Complexity and the supply chain
- The sources of supply chain complexity
- The cost of complexity
- Product design and supply chain complexity
- Mastering complexity
We have several times in previous chapters suggested that rather than refer to supply chains we should talk instead about networks. The idea of a chain suggests a series of linear one-to-one relationships whereas the reality is that the focal firm lies at the centre of a complex web of interconnected and interrelated yet independent entities.
Partly as a result of outsourcing activities that previously were performed in-house combined with the trend to offshore manufacturing, many companies have found that they have added to the complexity of their operations because the degree of interdependency across the network has increased. Thus an event or action taking place in one part of the network will often have unforeseen impacts somewhere else in the network. The unpredictability of these events is heightened by the growing volatility that characterises today’s business environment.
The well-known ‘butterfly’ effect seems to typify much of today’s supply chain turbulence. The idea is that a butterfly, flapping its wings somewhere over the Amazon basin, can cause a hurricane thousands of miles away! Whilst this example of what is sometimes described as ‘chaotic’ effects may be a little far-fetched, it provides a useful reminder of how the ‘law of unintended consequences’ applies to today’s highly interconnected supply chains.
In April 2010 a dormant volcano in Iceland erupted, sending a plume of ash into the upper atmosphere. A cloud of ash and debris from the eruption began to drift across the skies of Northern Europe. Because of a concern for aircraft safety most airports in the region were closed for the best part of a week. Whilst there was a considerable impact on individuals’ travel plans – many thousands of people were stranded away from home – there was also a less visible, but significant, impact on a number of supply chains. Many time critical components are sent by air freight or air express, and as major hubs in the UK and continental Europe were forced to close, the ‘butterfly effect’ was felt around the world.
In its strictest sense, complexity does not mean complicated (although complex systems often are complicated), but rather it describes a condition of interconnectedness and interdependencies across a network. A good example of a complex system is the weather. Many different influences combine to create a specific weather condition, each of those influences are themselves the result of interactions and hence a small change in one element can fundamentally affect the final outcome. Hence the difficulties that weather forecasters face trying to predict even tomorrow’s weather.
The outcome of complexity in a supply chain, as with the weather therefore, is uncertainty, and with that uncertainty comes an increased likelihood that forecast error will increase in line with complexity. This growing uncertainty brings with it a serious challenge to the classic practice of running the business on the basis of forecasts. It will be apparent that in conditions of stability – and hence lower uncertainty – forecast accuracy should generally be high. Equally, the converse will be true, i.e. as uncertainty increases so too will forecast accuracy reduce. Hence the argument that if uncertainty is to be the norm – at least for the foreseeable future – then a new approach will be required. Indeed, as we have previously observed, the challenge that organisations now face is how to reduce their dependence on forecasts and to become increasingly demand and event driven.
The sources of supply chain complexity
Complexity in a supply chain can arise from a number of sources and some of the most common causes are detailed below.
1 Network complexity
The more nodes and links that exist in a network then clearly the more complex it becomes. As a result of outsourcing non-core activities, many companies are today much more reliant on external suppliers of goods and services. Those external suppliers also are dependent upon a web of second-tier suppliers, and so on. There is a strong likelihood that the focal firm at the centre of the network will not even be aware of many of the second- or third-tier suppliers that feed their upstream supply chain. The potential for unexpected disruptions to the supply chain is clearly heightened by these extended networks as evidenced by the following example.
Following the shut-down of Dell’s American assembly line within days of the September 1999 earthquake in Taiwan the company set out to understand why this had happened.
To do this Dell studied where their tier-one suppliers did their shopping and this in turn soon yielded the first important answer – the Taiwan Semiconductor Manufacturing Corporation (TSMC). Dell’s executives realised that they were in fact buying hundreds of millions of dollars of chips each year from TSMC indirectly.
Abridged from Lynn, B.C., The End of the Line, Doubleday, 2005
One factor driving network complexity is the fact that whilst the entities that make up the supply chain are usually independent organisations, making decisions according to their own rules, the interdependencies are high. Consequently the way we network as a whole might be impacted by the actions of one element within that network is difficult if not impossible to predict. The way the system may have responded to a disturbance in the past is not necessarily a guide to how it might respond to a similar disturbance in the future. Thus conventional forecasting tools will be less effective the greater the complexity of the network.
