In the previous edition of this Journal, we outlined a three-phase research program to explore ordering and scheduling practices that lead to earliness and delays in materials and equipment delivery in capital projects. Usually, owner-operators and their EPC’s look to minimize the risk of schedule delays due to late materials and parts delivery by mandating that parts and materials be delivered far in advance of when they are needed. In contrast, many other industries, including automotive, retail and technology, coordinate orders and deliveries more closely with actual needs to better optimize overall system performance.
We hypothesize that holding large amounts of inventory long before it is needed on capital project sites is a symptom of sub-optimal supply chain operation. To better understand these issues, we have completed the first phase of our research program – a set of interviews with a variety of consultants and project and supply chain managers in the oil & gas industry, during which we asked them to respond to a set of questions so that we could better understand current industry views of supply chain management, inventory, risk management tools and related topics. In this digest, we summarize initial findings from the interview responses and outline steps to explore additional questions they have raised.
Keywords: Supply Chain; Supply Chain Operations; Supply Chain Flow; Procurement; Project Supply Chain; Supply Network
The majority of the investment made by an owner-operator during the delivery of a capital project is allocated across a complex network of product suppliers and service providers. Currently, it is often the case that owner-operators and their EPC’s look to minimize the risk of schedule delays due to late materials and parts delivery by mandating that parts and materials be delivered far in advance of when they are needed. This is in contrast to many other industries, including automotive, retail and technology, where the timing of orders and deliveries is more closely coordinated with actual needs in order to minimize system costs and better optimize overall system performance. We hypothesize that the existence of large amounts of inventory long before it is needed on capital project sites is a symptom of a sub-optimal supply chain operation. This view is supported by several studies which, in an empirical analysis, find that a majority of capital projects are completed late and over budget. A recent example is by McKinsey & Company [1], whose research concludes that 98% of mega-projects are facing cost overruns of more than 30%, and 77% of these projects are at least 40% late. A study conducted by E&Y [2] based on 350 mega-projects, specifically in the oil and gas industry, shows that 64% of projects suffer cost overruns and 73% of projects report schedule delays.
These observations support the notion that there is room for improvement in the capital project supply chain. To better understand these issues, we interviewed a variety of consultants and project and supply chain managers in the oil & gas industry. We asked them a set of questions so that we could better understand current industry views of supply chain management, inventory, risk management tools and related topics.
In a longer article to follow, we describe the framing and structure of the interviews, summarize the interview responses, identify several interesting findings and highlight additional questions for future research. In this brief digest, we merely summarize a number of provocative takeaways that frame the agenda of the 2017 Project Production Symposium, whose theme is on optimizing supply networks. We use the term “Supply Network” rather than “Supply Chain” to convey the notion that these structures are not simply linear sequential structures, but rather can be a complex web of relationships.
Our interviews generally gave us the sense that some of the managers we interviewed had a limited view of supply chain management. In most industries, supply chain management is understood to be the management of the flow of materials, information and finances, both within and among firms to optimize the overall system performance. In contrast, many of the managers with whom we spoke view supply chain management as essentially synonymous with procurement – managing costs and quality of suppliers.
While supply chain management (at least in progressive firms and industries) involves managing these flows in an integrated way, our interviews suggest a capital project “supply chain” consisting of independent islands connected by narrow bridges. Scheduling seems to work independently of procurement, which in turn doesn’t coordinate with suppliers.
This is not to say that the challenge is unrecognized – interviewees highlighted the use of regular multi-disciplinary meetings with stakeholders from the project and supply chain functions aimed at improving coordination. However, even when regular meetings exist to facilitate coordination, these efforts seem less than effective. Firms appear to lack a comprehensive model to optimize supply chain activities.
Because project schedules are often created and managed using vendor software with complex and sophisticated functionality, many believe that operations are optimized. In many cases, however, cost analysis is not integrated with schedule production. It appears from the interviews that most software tools appear not to have this type of functionality. Typically, according to our respondents, the schedule is an input to the procurement process.
