Value stream classification
Martin Christopher
Cranfield School of Management, Cranfield University, Cranfield, UK
Denis R. Towill
Cardiff Business School, Cardiff University, Cardiff, UK
James Aitken
Cranfield University, Cranfield, UK, and
Paul Childerhouse
Waikato Management School, The University of Waikato,
Hamilton, New Zealand
Abstract
Purpose – In the twenty-first century business scenario, most organisations supply a range of
products to multiple markets, so participate in several often quite different supply chains. Just as the
linear chain is a simplification of a supply network, the single channel is a simplification of the true
complexity many organisations face. When all products are pushed down a single channel, they
are paced by the slowest and customers are charged an average price resulting in many being
underserved. Manifestly, for most businesses “one size fits all” is not a viable option in delivery
pipeline design and operation. This paper aims to address these issues.
Design/methodology/approach – Traditionally, the requisite number of delivery pipelines
operated by a business is determined by “hunch”, as is the range of products flowing down each
channel to the marketplace. The information technology (IT) revolution, which in turn has spawned
the “analytic corporation” enables pipeline selection and product matching to be placed on a more
formal footing. In order to enable the tailoring of value stream pipelines to markets five classification
variables are proposed. These are duration of product life cycle (D), delivery window (W), annual
volume (or value) (V1), product variety (V2) and demand variability (V3).
Findings – Through the use of case studies drawn from real-world situations, the authors are able to
highlight the practical value of using appropriate taxonomies to identify appropriate supply chain design
strategies. A framework for the implementation of a scheme for value stream classification is proposed.
Practical implications – The proposed analysis, design, and implementation methodology is
summarised in flow diagram format. Emphasis is placed on the formation of a “natural group” task force
to execute this programme. Production, sales, logistics, and marketing are all essential “players” therein.
Originality/value – It is established that the DWV 3 classification system is suitable for framing
pipeline performance improvement programmes. By exploiting the “analytic corporation” IT
capability, the system has a dynamic role in determination of channel switching as products progress
through their individual life cycle phases.
Keywords Supply chain management, Classification schemes, Delivery, Lean production,
Agile production
Paper type Research paper
1. Introduction
Shewchuck (1998) summarised the drawbacks resulting from rigid adoption of the “one
size fits all” philosophy in attempting to satisfy customer demands in the present day
marketplace. The substance of his review shown in Table I is a useful starting point in
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1741-038X.htm
JMTM
20,4
460
Received November 2007
Revised August 2008
Accepted October 2008
Journal of Manufacturing Technology
Management
Vol. 20 No. 4, 2009
pp. 460-474
q Emerald Group Publishing Limited
1741-038X
DOI 10.1108/17410380910953720
understanding the complexities facing suppliers in an era of short life cycles, high
variety, and high-initial demand (especially of electronic “fashion” accessories).
Shewchuck (1998) thereby identifies four distinct manufacturing environments. Also
highlighted are the potential tensions between marketing and manufacturing. As we
have indicated previously, it is essential to properly exploit both the “lean” and “agile”
delivery paradigms to arrive at a satisfactory total solution (Christopher and Towill,
2000). This includes exploitation of both information and material flow de-coupling
points (Mason-Jones and Towill, 1997), and the leagility concept (Naylor et al., 1999).
We suggest these are important facets in the move to implement market-driven supply
chains, such as outlined by Sharifi et al. (2006).
The delivery pipeline design procedure adopted should be transparent to analysts,
designers, managers and users alike. Using too many classification variables (taxas)
means the analysis may become very messy and thereby lose valuable insight.
However, using too few taxa may not generate enough feasible solutions to eventually
arrive at a “best” design. The taxa need gradation on an agreed simple scale
(preferably binary). Finally, the pipelines output by the taxonomy should be capable of
relating to established delivery processes so that their actual engineering exploits “best
practice”. Finally, the total pipeline system must cater for real-time switching of
products between value streams according to pre-set triggering rules. In other words,
as noted by Fassoula (2006), adaptive operation of supply chains is highly desirable.
