Process innovation at the Sandy lumber mill.

Metlen, Scott ; Lawrence, John J.

CASE DESCRIPTION

The primary subject matter of this case concerns the value of quality control and continual system improvement. Secondary issues examined include project implementation, and quality management issues. The case has a difficulty level appropriate for undergraduate seniors towards the end of a semester quality management class and for graduate students. The concepts presented in the class are not trivial and several hours during class can easily be used to discuss all issues. A student's degree of understanding of process control issues, financial analysis, and quality management issues dictate the amount of time out of class each will spend to address case issues. Most students will need to spend a minimum of four hours to address all issues.

CASE SYNOPSIS

The Sandy sawmill produced dimensional lumber. The mill had recently completed a $2.6 million process upgrade that had allowed it to improve yields and better match lumber produced to market conditions. Revenues had increased from $29,000,000 to over $40,000,000 as a result. But challenges and opportunities remained. The milling process was theoretically capable of producing closer to specification, and operators still had difficulty identifying where in the process defects originated. Further, the mill manager was somewhat overwhelmed by the amount of data being produced and how best to use this data. The case ends with the mill manager wondering what he should do next to get the most out of the new system. This case was designed for use in a quality management class to facilitate discussion of the design and implementation of a statistical process control systems.

INTRODUCTION

Tony Flagor watched the computer screen display real time information on the boards being processed at the Sandy Saw Mill and marveled at the new control process now in place. He was proud of the changes that he and his team had made. Nevertheless, even though the new control process was vastly superior to the one that had been in place when Tony had first become mill manager, it still did not seem to be adequate. The milling process consisted of many steps, each one representing a unique opportunity for a quality problem. Unfortunately, the control process did not display which task caused an out of specification board, only that a given place on a given board did not meet specifications. Furthermore, the measurement system in place had the potential of creating an unfathomable number of data points per minute, and Tony was unsure how best to use all of that potential data. What Tony did know was that he needed to continuously improve the effectiveness of the mill in order to stay competitive, and he believed that effective real time process control was key.

The US lumber industry in which the Sandy Mill competed was highly competitive, and generating profits had become an increasingly challenging task for everyone in the industry. An 18% increase in lumber imports from 1997 to 2002 (Kelly, 2003) and increased effectiveness of processing technology had created a situation of high supply relative to demand despite diminished supplies of raw material due to closure of many national forests to logging to protect endangered species. From 1997 to 2003, 117 (14%) of the 829 United States softwood lumber mills went out of business (Spelter, 2003) and the number of trees harvested declined by 6.85% (United Nations Economic Commission for Europe, 2003), yet lumber production in the USA was static for the same time period (Kelly, 2003). The surviving operations had to increase effectiveness (extract more boards of higher value from a given log at lower operating costs) and gain economies of scale to stay in business. The Sandy lumber mill was a mill that so far had been successful in increasing lumber production effectiveness.

The mill manager in charge of milling operations and milling improvements at the Sandy mill was Tony Flagor. Tony came to the Sandy mill from a career in the Navy as an electrician where his responsibilities included electrical maintenance and training. After the Navy, Tony earned a BS and MBA from the University of Washington. Tony's first job at the Sandy mill was as maintenance manager after first working for one year as an electrician in an adjacent plywood mill, which was also owned by the company that owned the Sandy mill. The position of maintenance manager was for both the Sandy mill and the plywood mill. Within a year of taking the maintenance manager position, Tony applied for the mill manager position at the Sandy lumber mill. Much to the surprise and anger of some longtime employees of Sandy who had also applied for the same position, Tony became the new manager of Sandy. There were many reasons Tony was selected for the position. His personal work ethic and ability to work with everyone but still make and enforce tough yet fair decisions regarding subordinates enabled Tony to have productive work centers. As important as Tony's human resource ability was in management's decision to make Tony the new manager of Sandy, it was Tony's willingness and ability to incorporate new technology into the milling processes that cinched management's decision.