From the perspective of optimization theory, this is a greedy or myopic approach, in which the focus is first on one decision, and then (sequentially) on the next. By “greedy” we mean that without a global view of the system, only locally optimal decisions at each stage of a project can be made. Capital projects are typically complex supply chains with many thousands of decisions and constraints – without comprehensive tools and models, only limited, greedy improvements are possible. The resulting performance can be very far from the optimum achievable, because the impact of thousands of sub-optimum decisions can cascade throughout the system, having a cumulative impact that gets magnified exponentially.
Clearly, the total system cost can best be minimized for any level of performance if all decisions are considered simultaneously. Otherwise, each decision constrains the possible set of other decisions, and in general the best decision (for the entire system) will not be found.
This kind of greedy, myopic, sequential optimization leading to locally optimal solutions seems to be a significant issue in capital project supply chains. It is the sort of complex supply chain with many interrelated participants for which many locally optimal solutions tend to exist. To improve the overall objective in this setting, many simultaneous changes need to be made. Typically, finding these kinds of improvements is challenging – while a regular meeting of stakeholders (scheduling department, procurement department, project managers, suppliers, etc.) in capital projects can help find a good locally optimal solution, sophisticated tools are necessary to understand the impact of changing many (perhaps thousands) of decisions simultaneously.
The performance of a project is typically evaluated by comparing its execution with its planned schedule. However, some of the experts with whom we spoke believe that schedules are often not as effective as they could be. Since scheduling is primarily conducted separately from procurement, it is challenging to find the root cause of certain execution issues. For example, assume an essential component or piece of equipment is considerably delayed, leading to delays and overage for the overall project. Is this the result of uncertainty in delivery lead times, or is this the consequence of an insufficiently robust schedule? We call a schedule robust if the overall schedule objectives remain relatively constant when local changes are made to the schedule.
Consider the following example, consisting of a simple project with four different jobs. Job 1 is to precede Job 3. In addition, Job 1 needs equipment item 1 and Job 3 needs equipment item 2. Jobs 2 and 4 can use either piece of the equipment. Each job takes one time unit on average. However, there is some uncertainty in the duration of Job 1 and Job 2, perhaps due to procurement issues. Figure 1 presents the activity-on-node diagram for this project.
Figure 1: The activity-on-node diagram for the example project
In terms of minimizing the length of the critical path – the sequence of tasks that takes the longest time to complete among all possible sequences – both schedules presented in Figure 2 are equivalent.
Figure 2: (a) An optimal schedule that minimizes critical length (b) A robust optimal schedule that minimizes critical length
However, recall that Job 1 and Job 2 can extend to two units of time in the worst case. Then, the duration of the project will be 4 using schedule 2-(a) while it will be 3 using schedule 2-(b). This trivial example highlights the difference in robustness of alternative schedules. In practice, sophisticated tools are necessary to estimate schedule robustness, and an integrated approach to financial management, project scheduling and other components of supply chain management is necessary to determine a schedule that is robust enough to respond to changes in each of these areas.
Over the course of our interviews, we were struck by the vastly different points of view of our interviewees, particularly with respect to the future of capital project delivery, and to the possibility of more closely coordinating supplier deliveries with onsite need. While some of the interviewees believe that something closer to a just-in-time delivery model is entirely inappropriate for capital project delivery, others believe that it is essential for industry to move in this direction to achieve what manufacturing industries have achieved in their supply chain.
Several interviewers highlighted the importance of experience – one of the interviewees told us his confidence in his firm’s effectiveness increases with the experience of the planners. Of course, every system works better with the benefit of experience, but it is important to distinguish between the experience of the system and experience of individuals in the system. A comprehensive model for optimizing the supply chain would be a tool that enables a firm to improve performance over time by analyzing weaknesses on specific projects, so that these weaknesses can be addressed in subsequent projects – experience remains with the firm and benefits the firm. In contrast, our interviewees suggest that, particularly in oil and gas project delivery, experience remains with individuals, as decision-making is frequently based on intuition and unrecorded personal observations. If this is common practice, the limitations on knowledge transfer and therefore value to the firm are significant to optimizing future projects.
