In a 2015 report sponsored by PPI, Professors Iris Tommelein and Glenn Ballard of the University of California, Berkeley, analyzed the underlying principles of the construction practice of Advanced Work Packaging (AWP) and provided a critique to its standard design, making a number of predictions. Fundamental in their critique is the operations policies and parameters that dictate the final behavior of a project production system when AWP is applied. They predicted undesirable consequences to the performance of Project Production Systems including, but not limited to, inventory growth, longer cycle times, lower labor utilization and higher costs.
Using an Operations Science framework, this paper reviews a capital project example that implemented AWP and uses this real-life project experience to validate each of the predictions previously made by Professors Tommelein and Ballard. This paper also introduces the technical approach that the owner operator adopted to address the project performance gaps created by AWP, which included the application of CONWIP control protocols for more effective control of supply of materials and permanent equipment to the point of installation.
There is abundant published literature    , proclaiming that project cost and schedule overruns in the energy and industrial sectors have reached crisis levels. These cost and schedule overruns are impacting shareholder value and the ability for owner operators, including energy and industrial companies, to deliver and maintain their assets  based on business objectives.
How work is chunked into packages for the purpose of execution is not a new or innovative practice in the engineering and construction industry. But in recent years, these practices have evolved into established methodologies such as Workface Planning (WFP) and then Advanced Work Packaging (AWP) , gaining much interest among owner operators and Engineering-Procurement-Construction (EPC) firms. Industry organizations such as the Construction Industry Institute (CII) and the Construction Owners Association of Alberta (COAA) have invested in better defining and documenting these practices, resulting in the publication of a series of reports, such as the latest CII Report 272-12, “Advanced Work Packaging: From Project Definition Through Site Execution.”
Based on the interest of industry practitioners and through PPI-sponsored Research, Professors I. Tommelein and G. Ballard reported , including a subsequent paper , that WFP and AWP, as currently defined in the industry, will result in a Project Production System that is “under specified.” They proceed to identify several areas where system design choices remain to be made, indicating ways to optimize execution of work packages, and in so doing, support better control of project cost and duration including increased predictability. In , they proceed to formulate a number of predictions on the negative outcomes from the implementation of AWP resulting from the incomplete specifications in the Project Production System.
In this paper, we describe the unintended consequences of implementing AWP / WFP in the ability of project teams to effectively synchronize project supply flows, validating the predictions made in . After describing the production system resulting from the implementation of Construction Work Packages (CWP) and Installation Work Packages (IWP), we analyze the implications of its policies and parameters for supply flows, particularly the build-up in inventory of materials and permanent equipment arising from variability in both demand and supply and the ability of logistics systems and their capacity to cope with variability. Using Operations Science principles, we outline an alternative design of a section of the overall Project Production System to better synchronize supply with demand, leading project teams to unlock and capture value such as shorter cycle time, reduced inventory and greater reliability.
Unintended Consequences – Predictions by Tommelein & Ballard
Because Tommelein and Ballard made the case that WFP and AWP comprise an under-specified project production system, we will now review a summary of their predictions  as a list of unintended consequences resulting from inadequate polices and parameters, and then move to a case example that validates their predictions. Tommelein and Ballard summarized their findings in four predictions as follows:
- Failure to specify pull in AWP design, naturally results in push. We expect to see huge inventory growth on projects using AWP if they do not specify pull.
- The larger the transfer batch, the longer the duration of the process. We expect to see projects taking longer rather than being done more quickly, unless they reduce the transfer batch.
- Having defined work packages, coordinated between engineering and construction, does not reduce the challenge of coordinating massive flows of materials, information and resources to construction sites when needed. We expect labor utilization to get worse rather than better.
- Inventory growth, longer project durations, and higher labor costs – plus increased costs for expediting and firefighting – are expected to result in projects well over budget and time.
An owner operator invests several billion US dollars in an industrial processing facility. Due to its magnitude, the project can be classified as a classic megaproject. Procured through conventional means, different materials, parts, equipment (e.g., vessels) and modules from various suppliers are delivered to the construction site, resulting in large volumes of inventory residing onsite until demanded by the construction teams for installation. To better illustrate this, Figure 1 shows examples of what physically manifested onsite and different types of parts and materials that are stored in inventory, while Figure 2 provides a high-level representation of part of the overall Project Production System that is the focus of this paper, and more specifically, the onsite supply of materials to the point of installation.