Under Tony's direction, the Sandy mill undertook and accomplished a massive process redesign, which allowed the mill to increase the level of profits it generated for its parent company. Tony was pleased with the new processes because they were consistent with the strategic direction of the parent company, which involved focusing on those business units that had and could maintain a competitive advantage and that had potential to grow. Despite the success of the new process, Tony knew that the competition was not resting and that he would have to keep improving process effectiveness in the most cost effective manner possible to maintain and strengthen the competitive advantage that the mill currently had. Toward that end, Tony had to decide what the next step was to continue to improve the effectiveness of the mill.

COMPANY BACKGROUND

The parent company of Sandy was founded in the early 1900s. The company was a vertically integrated, diversified forest product company that owned 1.5 million acres of timbered land and operated 15 mills. The mills included lumber mills such as the Sandy mill, paper mills, particle-board mills, oriented strand board mills, and plywood mills.

The Sandy mill was a softwood lumber mill that had been in operation since 1972. It was nonunion, paid above industry average wages (one person year did cost the company an average of $50,000.00), and had the capacity of producing 85,000,000 board feet per year from logs ranging from five inches to 18 inches in diameter and eight to 18 feet long. Unlike many current mills, Sandy was expected to create lumber from multiple species of trees; Douglas, Alpine, and White Fir, Engelmann Spruce, Cedar, and, Lodge Pole Pine. To take advantage of this expected flexibility, the milling process required a greater amount of operator knowledge because mill operators had to understand the peculiarities of production inherent to each species. True flexibility also required a process that had fast, easy setups so that operators could quickly start processing of a different species to meet dynamic market demand. Up until 2002, the mill had an efficiency driven production process where the focus was only on making as much lumber as possible as cheaply as possible. Mill managers paid less attention to what species and dimension was currently in demand in the marketplace, but rather focused on how to get the longest production run for a given species. This efficiency focus was driven by the nature of the process in use at that time.

MILLING PROCESS, PRE-2002

The milling process as of 2002 consisted of a traditional lumber mill process that was labor intensive (Exhibit 1). The mill produced an average of 85,000 million board feet (MMBF) (one MMBF equals1000 board feet) of dimensional lumber per year. A board foot is a metric used in the lumber business measuring 1 inch x 12 inches x 12 inches and dimensional lumber refers to the width and depth of the boards produced. The original mill produced 2" X 4", 2" X 6", 1" X 4", and 1" x 6" dimensional lumber. To simplify calculations, all boards in this case will be assumed to be 2" X 4" where one lineal foot equals .67 board feet and the average board is 14 feet long. The mill operated 202 days per year with two, ten-hour shifts per day. The lumber produced sold at an average selling price of $347 per 1000 board feet (m.bd.ft) or one MMBF.

The start of the process consisted of a sawyer on a log carriage. The sawyer would quickly inspect a log and with partial help from a computer decide how the log should be cut to get the most lumber the quickest. The lumber the sawyer made would proceed by conveyor belt to inspectors who would sort the lumber by dimension, length, width, and depth (see exhibit 1 for a schematic of the process). Seventeen people per shift were involved in the total process; most making many qualitative decisions that would affect the quality and volume of process output. To help make these decisions more accurate and easier, and to keep all workers busy, large batch sizes of uniform logs of one species were the norm. Thus, to produce the volume of output needed, the mill often produced lumber from certain species and in certain dimensions that had lower market value relative to lumber with other dimensions or from other species in the interest of achieving higher milling efficiency and output. This practice created a focus on efficiency, not market effectiveness.

In addition to the focus problem created by the perceived need to produce large batch sizes effectuated by a labor-intensive process, lumber quality and yield per unit of raw material (a log) suffered. Although the process operators were experienced and good at the tasks they performed, the shear number of decisions made daily, tight dimensional market specifications required to meet lumber standards, the physically demanding and tedious nature of the work, and a large number of shifting process variables created a situation where many units of output were out of specification. The same problems also generated faulty decisions regarding the best way to get the most lumber out of a log, which in turn lowered the number of board feet of lumber produced relative to the board feet encompassed within each log. Not only was the board foot yield affected by sawyers' decisions, but the total dollar value per time of operation was also influenced negatively because the sawyer concentrated on dimensions that more easily derived the most volume of lumber the fastest, not the greatest amount of dollars.