Posited manufacturing
environment Typical products
Associated marketing
characteristics
Associated
manufacturing
characteristics
Compressed life cycle Home computers
Audio systems
Cameras
Medium product variety
High demand during
maturity phase
Short product life cycle
High volumes require
mass-production for
short periods
Danger of product
obsolescence
Compressed
time-to-market
Mobile phones
Palmtops
Fashion goods
Medium product variety
High demand during
maturity phase
Short product life cycle
and very short growth
phase
Competition so fierce
that mass-production
needed during start-up
Danger of failure to
penetrate market
Mass customisation via
assembly
Personal computers
Made-to-order bicycles
Three-day car
High-product variety
Single unit product
demand
Relatively short lead
time
Deep bill of material
(BOM)
Product variety enabled
via limitless
combination of
components
Mass customisation via
processing
Spectacle lens
Made-to-order shoes
Made-to-order clothes
High-product variety
Single unit product
demand
Very short lead time
Shallow BOM
Product variety enabled
via on-the-spot
processing
Source: Authors summarising Shewchuck (1998)
Table I.
Why “one size does not fit
all” in twenty-first
century delivery pipelines
Value stream
classification
461
2. Contribution of present paper
Christopher and Towill (2000) suggested on an a priori basis that there were five
underlying considerations when enterprise supply chain strategies were being
investigated. These were the likely duration of the lifecycle (D); the time window for
delivery (W); the anticipated volume of demand (V1); the potential variability of that
demand (V2) and the level of variety (V3). Whilst this framework has a strong logical
underpinning, it can be argued that it leads to the potential for a large number of
theoretical supply chain strategies. It also assumes that all five dimensions would be of
equal importance in all contexts. Hence, this approach should be seen as a preferred
starting point which in practice may result in amalgamation and consolidation of
product ranges and further system simplification as justified by that particular
enterprise. However, it is far better to start the analysis with too many variables rather
than too few, since the risk of wrongly selecting a particular “best” set of delivery
pipelines is thereby reduced.
This paper describes the reasoning behind the original DWV 3 framework and links
the resulting taxonomy to a number of prescriptions for organisational change and
supply chain development. It is suggested that “organisational agility” (Nagel and Dove,
1991) is a key pre-requisite for the adoption of a more flexible, differentiated supply
chain strategy. There are also significant issues relating to supplier choice criteria,
distribution channel design and governance. These concepts are evaluated and
discussed in the context of particular examples covering a range of industries and
enterprises. The context of our research can be seen from Table II. From the range of
variables considered by the various authors cited therein, it is evident that value stream
classification is an extensive field of interest. It obviously relates to many other concepts,
including customisation, postponement, plus the lean and agile paradigms
(Childerhouse et al., 2002). As observed by Turner and Williams (2005), despite such a
knowledge base being available current supply chains may still fall way below customer
expectations.
The uptake of each of the foregoing five variables is documented for six real-world
case studies. In no single case are all these taxa utilised. However, each variable is used
in at least one of the cases. Furthermore, the additional variables used in the cases
beyond the starting point of DWV 3 are also discussed. The empirical cases cover a
range of organisations both in size and industrial sector. Hence, they provide a
reasonably broad investigation of our value stream classification approach. We
demonstrate that DWV 3 is a robust framework for developing the necessary pipelines
for enhancing enterprise competitiveness. In practice, this system results in a limited
(and therefore handleable) number of aggregate pipelines tailored to individual
business needs. Thus, a company with many thousands of SKU’s will be best served
with a few (usually between 4 and 6) carefully selected, designed, and implemented
pipelines matched to requisite customer service levels and other key business metrics.
The paper concludes with a description of the underpinning principles that should
enable the effective design and implementation of responsive supply chains.
3. Searching for the key classification variables
As stated earlier in the paper, Christopher and Towill (2000) argued that there are five
key characteristics that influence decision making on the design of value stream
delivery strategies. “These attributes are duration of life cycle; time window for
JMTM
20,4
462
delivery; volume; variety; and variability (DWV3).” Table III summarises the key
reasons why each of these five classification variables was included as discrete taxa
within our scheme. Note that cross-reference is made to important citations in the
literature as appropriate. Duration and stage of product life cycle have been noted by
many authors as key characteristics influencing a demand chain to adopt tailored
strategies. For example, Fisher (1997) uses the duration of product life cycle as a
classification variable to differentiate between functional products with long life cycles
(then set at , two years) and innovative products with short life cycles (three to 12
months). He particularly emphasises the need for specific responsive delivery
strategies for the latter to best exploit the short windows of opportunity.