For example, in most manufacturing firms, safety times used to mitigate costs of delay are determined based on the tradeoff between capital tied up in the inventory, risk of delay and other relative financial metrics. Based on our interviews, it seems that in this industry these kinds of decisions are instead made using rules of thumb – there is clearly an opportunity, if not an imperative, to use comprehensive models to improve this kind of decision-making. Most software tools used for entering project schedules and risks appear not to have this comprehensive modeling capability.
The impact of uncertainty on project performance is clear – uncertainty drives risk and cost, safety times and due dates, and inventory strategies. That said, uncertainty is not static. It typically resolves over time, though at what cost? It is not clear that this notion of dynamic uncertainty has been integrated into project plans.
As activities get closer, the amount of uncertainty associated with their start times and durations typically reduces. In principle, the plan could be adjusted to account for this reduction. For example, consider the impact of a mandated four-month safety time. All decisions regarding delivery or relevant parts and equipment must therefore be made sooner than if this time were shorter. Reducing the safety time to a week or two would not only decrease the necessary working capital, it would also reduce the amount of uncertainty that must be accounted for.
The first step in resolving a problem is to acknowledge its existence. Based on our interviews, it seems that many of the participants in capital project supply chains believe that the supply is operating close to optimal. We hypothesize that this mindset is a consequence of, among other things, local or greedy optimization. Since there is no obvious way that small, local changes can lead to improvements, participants believe that the current approach is optimal. A comprehensive model-based analysis could help to highlight the inefficiencies of current approaches.
Building this kind of comprehensive model will be challenging. Many variables and constraints, and the relationships between them, must be considered. A sequence of locally optimal models is clearly easier to build and use, although for the reasons highlighted above, this approach will ultimately prove ineffective. Decisions that locally decrease cost or risk may end up increasing cost or risk for the entire project.
Often, when we help managers analyze their supply chains, we begin with the following question: Imagine that you own every part of the supply chain – how would your decisions differ? In this setting, most managers realize that their decisions would differ, and this can lead to an analysis of how the supply chain can be globally optimized, and how the advantages of doing so can be shared.
Of course, while the benefits of globally optimizing the supply chain are incredibly promising, there are reasons why it is not typically done. This sort of comprehensive model is challenging to build, validate, and optimize due to the size and the decentralized nature of a typical capital project supply chain. Furthermore, most suppliers have multiple customers, each with their own competing needs. How will suppliers act and react to incentives in a given system design? Are suppliers predictable? Are the prices they offer related to the other jobs they might have previously accepted, or to anticipated future business? Will suppliers use their own models to make decisions? The literature in this area is limited.
Some experts believe that standardization of the procedures and modularization of designs can help address some of these issues and boost the efficiency of capital project supply chains. However, given the lack of comprehensive models of these kinds of supply chains, the scope of this impact is unclear. A more comprehensive modeling approach could also help to answer some of these questions.
Our initial conclusions from the interview responses suggest that a logical next step is to show how to build a comprehensive optimization model to coordinate project planning and scheduling with other supply chain-related decisions in order to globally optimize the supply network in terms of project delivery schedule, cost, and risk management. In a longer article to follow, we will elaborate on the analysis of user responses and detail the next steps in such a research enterprise.
Arman Jabbari received his B.Sc. and M.Sc. in Industrial Engineering from Sharif University of Technology in 2013 and 2015. Given his interest in optimization, he found it optimal to pursue his education at one of the best engineering schools in the world: UC Berkeley.
He received his second M.Sc. from UC Berkeley in 2016, and he is pursuing his Ph.D. in Industrial Engineering and Operations Research, which is expected to be granted in 2020. During his graduate studies, he developed an interest in project supply chain management, where he can apply his passion for optimization to overcome real-world obstacles.
Phil Kaminsky received his B.S. in Chemical Engineering from Columbia University in 1989, and his M.S. and PhD in Industrial Engineering and Management Science from Northwestern University in 1997. Before graduate school, he worked in production engineering and control at Merck in Rahway, New Jersey.
Professor Kaminsky is the faculty director of the Center for Entrepreneurship and Technology, and director of the Initiative for Research in Biopharmaceutical Operations. He became department chair of Industrial Engineering and Operations Research at UC Berkeley in 2012.