The mill's efforts to control quality during the production process consisted of having personnel measure eight boards in eight places within the first 10 minutes of start up to see if production was within specifications. The focus was on whether the process was producing boards within process specification at start up, as opposed to whether the process was in or out of control in real time. Real time process adjustment decisions were seldom made based on the direct output from the QC process after start up. Process decisions made by operators were primarily a function of their tacit knowledge of the inherent variability embedded in the process system, and how that variability affected outcomes.

Required lumber dimensions were determined by industry standards (i.e., a dried two by four needed to be a minimum of 1.50" deep and 3.50" wide), and the mill experienced increased costs as a result of being either under or over these dimensions. The more costly problem was being under specification. Product that was under specifications had to be sold on a secondary market at a significant discount. Tony estimated that being under specification cost $20.00/1000 board feet, which could cost the company up to $8,000.00 per day in lost profit if the entire production for the day was under specification. Because of the high cost of being under specification, the process was routinely set to overcut .350" (Standard deviation (SD) .11) over market specification for width and .210" (SD .075) for depth. Of this overcut, however, .006 inches was necessary for shrinkage during the drying process to reduce the moisture content of the wood to 12-16% and another .015" was necessary to allow for variation in saw kerf (width of the cut made by the saw that is a representation of the output of the saw sharpening process). The remainder of the overcut was to protect against the possibility of producing output that was under specification due to other system variability. With the mean overcut listed above it was felt that the milling process would produce few boards under market specification due to the normal variability of the system. Tony was not sure what the long run average variability of the system was exactly, but knew that the theoretical minimum variability was the variability associated with variation in kerf.

Cutting product that was over specification did not impact the product in the marketplace as any boards that were produced over specification were corrected in the final planing process after the green lumber (the term green here is the term used in the industry to describe lumber that has not yet been dried) had been dried. Cutting the lumber over specifications did waste material, however. Tony estimated that every 1/1000 of an inch cut over market specification on both width and depth dimensions together cost the company $20,302.00 per year in wasted wood. The plant tended to operate with a bias toward avoiding undersized lumber, and oversize system drift was seldom caught or worried about and at times whole shifts would produce boards that exceeded the targeted overcut.

PLANNING FOR CHANGE

When Tony took over as manager the board feet of lumber produced as a percent of the number of true board feet of logs introduced into the production process was 85%. Even though the mill was making money, industry competition was eroding profits. Tony knew he had to create a production process that increased yield by producing more boards per log by reducing the overcut and optimizing the cut decisions based on both number of board feet and market value per board foot for a given log. To accomplish the yield and market value increase, the number of human decisions had to decrease, and an effective QC program had to be implemented. An effective QC program would give operators the necessary ability to manage board dimension in real time given the need for narrower process settings to produce boards closer to market specification.

Tony knew that changing the process would be challenging from a human resource perspective. The mill was located in a small town and was operated by people with little education beyond high school. Up to three generations of some families worked at the mill at any one time, and traditional milling practices were revered. Due to this reverence for traditional practices and the low level of education, workers were going to be particularly suspicious of the new technologies that Tony knew were necessary to keep the mill competitive.

When Tony became sawmill manager he held a meeting with the mill's employees to explain why the traditional milling process was no longer adequate and to explain that the process and the way operators managed the process would have to change if the mill was to remain competitive. Process improvement had to happen in all areas of the milling system or everyone would be out of a job. The targets Tony gave were to reduce the average board overcut by .06" for width and .054" for depth, increase the board feet of lumber recovered to true log board feet bought to 95%, and to increase the per thousand value a minimum of $30/1000 by cutting to market demand. Then Tony did something the employees were not used to--he asked them how they thought they should proceed and promised there would be no lay offs from the company due to process change. After numerous meetings between Tony and his employees and Tony and his managers, everyone agreed that the current processing system was no longer adequate. The process operators were expert, dedicated employees who were limited by the current process in how much product they could produce and the quality of that product. Thus, it was agreed that the mill employees needed to design and implement a completely new, automated process utilizing the latest technology if the Sandy mill was to stay in business for the long term.