2 Process complexity
Underpinning every supply chain are innumerable processes – processes internal to the firm as well as those processes managed by upstream and downstream partners. Often these processes have been developed in a haphazard way and have been added to and modified to reflect current requirements and as a result have become more complex. This complexity is manifested in processes with multiple steps, often performed in series rather than in parallel.
Lengthy processes containing many different activities will not only create extended lead-times but are also more prone to variability in performance. The more steps in a process and the more ‘hand-offs’ that exist, the greater the likelihood that there will be frequent discrepancies between planned and actual outcomes.
There is a need for a constant review of process structure and a consequent re-engineering if this pervasive source of supply chain complexity is to be kept to a minimum.
When end-to-end supply chains are examined in detail it usually transpires that the majority of time is non-value-adding time. More often than not this non-value-adding time is idle time – in other words time spent as inventory. This non-value-adding time is itself generated by the processes that underpin the supply chain.
We previously introduced the idea of BPR in Chapter 6. Experience has shown that many business processes are ‘legacy’ processes. This means that they were designed for a particular purpose at a particular time and may no longer be fit for purpose in today’s changed conditions. Organisations intent on reducing process complexity must constantly question the purpose of every process and continuously explore the opportunities either to remove them altogether or to reduce the number of steps involved.
3 Range complexity
Most business organisations find that the range of products and/or services that they offer to the market has a tendency to grow rather than reduce. The rate of introduction of new products or services, new pack sizes or variants and brand extensions seems to outpace the rate at which existing products or services are eliminated. The general effect of this mushrooming of the product/service portfolio is to extend the ‘long-tail’ of the Pareto distribution.
Typically, as more variants are added to a range, the demand per variant will reduce with a subsequent impact on forecast accuracy. Consider the difference between the Ford Motor Company at the time of Henry Ford I producing a single model – the Model T, with the reputed offer of ‘any colour you like as long as it’s black’ – with the company today.
Ford, today, offers a vast range of models with extensive options. In theory there are possibly millions of different variants! This multiplication of the product range means that, inevitably, average demand per variant is very low. Hence the difficulty of forecasting at the individual variant level and thus the typically large inventories that build up as a result of forecast error.
Range complexity also adds to cost because of its impact on inventory levels. Every time a new variant is introduced (e.g. a new pack size, a new colour or flavour) it becomes a new SKU with a consequent need for its own safety stock. Whilst customers clearly value choice, it is important to understand what is the level of variety sufficient to provide a degree of choice that will lead to an outcome where marginal revenue is greater than marginal cost. One company that has sought to implement a structured approach to managing range complexity is Hewlett Packard as described in the box below.
Managing Range Complexity at Hewlett Packard
Hewlett Packard (HP) is a leading manufacturer of PCs, printers, servers and other related products. It offers a vast range of products with many different variants available. For example for its laser printer range alone there are over 2,000 SKUs!
Whilst offering this level of customer choice enables high levels of sales to be achieved, it can come at a cost – particularly inventory related costs but also too the managerial resources required to manage the resulting complexity.
To help meet this challenge, HP has developed two decision support systems designed to enable it to better understand the cost/benefit of new product introductions and to improve availability with less overall investment in inventory. As a result HP has streamlined its product offering and dramatically reduced cost whilst at the same time achieving quicker response times and higher levels of customer satisfaction.
Source: Ward, J., et al., ‘HP Transforms Product Portfolio Management with Operations Research’, Interfaces, Vol. 40, No. 1, pp. 17–32, 2010.
4 Product complexity
The design of products can have a significant impact on supply chain complexity. It can be argued that the supply chain begins on the drawing board in that decisions on the choice of materials and components can directly or indirectly impact total life cycle costs as well as agility and responsiveness.
Product complexity can arise because the number of components or sub-assemblies is high, or because there is little commonality across the bills of material for different products. The less the commonality at the bill of material level the less is the flexibility to vary product mix or volume.
A further unforeseen impact of product design decisions is that if components or materials are specified which happen to have lengthy replenishment lead-times then the ability to respond rapidly to changes in demand for the product will be impeded.
By involving logistics and supply chain planners early in the design process much of the subsequent complexity can be avoided. For example, at Motorola all new product ideas are screened for complexity1 before they can be considered for commercialisation.