Authors Classification variables Taxa
Hoekstra and Romme (1992) Position of de-coupling point Make and ship to stock
Process constraints Make-to-stock
Product-market constraints Assemble to order
Delivery service requirements Make-to-order
Inventory cost considerations Purchase and make-to-order
Fuller et al. (1993) Annual sales revenue Systems orders
Annual unit volume Customer inventory replenishment
Co-ordination requirements Rapid response
Destination volume Nuts and bolts
Handling characteristics Slow movers
Customer order fulfilment interval Bulk cable
Lampel and Mintzberg (1996) Level of aggregation Pure standardisation
Level of individualisation Segmented standardisation
Customised standardisation
Tailored customisation
Pure customisation
Fisher (1997) Product innovation Functional products
Demand volume stability Innovative products
Product life cycle duration
Made-to-order lead time
Product variety
End-of-sale mark down
Pagh and Cooper (1998) Product life cycle Full speculation
Product customisation Logistics postponement
Product variety Manufacturing postponement
Product value Full postponement
Relative delivery time
Delivery frequency
Uncertainty of demand
Naylor et al. (1999) Cost and quality Lean
Lead time and service level Agile
Stability of demand Leagile
Lamming et al. (2000) Product innovation Innovative-unique and complex
Product uniqueness Innovative-unique and noncomplex
Product complexity Functional and noncomplex
Source: Authors
Table II.
Classification of value
streams within the
literature
Value stream
classification
463
The time window for delivery (or delivery lead time) reflects the speed requirements
placed on the demand chain. In combination with manufacturing and logistics lead
times, it identifies the most feasible position of the de-coupling point (Hoekstra and
Romme, 1992). Fuller et al. (1993) use delivery lead time as a classification variable,
whilst Christopher (1992) emphasises the competitive advantage of offering increased
customer service via shorter order cycle times. Volume is the third of the DWV 3
variables. Fuller et al. (1993) use this as a dominant variable in their classification of
logistically distinct businesses which resulted in the pipeline set summarised in
Table IV. However, they are not generic, which is the goal of this paper. In similar vain,
Parnaby (1994) explains the “Lucas Way” by which the manifold manufacturing
processes are segmented depending on volume. Three such categories are often used,
called “runners”, “repeaters” and “strangers”. They are typically based on a Pareto
analysis of present/forecast demand.
Product variety, the fourth DWV 3 variable is constantly increasing in the
marketplace as demand chains attempt to compete on the basis of added value in
relation to such attributes as colour, form and function (Gottfredson and Aspinal, 2005;
Scala et al., 2006). Variety has a major effect on a demand chain in relation to the point
of differentiation as increased changeovers, larger inventory pools and scheduling
issues often result. The fifth and final variable is demand variability. Fisher (1997)
emphasises the impact that unpredictable demand can have on the chain either in the
Classification variables Key reasons for use within value stream classification system
Duration of life cycle (D) Short life cycles require rapid time to market and short end-to-end
replenishment pipelines
The value stream is required to “fast track” product development,
manufacturing and logistics
Replenishment lead times vary according to stage within the
product’s life cycle
Time window for delivery (W) Rapid response is required to replenish fashion goods that are
selling well at a particular point in time
Competitive pressures are continually reducing acceptable
response times
Many value streams compete on the basis of very short windows
for delivery of customised products
Volume (Vol) High-volume mass markets allow for lean-type production and
make-to-forecast strategies
Lower volume markets benefit from flexibility throughout the
entire demand chain
Variety (Var) Greater variety results in a larger number of SKU’s
Continuous appraisal of the split between variants required during
the product’s life cycle
Variants popular at the introductory stage may be less popular in
the decline stage
Variability (Vly) Variability relates to both demand and supply predictability
Spikes drastically affect capacity utilisation and resultant
production techniques
Unpredictability increases the risk of obsolescence and lost sales
Source: Authors
Table III.