NEW MILLING PROCESS

During 2002 and part of 2003, under Tony's leadership, Tony, his employees, and equipment suppliers/consultants installed an automated milling system that they had designed (Exhibit 2). The system consisted of two major components: (i) a full log scanner and the corresponding software and controls to allow computer optimized canter saw cuts; and (ii) new scanners for the board edger and trim saws and the corresponding software and controls to allow for computer optimized trimming of the individual boards produced. The new system also include a new method of saw maintenance that reduced the saw kerf from .135 inches to .115" and the SD of the kerf from .005" to .002". Exhibit 2 provides a process flow diagram of the new system.

This original installation cost $2.6 million and was completed within the time specified, but it was a stressful time for the organization. To help reduce stress and keep employee buy in, management provided extensive training to the process operators so that they could not only operate the new milling system, but also understand the technology that enabled the system. Further, as promised, the management at the Sandy mill did not lay off any employees due to the new system, although one employee was reassigned to another job within the company. Tony made sure he took time to ski, bike, and spend time with his family during this time as a way to cope with the stress, and he encouraged his employees to also maintain balance in their lives.

The new system improved log yield to 95% and helped the mill become more market driven as the decisions on how to cut the logs and boards was transferred from operators to sophisticated computer software. The decision rule the computer followed was based on the size and quality of a log, and the dimensions that were in demand and had the highest prices in the marketplace. While the mill still operated the same number of hours as before, production increased from 85,000 MMBF to 116,500 MMBF. Some of the increased production was due to an increased speed of production (as a result of faster decisions) and some due to less waste (as a result of better decisions). Further, completely automating the sawing task helped reduce setup times and enabled shorter batches by species and made it practical to have a greater mix of log sizes within a species batch and still realize the optimal number of board feet of lumber relative to the board feet in a log. The ability to run smaller batches and to optimize the amount of lumber based on the value of specific dimensions increased the average selling price per thousand by $25/MMBF.

One disappointment that the team experienced with the initial installation was the team's inability to meet its targeted reduction in overcut. The team was able to reduce the overcut by .03 inches (.320" width, .18" depth overcut, SD .099" width, .059" depth), but this was only about half of what was targeted. The primary limitation was that the initial install did not provide a feedback mechanism such that the production system could be monitored for board dimensions on a real time basis. The scanners used at the trim saw and the board edger could have been used for this purpose, but the software was not designed to do so and would have been expensive to convert. The trim scanner inspected a board to determine the optimal length or lengths to make that board determined by defects within the board and original length of the board. Standard lengths ranged from eight feet to 20 feet in two-foot increments. For example, a 19 foot, defect-free board could be made into an 18 foot board, an eight and ten foot board, or two eight foot boards. Market price would dictate the optimal decision. If there were a defective spot on the board, the algorithm used in the computer software would determine how to cut the spot out and still produce a valuable board or boards. Similarly, the scanner for the edger inspected the rough boards that had round edges to determine the optimal width of the board that would remove the round edges from the board. About 1/3 of production went through the edger and all production ultimately passed through the trimmer operation. Both of these devices essentially took a digitized picture of a board where 95% of the measurement error will range between +/- .008". The scanner software was not designed to capture and consolidate this data, however, so even though the original install did achieve less waste by making better decisions about the width and length of a board to make, it did not include an easy method of determining real time status of the production system.

NEW QUALITY CONTROL PROCESSES

To achieve real time system status so machine settings could be set tighter, Tony immediately had two additional scanners placed in the system shortly after the initial system install. These scanners just monitored board thickness and width, and were accurate to +/- .002" when the lenses were clean. The first scanner was placed in the process just after all the boards from both sets of vertical band saws were cut to the optimal width. The second scanner was placed in the system following the trim task where all boards going through the system were cut to optimal length/s (see Exhibit 2). Each scanner could scan 1000 measures per second on boards that were being conveyed at approximately 700 feet per minute thus, the QC system had the ability to measure all boards in both width and depth dimensions every 1/8 of and inch. This represented an incredible amount of data that Tony was still trying to figure out how best to use. Due to the nature of the current process, however, it was not possible to tell which board came from which band saw on the first scanner, or which board came from the band saws or the gang saws or which saw on the gang saws on the second scanner.