This approach was developed initially for their mobile phone manufacturing operations – a business they have now disposed of. Prior to the implementation of this screening process there was often little commonality of parts across their mobile phone range. For a single mobile phone there could be over 100 possible configurations, i.e. four different colours and 30 software choices. Furthermore, these product variations were made ahead of demand to a forecast that was only accurate 3 per cent of the time! To tackle this problem Motorola devised a ‘complexity index’ for each product, which included the number of components, the degree of commonality, lead-time of supply, and so on. Ideas for new products with high scores on the complexity index tend not to be proceeded with.
5 Customer complexity
Customer complexity arises as a result of too many non-standard service options or customised solutions. The costs of serving different customers can vary significantly. Each customer will exhibit different characteristics in terms of their ordering patterns, e.g. frequency of orders, size of orders, delivery requirements and so on. These differences will be increased further as a result of the availability of different service options or packages and/or customisation possibilities.
Gottfredson and Aspinall give an example of how too extensive a service offer can add complexity to the sales process2:
One telecommunications company, for example, has used the power of IT to slice and dice its service set into ever-more finely differentiated options. The firm hoped it would boost revenues by more precisely fulfilling the needs of every imaginable buyer. But offering so many options has had the opposite effect. The company’s customer service reps are now forced to sort through more than a thousand promotion codes whilst they’re talking to a potential customer. Most of the promotions offer distinct levels of discounts and product benefits. Making sense of them all is an overwhelming task.
Even though from a sales and marketing perspective there may be advantages to be gained from offering a range of options to customers, these decisions must be tempered by a detailed knowledge of their cost and agility implications. Ultimately the only complexity that can be justified is that complexity which delivers real value for which customers are prepared to pay.
A problem that is faced by many businesses is that they have a limited understanding of the true costs of servicing individual customers. It is quite possible that because some customers generate a high cost-to-serve and order products with relatively low margins they could actually lose money for the company. Using tools such as ABC can help identify those customers whose cost-to-serve is high relative to the revenue that they generate. Using this information, alternative service options might be devised that could improve the profitability of those customers.
6 Supplier complexity
The size of the supplier base can add to supply chain complexity by increasing the number of relationships that must be managed as well as increasing total transaction costs. Because one of the prerequisites for agility is a high level of collaborative working with key suppliers, this implies a high level of active supplier management and supplier involvement in process integration. It is unlikely that this degree of closeness can be achieved across a diverse supplier base and hence the need for rationalisation. The implications of such a supply base rationalisation are profound. It is clear that careful regard must be paid to the effect of a smaller number of suppliers on the resulting supply chain risk profile. Too high a level of dependence on just a few critical suppliers can be dangerous. Instead a better option, if available, is to have a lead supplier across a category of products who takes responsibility for the management of that category across a number of suppliers, for example in the same way a logistics service company such as UPS might co-ordinate a number of logistics and transport providers for a client company.
Alternatively, the risk inherent in single sourcing might also be mitigated by adopting a strategy to single source by site or by model. Thus a manufacturing company with multiple factories might single source a particular material from supplier ‘A’ for one factory but use supplier ‘B’ for another factory. Likewise a company making a number of different models in a product family, e.g. a white goods manufacturer, might single source the procurement of components by model. The advantage of such a strategy is that it further reduces complexity by enabling the adoption, if appropriate, of solutions such as VMI. Usually VMI is not practical when there are multiple suppliers of the same item.
With a smaller supplier base, a company can more pro-actively manage supplier relationships through ‘supplier development’ programmes. Typically such programmes involve the company working closely with individual suppliers to identify opportunities to improve not just product quality, but also process quality and to work jointly on cost-reduction initiatives.
Supplier complexity can be further reduced if it is possible to standardise the processes that are used to connect and communicate with suppliers. Thus for example a common ‘procure-to-pay’ process where all transactions with suppliers are handled in the same way.
7 Organisational complexity
Most businesses have traditionally organised around functions and departments, and their organisation charts have many levels and tend to be hierarchical in their structure. Such ‘vertical’ organisational arrangements are no doubt administratively convenient in that there can be a ‘division of labour’ between functions as well as effective budgetary control. However, they tend to inhibit agility because they are, of necessity, inwardly looking with a focus on efficiency rather than customer facing with a focus on effectiveness. A further problem is that over time the functions have a tendency to become ‘silos’ with their own agendas, and they can lose sight of the fundamental purpose of the business, i.e. to win and keep profitable customers.