Figure 1: Examples of inventory at site
As the project team implemented AWP / WFP, materials were then kitted according to specific IWPs and placed on trailers for transportation to the point of installation, as illustrated in Figure 3. This is another representation of onsite inventories, but in this case, the difference is that inventory is combined with capacity (e.g., trailers), not to mention the use of space, as well as the fact that trailers are destined to a specific area becoming an engineered-to-order (ETO) supply process (not interchangeable between areas or IWPs).
Figure 3: Trailers used for transportation of materials to the point of installation
Despite significant investment in capacity (labor, trailers, staging areas, etc.), the installation and associated onsite supply flows (from staging areas to the point of installation) experienced severe problems resulting in lack of synchronization between onsite supply and demand. The installers were saying “we cannot get what we need, when we need it.” On the other side, those in charge of onsite supply were saying “installers don’t know how to plan – they change their minds too often and don’t take what they ordered.”
The combined effect of this situation was schedule delays (e.g., site teams did not get what they wanted it) and additional costs (e.g., higher than expected logistics costs and cost of schedule delays), not to mention the exposure to loss of revenue by the owner operator because of the delays in the project coming online. Safety risk levels and quality standards were also compromised because of the increased and constant handling of inventory (e.g., offload, sort, tag, load, transport, offload, move to the point of installation).
Project Production Management Analysis
We should start by asking ourselves: what was the project team (construction and onsite supply teams) doing that caused these outcomes?
To answer this question, it is important that we analyze the situation described herein from a production perspective, including what policies were implemented. To achieve this, we must start by visualizing the production system including how transformation, stocks and flow of work occur. Figure 4 shows a Process Flow Diagram that provides a more detailed representation of the production system illustrated in Figure 2. It shows the sequence of main activities as materials and information flow from the makeup of CWPs to the assembly, storage and transportation of IWPs to the workface and their respective execution.
Figure 4: Production System covering CWPs and IWPs
Based on this Process Flow Diagram as well as other information gathered during this analysis, the following can be stated about the design of this production system:
- While a CWP comprises all work for a given area for multiple trades / disciplines (e.g., steel, piping, mechanical, electrical, instrumentation, etc.), each IWP was developed per discipline (e.g., piping, electrical) and based on approximately 1500 to 3000 man-hours of work, the equivalent to approximately 5 trailers-worth of materials and 2-3 weeks of work for a crew of between 7 to 10 people (this could vary).
- A project policy mandated that an IWP shall be created 12 weeks before site needs it. This represents a process that starts with compiling the engineering for each IWP and preparing the Bill of Materials (BOM) plus developing a plan for the execution of the single-discipline IWP. This also includes checking if materials are onsite and if the required capacity is available.
- With an IWP created (at a document stage including the required verifications), construction teams are required to communicate the demand for an IWP to those in charge of onsite supply 21 days in advance of the need date (T-3 weeks) by completing and submitting a Field Materials Request (FMR) form. FMRs were used as triggers to start the physical creation of an IWP (primarily one FRM per IWP). As the size of an IWP is about 2-3 weeks of work (could be less), a single FMR represented this volume of materials. Level 4 schedules were used not only to determine required-at-site (RAS) dates but also to establish when an FMR was to be generated (refer to Figure 5).
- Upon receipt of an FMR, the onsite supply process was initiated, including locating and flagging the materials, loading empty trailers with IWP materials (was uncommon to see materials for multiple IWPs loaded on the same trailer), staging the trailers waiting for site teams to come and collect them, and then moving the trailers to the workface for installation.
- A situation that became common was that not 100% of the materials for a given IWP were onsite as delays in engineering caused a ripple effect in procurement, fabrication and supply to the site. Therefore, the team implemented a policy that only FMRs with 100% of the materials listed in the BOM physically on site can be submitted to the onsite supply team.