The data from each scanner was graphed in real time against targeted process specifications and was used in an Engineering Process Control (EPC) protocol. When 5 out of 10 boards, 30 out of 100 boards, or 50 out of 300 boards were above or below the process specifications, the computer alerted the operators. Once alerted, the operators determined and fixed the problem that was causing boards to be produced beyond process specification. Because operators had no way to tell how the out of specification board came through the system from the displayed data, the system could not be shut down until the operator determined the problem. There were at least 12 different places in the total system where adjustments could change, and out of specification boards would be produced as a consequence. At the production rate of one board per second, if all 12 areas were producing under size boards, 8333 board feet with a $20 penalty per MMBF could be made in the ten minutes it usually took the operators to find the problem.

In addition to the computer-generated alerts that the operators received from the system, the production supervisors had continuous access to the real time graphical process output that the system generated in the mill's control room. Tony had yet to establish a clear protocol on how supervisors should make use of this graphical output. Sometimes supervisors noticed system drift or patterns evident in the graphical output and took action prior to any alert being generated by the system. Further, sometimes supervisors took action to make a correction when the amount of overcut moved higher than the established overcut target. Other times, Tony was not even sure if supervisors looked at the output except at the end of the shift. Tony also recognized that the data being generated by the system provided the potential to evaluate the effectiveness of a given run or shift. Tony was unsure about the wisdom of using the data for such purposes and how best to understand the causes of the variations he saw in the data within a given shift and between different shifts.

The addition of the two scanners to the system allowed the overcut to be reduced to .290" (SD .09") for width and .156" (SD .05") for depth because of the reduced standard deviations of overcut. The mill continued to follow the EPC protocol described above, and this protocol resulted in the system being shut down at least four times on a typical day to adjust at least two of the 12 adjustable areas each shut down. Unfortunately, the new QC system did not provide a method of determining where the out of specification boards were processed in the system (such a system would cost approximately $10,000.00), nor was the software set to determine what part of the board was out of specification (such software would also cost approximately $10,000.00).

Knowing what route a specific board came through the production system would allow the system to be shut down as soon as a problem was detected because an operator would know which of the 12 areas was creating the problem. Knowing which part or parts of the board was/were out of specification would also provide clues as to what the problem at a specific area might be. Tony knew that the causes of some problems were so certain that automatic adjustments could probably be designed into the system such that the system could make the necessary corrections before the operator was even aware there was a problem. Tony knew of several locations in the process where such automated adjustments could be designed in and assumed that with the help of mill operators, all twelve areas of adjustment could be automated. Tony estimated that it would cost between $5,000 and $15,000 per location for automated adjustments. Tony expected that if these instruments were put in place in conjunction with knowing immediately where an out of specification board came from, the average overcut could be reduced another .06" (SD reduced further to .057") for width and .02" (SD reduced further to .026") for depth.

WHAT NEXT

Tony and the sawmill operators were pleased with many of the outcomes produced by the new process system, but knew they had to keep improving the process. Because operators could not react fast enough when the system started producing product out of specification, the amount of deliberate overcut was still large relative to the lower bound of milling system variation embedded in the flex of the saw blades. The QC system was producing massive amounts of data, but the operators did not seem able to utilize the information to produce boards closer to market specification, nor were the operators able to determine the source of specification problems from data alone. Tony knew that he had to revamp the QC system, preferably so that the milling process system would adjust automatically to information produced from the QC system. However, Tony was worried that upper management might balk at putting more money into the QC system and that the milling process operators might not understand a more sophisticated QC system and not be able to take advantage of all the possibilities for milling system improvement and control. Tony now had to decide whether or not to make further improvements in the QC system, keeping in mind the significant cost of such changes and the capabilities of his operators to utilize a more advanced system effectively.

[ILLUSTRATION OMITTED]

Task descriptions pre 2002.

Task 1, Debarker: The first step is to remove the bark from the log. The bark is used to generate steam for other milling processes.

Task 2, Mark II Canter Saw: The log is cut into a certain sized rectangle by removing slabs from the log that are a given thickness depending on the dimension of board required and the initial log dimensions. This task is critical in determining the board feet of lumber salvaged from a log. If the first cut is made in the wrong place, fewer usable boards or boards with lower market value will be produced. The decision on how to cut each log is made by the sawyer based on a visual inspection of the log and the help of a decision support system.