The challenge is to find a way to break through these silos and to re-shape the organisation around the key value-creating and value delivery processes. Such process-oriented businesses are ‘horizontal’ rather than ‘vertical’ in their orientation. They are cross-functional and hence there is a stronger emphasis on teams and on process improvement in terms of speed and reliability.
As organisations grow, either organically or through merger and acquisition, the likelihood is that they will become more cumbersome and less able to respond rapidly to change. Consequently there is a constant need to re-engineer existing processes and to root out the complexity that will inevitably arise if things are left to themselves. Organisational complexity can also be exacerbated by having to work across time zones and cultures as a result of the globalisation of business. Frequently this added complexity is an unintended consequence of low-cost country sourcing and/or cross-border mergers.
8 Information complexity
Today’s supply chains are underpinned by the exchange of information between all the entities and levels that comprise the complete end-to-end network. The volume of data that flows in all directions is immense and not always accurate and can be prone to misinterpretation. Visibility of actual demand and supply conditions can be obscured through the way that information is filtered and modified as it passes from one entity or level to another. The so-called ‘Bullwhip’ effect is a manifestation of the way that demand signals can be considerably distorted as a result of multiple steps in the chain. As a result of this distortion, the data that is used as input to planning and forecasting activities can be flawed and hence forecast accuracy is reduced and more costs are incurred.
In a sense, information complexity in a supply chain is directly or indirectly influenced by the preceding seven sources of complexity. Network and process complexity will impact the number of stages, steps and levels through which the information must pass; range and product complexity add variety and lead to multiple bills of material and hence more data; customer and supplier complexity means that the exchange of data increases significantly and organisational complexity implies more levels through which information must pass as well as more hand-offs from one function to another.
The antidote to information complexity is firstly a reduction in the other seven sources of complexity as well as greater visibility. A key to that visibility has to be a greater level of collaborative working across the supply chain, where information transparency is seen as a vital prerequisite for a more efficient and effective value delivery system.
In recent years, many companies have begun to use advanced data analytic tools to enable them to ‘mine’ the massive amounts of data that is now available. This data could relate to transactions, demographics, customer profile information and a whole host of other relevant data sources. What has now come to be known as ‘big data’ represents a great opportunity for companies to develop real customer insight. Whereas in the past it has difficult to see through the ‘fog’ surrounding this sea of data, now, using these new analytic tools, it is possible to master this particular source of complexity.
The cost of complexity
It can be argued that an increasing proportion of total end-to-end costs in the supply chain are driven by complexity in one form or another. Often these costs may not be readily transparent as they are hidden in general overheads or the costs of carrying inventory which, as we observed in Chapter 4, are not always properly accounted for.
Underlying much of the cost of complexity in the supply chain is the Pareto Law (the so-called 80:20 rule). Vilfredo Pareto (1848–1923) was an Italian industrialist, sociologist, economist and philosopher. In 1909 he identified that 80% of the total wealth of Italy was held by just 20% of the population. Thus was born the 80:20 rule that has been found to hold across many aspects of social and economic life. In Chapter 2 it was suggested that an 80:20 relationship exists with regard to customers and products, i.e. typically 80% of the profit derives from 20% of the customer and likewise 80% of the profit comes from just 20% of the products. Generally this 80:20 relationship applies across most elements of the supply chain and is a key contributor to complexity and hence cost.
Most businesses will find if they perform an 80:20 analysis that they have a ‘long-tail’ of customers who, whilst significant in numbers, actually contribute very little to overall profitability – indeed some may actually make a loss. Likewise, the same conclusions would probably emerge from an 80:20 analysis of products.
Sometimes when performing the 80:20 analysis across the product range, it is tempting to suggest that where a ‘long-tail’ exists it should be removed through product rationalisation. However, there may be strategic reasons for maintaining a high level of variety, or indeed there may be opportunities to use alternative strategies to manage the slow movers to make them profitable. For example, it has been suggested that if an Internet distribution channel is available, then the ‘long-tail’ can become a source of profitable business.3 Because the ‘long-tail’ represents such a large number of products, even though individual item sales levels are low, if inventory and distribution costs can be reduced by creating a single, virtual inventory through working with partners across multiple channels, the economics may be transformed. To a certain extent this is the approach that Amazon has taken, enabling it to offer a vast range of book titles (and other products) but with minimal inventory.