Figure 5: Project Controls Driving Deliveries to the Point of Installation (Current State Design)
Given the above description of how the production system was designed including policies and parameters, Operations Science allows us to deduce consequences of the system choices shown in Figure 4 and Figure 5, which in fact will begin validating the predictions of Tommelein and Ballard. Based on this, the following conclusions can be made:
- AWP / WFP uses a strategy of placing large inventory buffers of materials and information to shield installation craft, but ignores managing predecessor WIP, capacity and variability.
- The choice of using Level 4 schedules, typically used for monthly reporting and forecasting of progress (project controls ), as the means to set delivery dates is essentially a control mechanism that ends up incentivizing the push of materials to site because these type of schedules rarely reflect the real status of work in the field. This is compounded by the fact that construction teams were required to forecast the demand for materials three weeks in advance (21 days) using these schedules. As daily reliability of planning was between fifty and sixty percent, meaning every five / six of ten activities were not being completed as planned every day, construction teams did not stand a chance of providing a reliable forecast (let’s remember that by definition, all forecasts are always wrong), so the accuracy of demand was very low. Photos of trailers presented before provide hard evidence that the creation of a huge inventory of materials and capacity (the trailers and space requirements for the trailers) is a result of an unsynchronized supply system. This supports the first prediction.
- Little’s Law (Work-in-Process = Throughput x Cycle Time) allows us to deduce the Work-In-Process (WIP) in the system measured from the point where an FMR is submitted and the time an IWP is to be completed. It was estimated that, on average, a single construction team required up to 20 trailers per day (Throughput = 20). As the cycle time for the delivery of trailers was set as 21 days (from order to fulfillment), we can calculate that at any given time, a WIP equivalent to 420 trailers (=20×21) should be expected in the onsite supply system.
In fact, due to variability in the system, this number was even greater, reaching about thousand trailers at one point. The transfer batch (the size of the FMR) was large and set the cycle time to 21 days, and within that time window, demand was highly variable so not all trailers that were moved to the point of installation were used when requested. This resulted in a decision by management to increase the number of trailers (more capacity and more WIP). It is important to highlight that because site work was exposed to high levels of variability, it meant that construction teams needed other IWPs immediately that were not yet on trailers, creating changes in demand. But because supply was set for 21 days of cycle time, the system was not flexible enough to account for this, contributing to delays in the schedule as well as low capacity utilization (people waiting for materials). This validates the third and fourth predictions.
In summary, order batches were too large which contributed to longer than needed cycle times. Variability in demand was high which led to excessive WIP. Excessive WIP caused extra costs (e.g., more trailers and the required capacity to maintain those in use) and the subsequent schedule overruns. This combination of factors conspired to yield the delays and cost overruns we have described, fulfilling all four predictions made by Professors Tommelein and Ballard in their report .
How to Address the Situation through Project Production Management
The previous analysis has exposed four important characteristics of the production system: 1) highly variable demand, 2) excessive WIP levels, 3) large size of order batch and 4) long cycle times. From a production perspective, and to achieve better performance, the proposed Project Production Management (PPM) strategy consisted of the following elements:
1. Increase the reliability of demand through the implementation of a robust Project Production Control solution to enable variability reduction (variability cannot be eliminated, but it can be minimized)
2. Reduce the size of the order batch to one day of work (adopt the concept of production packages)
3. Reduce cycle time as to fit order fulfillment within the window of reliability of site teams
4. Discontinue the use of Level 4 schedules as the means to drive the delivery of materials to the point of installation and implement a Constant-Work-in-Process (CONWIP) system of control using a production schedule as the means to trigger supply
5. Optimize the production system during execution as to further reduce WIP and variability, and therefore reduce cycle time
Figure 6 illustrates the proposed adjustments to the production system, based on Operations Science. For the above-mentioned strategy to instill the desired behavior in the production system, new policies needed to be defined and made operational. Examples of these policies included, but were not limited to:
- Initiate production of a production package from the work site (CONWIP control)
- Contain no more than one day of field installation work in a production package
- Include all necessary technical information in a production package
- Check for accuracy of the production package at the fulfillment point and notify of issue
- Kit today what will be delivered tomorrow
- Optimize offloading sequence and parts presentation / arrangement to the extent possible
- Optimize kitting capacity within the day, no more
- Deliver today what will be installed tomorrow
- Ensure production package in sound when receiving at point of installation
- Use trailers for staging when possible
- Offload from trailer to point of installation when possible
- Remove from site what is not consumed daily (waste, package materials, rejected production packages, etc.)