Key reasons for the
choice of DWV 3
variables to classify
demand chain types
JMTM
20,4
464
form of stock-outs and resultant lost sales, or alternatively excessive obsolescence
costs. Subsequently, Naylor et al. (1999) use stability of demand to appropriately match
either lean or agile paradigms to best satisfy marketplace demands. This is an
important concept for location of the material flow de-coupling point so as to generate
appropriate value stream delivery pipelines. Whilst we have tended to focus on
demand variability it should also be recognised that supply or process variability can
be equally important in the choice of supply chain strategy. Indeed, in these days of
increased global sourcing and off-shore manufacturing, this source of variability needs
to be factored into the equation. We would propose therefore that our original DWV 3
model be amended to define “variability” as relating to the uncertainty in the wider
supply chain, not just in demand.
4. Applying the value stream classification system
Selecting a value stream classification is, of course, a key element, but still just one
element in any business process improvement (BPI) programme to match pipelines to
market-place for that particular enterprise. Considerable data collection and processing
is needed to profile the product range prior to detailed design. The complete BPI
procedure is shown in Figure 1. It covers agreement on the classification variables to be
used and codification levels (H/L; H/M/L) as appropriate. If too many levels are
selected, then subsequent analysis becomes more complicated. Indeed, the benefits
from using a formal classification system may be lost and the exercise de-generate into
chaos as vested functional interests within the enterprise argue strongly in favour of
“their” special cases. Such judgements need to be made at the design stage when the
full facts have emerged.
Undertaking the codified profiling of an extensive product range is obviously a
computer-based task. Nowadays, this would rightly be regarded as an integral activity
of the “analytic corporation” as described by Davenport (2006). Such profiling is
followed by cluster analysis to determine how much traffic might use a particular route.
For example, the IMATEL (Fuller et al., 1993) industrial business process-reengineering
programme generated 384 theoretical routes. However, with their then current business
only 300 of these were actually populated, but many of them likely to be quite sparsely.
Pipeline description Associated product attributes
System orders Co-ordinating and marshalling orders for installers
Dedicated support services
Customer inventory replenishment Shipping regular “Top-up” orders
Stocking points located near largest customer
Rapid response Order-specific shipping
Low-cycle time order entry and handling
Nuts and bolts Ordered and stored in bulk
Typically available in “barrels”
Slow movers Dedicated product with unique pricing
Stocks held by customers
Bulk products Independent transportation
Dedicated handling facilities
Source: Fuller et al. (1993)
Table IV.
Delivery pipeline
implemented at
IMATEL Inc.
Value stream
classification
465
Figure 1.
Flow diagram
summarising BPI
programme for
implementation of value
stream classification
schema
R
es
ou
rc
in
g
Ph
as
e
A
na
ly
sis
a
nd
Id
ea
s T
es
tin
g
Ph
as
e
D
es
ig
n
Ph
as
e
Ex
ec
u
tio
n
Ph
as
e
Logistics Production
Sales
Set Up
“Natural Group”
Task Force
Marketing
Agree Target
Service Levels
Select
Classification
System
Product
Profiling
Cluster
Analysis
Select
Contender
Pipelines
Modifications
Aggregation
and Check
Pilot Run
Full-Scale
Implementation
Source: Authors
JMTM
20,4
466
This is the BPI stage where computer-based experiments with adjusting codification are
appropriate. Thus, the initial assumption on the volume breakpoint may have been
(greater than/less than) 500 units per week. Given the right software to repeat the
profiling for this variable changed to 1,000 (or any other sensible value) units per week is
straightforward. However, realistic initial selection is preferred as the need for excessive
experimentation can cloud the issue, rather than provide the required transparency.
Further aggregation to eventually produce a realistic number of value stream
pipelines requires informed judgement. As we shall see in the next section, IMATEL
reduced the 300 “paper” routes into just six very clearly definable pipelines.
Furthermore, it is