However, for most companies it is likely that a selective rationalisation of slow-moving lines will have a positive impact on overall profitability.
Product design and supply chain complexity
It is important to recognise that often a significant source of supply chain complexity is the actual design of the product itself. It has long been known that a large part of total through-life costs are determined at the drawing board stage – sometimes as much as 80%.4 There are a number of ways in which product design decisions can impact subsequent supply chain complexity and hence costs. For example:-
- Time-to-market and time-to-volume
Decisions on the functionality of products can increase manufacturing complexity and reduce flexibility and responsiveness. For example, the more features a product has will likely impact the bill of materials and response times. - Added complexity through lack of commonality
Decisions on product design impact the bill of materials. Low levels of component commonality across product families will add complexity. - Increased replenishment lead-times
Some design decisions will determine the choice of supplier and therefore could impact replenishment lead-time, e.g. where the supply source is offshore. - Supply chain vulnerability
Again, if the design decision involves unreliable supply sources this could potentially increase the chance of supply chain disruption. - After-sales support
For those products requiring after-sales support, e.g. service parts, the design of the product will have implications for inventory levels. - Late stage customisation
The ability to postpone the final configuration or the packaging of a product will be enhanced or constrained by product design decisions.
Many companies are now seeking to address these problems by adopting a ‘design for the supply chain’ approach. This entails a high level of integration between the design team and those responsible for logistics and supply chain management. The idea is to ensure that the ‘through-life’ impact of product design decisions is fully understood before the business commits to launch the product. Thus, for example, Zara the leading global ‘fast fashion’ company has always brought its designers and its supply chain planners together to make sure that every new garment design can be brought to market cost-effectively within the shortest possible time.
Mastering complexity
Because supply chain complexity is such a major source of total end-to-end pipeline cost as well as being a significant inhibitor of responsiveness, it is essential that complexity reduction becomes a priority. It can be argued that today’s supply chain managers need to be ‘complexity masters’, such is the importance of containing and removing this impediment to enhanced profitability.
Figure 9.1 suggests a five-stage process for bringing under control.
The first step in managing supply chain complexity is to understand where it is coming from. A good starting point to identifying the source of complexity is to review the eight categories previously identified, i.e. network, process, range, product, customer, supplier, organisational and information complexity.
Network and process complexity can be identified through the use of mapping procedures such as those described in Chapter 8. Because networks and processes are not often managed holistically, i.e. they tend to be managed by individual activity rather than as a whole, the likelihood is that they will contain the potential for unnecessary complexity, e.g. too many echelons, poorly managed interfaces and too many activities that do not add value. Network simplification and process re-engineering should be ongoing in every supply chain that seeks to become less complex.
Range, customer and supplier complexity can be identified through Pareto analysis. In other words what proportion of total revenue, spend or inventory is accounted for by what proportion of customers, suppliers or SKUs? By focusing on the ‘long-tail’ previously discussed, it should be possible to identify opportunities for rationalisation. Again, it should be stressed that such rationalisation needs to be addressed cautiously with regard to the wider business strategy and financial consequences.
Product complexity will be revealed through a detailed analysis of the bills of materials of each product in the range. The goal is to both minimise the number of components in each product and to maximise the commonality of components, sub-assemblies or platforms across the range.
Organisational complexity is partly driven by the number of levels in the business and by the decision-making structure. Typically organisations with many levels and with many functional ‘silos’ tend to be slow to respond to changed conditions and slow in new product development and introduction. One effective way to reduce this source of complexity is by a greater emphasis on working across functions, particularly by creating process teams – an idea to which we shall return in Chapter 14.
It should however, be recognised that not all complexity is bad. In some respects it is through complexity that organisations differentiate themselves from their competitors. For example, customers often seek product variety, they are not prepared to settle for the previously quoted Henry Ford I offer of ‘any colour you like as long as it’s black’!
The challenge for supply chain managers is to understand the value that customers seek and to find ways to deliver that value with least complexity.
Also it can be argued, perhaps paradoxically, that a focus on complexity reduction could increase supply chain risk. For example, a too-ambitious programme of supplier rationalisation could leave the company vulnerable to disruption if for whatever reason a critical supply source were to fail.