- Inspect and accept completed work on a daily basis
Figure 6: Proposed Production System Adjustments
Although all the above elements of the strategy are critical to achieve better performance, the reduction of cycle time is key to this strategy. A shorter cycle time enables reduction of WIP and the capacity required to maintain the WIP. In this case, fulfilling demand within five days was doable according to the chart presented in Figure 7. However, teams did not trust the behavior of the old system and initially resisted to a 16-day cycle time reduction (from 21 to 5 days). Because of this, gradual adjustments were made to cycle time starting with a reduction of 11 days to set cycle time to 10 days. Compared with 21 days from request of an IWP to delivery to point of installation, even a cycle time of 10 days had direct implications for the amount of WIP in the system. For instance, with required average throughput being 20 trailers per day and with a cycle time of 10 days, WIP is reduced from 421 trailers to 200 trailers (=20×10). With a cycle time of 5 days, WIP is further reduced to 100 trailers – less than 25% of the WIP in the original system.
|AUTHORIZE TO LOAD||LOAD||MOVE TO POI||AT POI||INSTALL|
|Day -4||Day -3||Day -2||Day -1||Day 0|
Figure 7: Five Day Cycle Time to Pull a Production Package
We described an actual case example that supports and validates the predictions of Professors Tommelein and Ballard in their critique of WFP and AWP  . Operations Science provides a very robust technical framework to analyze and identify opportunities to improve production system performance including how to control WIP, reduce variability and cycle time, even in cases like the one presented herein that addressed onsite supply to the point of installation.
Through the case example, this paper illustrates several fundamental Operations Science principles of Era 3 project delivery thinking compared with classical Era 1 and 2 thinking . It also validates the fact that AWP / WFP uses a strategy of placing large inventory buffers of materials and information to shield installation craft, but ignores managing predecessor WIP, capacity and variability. Although the impact of variability can be buffered through a combination of capacity, time and inventory, the larger the buffers, the higher the costs.
As illustrated in this case, real constraints on capacity (e.g., limited number of trailers and space) and time inevitably lead to negative implications for WIP, which means cycle times will be longer. Through the proposed design of the production system, variability can be further mitigated (but not eliminated) as well as WIP better controlled as to reduce cycle time. Finally, it is imperative to move from controls (as in project controls) as the means to trigger supply to a more effective means of control and optimization, such as the application of a CONWIP control protocol.
 D. Tideman, H. T. Tuinstra and B. J. Campbell, “Large Capital Projects In The Oil And Gas Sector & Keys To Successful Project Delivery,” Strategy&, 2 October 2014.
 Ernst & Young, “Spotlight On Oil and Gas Megaprojects,” Ernst & Young, 2014.
 S. Changali, A. Mohammed and M. van Niewland, “The Construction Productivity Imperative,” McKinsey Productivity Sciences Center, 2015.
 F. Barbosa, J. Woetzel, J. Mischke, M. J. Ribeirinho, M. Sridhar, M. Parsons, N. Bertram and S. Brown, “Reinventing Construction Through A Productivity Revolution,” McKinsey Global Institute, 2017.
 W. J. O’Brien, F. Leite, O. Hamdi and S. Ponticelli, “Advanced Work Packaging: From Project Definition Through Site Execution,” Construction Industry Institute, Austin TX, 2016.
 I. D. Tommelein and G. Ballard, “A Critique of Advanced Work Packaging,” Project Production Institute, San Francisco, 2015.
 I. D. Tommelein and G. Ballard, “”Advancing” Advanced Work Packaging,” Journal of Project Production Management, vol. 1, no. Winter, 2016.
 R. J. Arbulu, H. J. Choo and M. Williams, “Contrasting Project Production Control with Project Controls,” Journal of Project Production Management, vol. 1, no. Winter, 2016.
 R. G. Shenoy and T. R. Zabelle, “New Era Of Project Delivery – Project As Production System,” Journal of Project Production Management, vol. 1, no. Winter, 2016.