for supply chain improvement. chain management processes. The International Journal of Logistics
Management 12 (2), 13–36.
Eckes, G., 2001. The Six Sigma Revolution. John Wiley & Sons, Chichester.
El-Haik, B., Al-Aomar, R., 2006. Simulation-based Lean Six Sigma and
8. Recommendations and further research opportunities
Design for Six Sigma. Wiley, New York.
El-Haik, B., Roy, D.M., 2005. Service Design for Six Sigma. Wiley,
It is recommended that further research opportunities New York.
George, M.L., 2003. Lean Six Sigma. McGraw Hill, USA.
exist by the application of the presented order fulfillment Gunasekaran, A., Patel, C., McGaughey, R.E., 2004. A framework for
model in the following ways to further contribute to the supply chain performance measurement. International Journal of
understanding of its usefulness and validity: Production Economics 87, 333–347.
Klefsjö, B., Wiklund, H., Edgeman, R.L., 2001. Six sigma as a methodology
for total quality management. Measuring Business Excellence 5 (1),
 The model be applied in small to medium sized 31–35.
Lambert, D.M., Cooper, M.C., Pagh, J.D., 1998. Supply chain management:
enterprises that do not have advanced supply chain implementation issues and research opportunities. The International
management practices in place. Journal of Logistics Management 9 (2), 1–19.
 That the model be used in different positions within a Lummus, R.R., Vokurka, R.J., Krumwiede, D., 2008. Supply integration and
organizational success. SAM Advanced Management Journal 73 (1).
supply chain and its use be extended further across a
Martin, J., 2007. Lean Six Sigma for Supply Chain Management. The 10
chain. Step Solution. McGraw-Hill, USA.
 That the model be used in further research to other McKay, D.S., 2003. How a systems engineering education benefits supply
supply chain processes. chain managers logistics spectrum, July/September /http://www.
findarticles.com/p/articles/mi_qa3766/is_200307/ai_n9254070S.
 As the use of the model could necessitate the use of Peplinski, J., 2004. The Importance of Modelling and Simulation in DFSS.
specific software, research could develop suitable Statistical Design Institute, LLC.
Y. Amer et al. / Int. J. Production Economics 127 (2010) 278–291 291

Rath, Strong, 2002. Design for Six Sigma. Rath and Strong Publishers, Yang, H.M., Choi, B.S., Park, H.J., Suh, M.S., Chae (Kevin), B., 2007. Supply chain
Massachusetts. management six sigma: a management innovation at the Samsung
Simatupang, T., Sridharan, R., 2005. An integrative framework for supply group. Supply Chain Management: An International Journal 11 (2),
chain collaboration. The International Journal of Logistics Manage- 88–95.
ment 16 (2), 257–274. Yang, K., 2008. Voice of the Customer: Capture and Analysis. McGraw-
Summers, D.C.S., 2007. Six Sigma Basic Tools and Techniques. Pearson Hill, New York.
Prentice Hall, USA. Yang, Kai, El Haik, Basam., 2003. Design for Six Sigma. McGraw-Hill.
Supply-Chain Council, 2009 /http://www.supply-chain.org/S. Zhang, Y., Xu, Y., Wan, Z., 2007. Research on the application of six sigma
Taguchi, G., Chowdhury, S., Yuin, W., 2005. Tagushi’s Quality Engineering method in logistic corporation. In: Proceedings of the IEEE Interna-
Handbook. Wiley, New Jersey. tional Conference on Automation and Logistics, Jinan, China, August
Xu, H., Koh, L., Parker, D., 2009. Business process inter-operation for 18–21, 2007.
supply network co-ordination. International Journal of Production
Economics, in press, doi:10.1016/j.ijpe.2009.05.015.

You might also like