Complexity management in the supply chain has to be a careful balance between over-simplification on the one hand and a focus on cost and efficiency on the other. The aim should be to reduce or eliminate any complexity that does not add value to the customer or that does not protect against supply chain risk. The impact that complexity can have on supply chain risk is well illustrated by the case of the Boeing 787 described below.
The Boeing 787 Dreamliner: an outsourcing nightmare
On 15 December 2009, over two years later than originally planned, the Boeing 787 – the so-called ‘Dreamliner’ – made its maiden test flight. The 787 was a radically new concept embodying highly innovative technology and design features. The market positioning of the aircraft had proved to be highly successful, with pre-launch sales options from airlines around the world in the region of 850 planes. With a passenger capacity of up to 330 and with a range of 8,500 nautical miles the 787 would use less fuel and operate at a cost-per-seat-mile somewhere in the region of 10 per cent less than other comparable aircraft. These savings were enabled primarily by the lower weight of the 787, which was achieved through the use of novel composite materials, and new engine technology.
Even though most industry commentators expected that in the long-term the 787 would be a great success, there was no doubt that the delay in the launch had impacted negatively on Boeing’s financial performance.
Clearly, a design as innovative as the 787 brought with it many challenges because much of the technology was untried and untested. Beyond this however, there were a number of risks that were systemic, i.e. risks that arose as a result of decisions taken by the company on the precise form of the chosen supply chain architecture.
Traditionally Boeing has built most of its aircraft in the own facilities in Washington State, USA. In the case of the 787 the only part manufactured in their Washington factory is the tail fin (and even this manufacturing is shared with another facility outside Washington). The other parts of the aircraft are manufactured as sub-assemblies by a myriad of external suppliers around the world. For example, the forward fuselage and nose are made by Spirit AeroSystems in Witchita, Kansas, whereas parts of the midsection are manufactured by Alenia in Italy and the wings and a further fuselage section are built by companies in Japan. The final assembly of the aircraft takes place in Boeing’s facilities in Everett, Washington and Charleston, South Carolina.
Not only has the manufacture of most of the sub-assemblies been outsourced but those same suppliers were also involved in much of the detailed design of the sections/systems they were responsible for. Perhaps not surprisingly a number of problems were encountered.
Many of the suppliers found that the innovation involved challenged both their design and their engineering capabilities. Boeing had to send its own staff to help the suppliers sort out these problems. Often sub-assemblies would arrive at Everett incomplete or wrongly manufactured, requiring disassembly and re-building. Months were lost in the process of putting things right. These delays had financial consequences and the cost of additional design, re-work and penalty payment ran into billions of dollars.
The paradox is that the business model adopted by Boeing, i.e. outsourcing the design and manufacture of sub-assemblies to supply chain partners, was motivated by the aim of speeding up time-to-market. The original view at Boeing was that using external specialists would enable a more flexible supply chain, capable of responding more rapidly to customer demand. In the event the outcome was a significant delay in time-to-market and a major cost over-run.
Undoubtedly a product as innovative as the 787, embracing as it does entirely new materials and technology, would always face significant challenges. However, beyond this, Boeing’s experience highlights the fact that whilst companies might outsource the execution of an activity they should never outsource its control.
SOURCES: ‘DREAMLINER MAKES HISTORY WITH PLASTIC, OUTSOURCING, DESIGN – AND DELAYS’, THE SEATTLE TIMES, DECEMBER 12, 2009.
‘JET BLUES: BOEING SCRAMBLES TO REPAIR PROBLEMS WITH NEW PLANE’, THE WALL STREET JOURNAL, DECEMBER 7, 2007.
References
1. Whyte, C., ‘Motorola’s Battle with Supply and Demand Complexity’, Supply and Demand Chain Executive, 12 August 2004.
2. Gottfredson, M. and Aspinal, K., ‘Innovation vs Complexity: What is too Much of a Good Thing?’, Harvard Business Review, November 2005, pp. 62–71.
3. Anderson, C., The Long Tail: Why the Future of Business is Selling Less of More, Hyperion, New York, 2006.
4. Appelqvist, P., Lehtonen, J.M., Kokkonene, J., ‘Modelling in Product and Supply Chain Design: Literature Survey and Case Study’, Journal of Manufacturing Technology Management, Vol. 15, No. 7, 2004, pp. 675–686.