Task 3, Gang Saw: The rectangle produced by the Heading Sawyer is automatically cut into individual boards. The Gang Saw consists of up to 13 horizontal circular saws that have to be precisely set to produce boards of equal thickness.

Task 4, Hand Sort: boards are sorted by width, thickness, and approximate length. Boards that are the correct width and thickness are trimmed to an optimal length and stacked by length to go to the Kiln dryer where the lumber is dried to 12% moisture content from a normal green condition of 20 to 40% moisture content. Boards that are not the correct thickness or have too much wane are sent to Task 5.

Task 5, Slap Resaw: boards from the Canter Saw that are too thick are adjusted to the correct thickness and sent to Task 6.

Task 6, Board Edger: boards from the Slap Resaw are edged to the correct width and sent to the trim saw.

Task 7, Trim Saw: boards are cut to optimal length and sent to be stacked by hand for transport to the Kiln.

Task 8, Chipper: wood that cannot be made into lumber is ground into chips that will be burned to make steam or be used to make paper.

[ILLUSTRATION OMITTED]

Task descriptions post 2002 (with new technology/systems shown in bold).

Task 1, Debarker: The first step is to remove the bark from the log. The bark is used to generate steam for other milling processes.

Task 2, Log Scanner: scans are taken of the log using infrared scanners and the optimal number and thickness of slabs are determined automatically based on the shape of the log and the value of the lumber.

Task 3, Computer controlled Mark II Canter Saw: The log is cut into a certain sized rectangle by removing slabs from the log that are a given thickness depending on the dimension of board required and the initial log dimensions. This task is critical in determining the board feet of lumber salvaged from a log. If the first cut is made in the wrong place, fewer usable boards or boards with lower market value will be produced. This task is now controlled by a computer and optimal cuts are made to produce the greatest number of boards with the highest market value from each log.

Task 4, Gang Saw: The rectangle produced by the Canter Sawyer is automatically cut into individual boards. The Gang Saw consists of up to 13 horizontal circular says that have to be precisely set to produce boards of equal thickness.

Task 5, Scanner for Edger: scans using infrared scanners are taken of the slabs from the Canter Saw to determine automatically what board or boards can be made from the slab and then the boards are sent to Task 6 where they are edged to the proper width.

Task 6, QC Scanner 1: slabs are measured using infrared scanners to determine thickness to determine if the Heading Saw blades are set correctly.

Task 7, Board Edger: boards are edged to the correct thickness.

Task 8, Scanner for Trim Saw: boards are scanned to determine what optimal length of boards can be made, this information is used after Task 8 to determine which stack each board is sent to by length.

Task 9, Trim Saw: Boards are cut to optimal lengths and sent to be stacked for transport to the Kiln. A one to three board solution is possible.

Task 10, QC Scanner 2: all boards are measured using infrared scanners to determine width, thickness, and length.

Task 11, Sorting: boards are sorted automatically from information from Task 8 by length to go to the Kiln dryer where the lumber is dried to 12% moisture content from a normal green condition of 20 to 40% moisture content.

REFERENCES

US Census Bureau. (2002). Economic Census. Retrieved May 22, 2004, from www.census.gov/econ/census02/advance/TABLE2.HTM.

United Nations Economic Commission for Europe. (2003). Timber Bulletin ECE/TIM/BULL/56/2. Retrieved May 22, 2004, from http://www.unece.org/trade/timber/database/fps98_02.xls.

Kelly, Thomas, (2003) Lumber Statistics. United States Geological Survey. Retrieved May 22, 2004, from http://minerals.usgs.gov/minerals/pubs/of01-006/wood.xls.

Spelter, Henry & Alderman, Matthew (2003). Profile 2003: Softwood Sawmills in the United States and Canada. United States Department of Agriculture. Research Paper FPL-RP-608. Retrieved May 22, 2004, from http://www.fpl.fs.fed.us/documnts/fplrp/fplrp608.pdf.

Scott Metlen, University of Idaho

John J. Lawrence, University of Idaho