Optimized Wood
Skip Navigation Links  Optimization Planning of Veneer, Plywood, Engineered Wood at Boise Cascade by Joseph Lilly

It was in the mid 1990's, while in route to visit a mill in Medford, Oregon that Chuck, the regional manager and Tom the corporate planning manager first discussed the need for a better planning system.  The operation was getting too complicated.  Chuck was a veteran in the timber and wood industry and at Boise Cascade, he was regarded as the guru of plywood and veneer.  In the Pacific Northwest, the company was a leading player in the industry and owned and controlled millions of timberland acres.  Chuck managed the supply chain at the company's lathe and plywood mills in Oregon, and engaged in buying and selling veneers, the thin sheets of wood used in making plywood.  Chuck probably understood the economics of the market as well as anyone and yet he conceded that he was becoming overwhelmed by the complexity.

The company had recently entered into the Engineered Lumber Products (ELP) business and had just began operating their new, showcase mill in White City, Oregon.  This mill produced high valued laminated beams and I-joists.  These ELP products allow building designs of much longer floor and ceiling spans which were becoming favored in the home building market.  Yet in manufacturing these engineered products, the strength specifications require the best strength-graded veneers and that the laminations be thinner than a comparable sheet of plywood. 

Now maintaining a steady supply of one-tenth inch veneer was causing Chuck a serious problem.  He was generating lots of one-tenth inch veneer that didn't make the strength-grade, and for which he had no real plywood or market demand. 

Chuck was aware of a planning method that he thought could help resolve the situation.  Earlier in his career, he’d been involved in a linear programming project.  Although not a big technology guy himself, he did have an awareness and appreciation for the use of LP as an effective management tool to optimize resources.  So in his conversation with Tom, Chuck declared that he needed an LP.

It was at about that time that I'd given a talk at a Western Wood Products Association trade show in Spokane.  I'd met some folks from Boise Cascade and was invited to Boise to give a presentation to the corporate planners in Tom's group and to a group of mill managers.  As an optimization consultant, I was giving a sales pitch on applying optimization in the wood industry.  I'd prepared a small demonstration from software I'd developed and with it, we interactively solved a simple lumber model, made lots of "what-if" changes and resolved to see the impact of changes in price, grade yields etc.

Optimization, (or linear programming), is simply an application of algebra.  It is a way to maximize or minimize some objective such as profit or cost.  The details of an operation can easily be described with simple algebraic equations (or inequalities).  Optimization uses a special algorithm to solve these equations and reveal the combination of decisions that are optimum.  The method considers all the tradeoffs, and determines the best program with respect to the entire operation.

Tom, coincidentally had just had this conversation with Chuck, and so indicated that the company needed a plywood model for Medford.  It was several months later that I was contracted to do this project.  I was introduced to Greg, who had been assigned as the project manager.  We initially met with the folks at Medford in an open meeting with all the key mill managers, superintendents, supervisors and foremen to establish the project goals and objectives and to set out a plan, including responsibilities and timelines.  After working with this group, I realized that I was lucky first, to have such widespread interest and involvement and second, to be among such an experienced staff.  As the project was taking shape, I convinced Boise Cascade to develop a modeling system, not just a model.  Although I generally promote this idea, it was apparent that to treat this as a one-time “study”, in which the consultant concludes the project with a recommendation, would be a mistake.  Boise Cascade was faced with issues that were perpetually changing; market prices and availability of materials, timber supplies and characteristics, and the nature of the market for finished building products.  Boise Cascade needed a modeling system, something that could be reused, and was easy and flexible for designing new modeling scenarios.

At that time, I’d been a consultant for a few years and had developed such software applications.  Yet, these were designed to support a fixed model.  I was searching for something more flexible and robust.  Now I had also been teaching some courses in Production and Operations Management at Cal State Stanislaus.  In particular, I’d covered linear programming in a Quantitative Analysis course.  In my lectures, I’d routinely illustrate an LP model on the chalkboard by drawing something of a flowchart.  It would usually look like boxes and connecting lines with arrows that simulated the flow.  So these events lead to my idea of developing a software application for drawing a model graphically, and solving it with optimization technology. 

 During the next several months, the project progressed on two tracks: a software development track, and a model development track.  I was designing software with a particular model, but with the intention of applying it to many different models.  I adapted a graphical user interface in Visual Basic from some of my previous work, designed an Access database, and interfaced this with XA, the optimization solver by Sunset Software.  The GUI allowed a user to create a model, or a description of the production configuration.  Users were able to place objects on the screen to graphically represent a facility, or machine or purchase of material, or transportation routing.  Each object could be defined with various properties, such as the database address for data details.  The modeling GUI allowed for an independent model of a single facility, or a whole configuration of multiple facilities.  The software interpreted the user’s drawing configuration, retrieved the corresponding data, programmatically formulated an optimization model, solved and returned the solution.

In developing the system, we needed a database primarily to maintain yield and throughput data for each material for each processing point.  I realized that the number of data records might be quite large so the database was designed for scalability.  I was fortunate to be working with Greg, who became my indispensable partner in this project.  He setup and conducted many yield tests to gather data that we needed for our initial model and was my primary software user and tester.  I was also fortunate to be working with Chuck who really, taught Greg and I the veneer and plywood business.  I was constantly questioning just how much detail was important, why processing was done in a certain way, etc.   

Over the course of the next several years, Greg and I moved throughout Boise Cascade’s Timber and Wood operations, and working with the staff at each location, setup initial models and deployed optimization software.  At this stage, the work turned primarily to data gathering and modeling.  In some cases, I extended the software to accommodate unique circumstances or to handle custom modeling applications.  For example, we studied and modeled the lathe yield at a green-end mill so that with a few log/block parameters such as species, diameter, position of the heartwood and sapwood rings we could fairly accurately predict and generate the yield data.  We did some special modeling and analysis of breakeven price for open market timber purchasing and sales.  We modeled the manufacturing throughout the vertically integrated supply chain including the green-end, plywood and engineered wood facilities and solved for the optimal regional configuration.  We analyzed alternative peel thickness and alternative “constructions” or bill of materials of plywood and engineered products.  This especially served to resolve Chuck’s initial material imbalance.

In one particularly interesting project, I met with George, Boise Cascade’s CEO, in a daylong planning session.  At that time, the company had several Vice-Presidents, Regional Managers and many Mill Managers in its Timber and Wood Division.  The compensation of these executives and managers was partly based on performance and in particular on the measure known as “unit-manufacturing-cost”.  I challenged this practice and with the use of the Optimization modeling system, did demonstrate that the UMC objective induced decisions that kept the company in the low end of the market.

Then in 1998, a devastating fire burned the Medford plywood mill to the ground.  460 employees were instantly out of work.  At that time the plywood market was not very good and so the company was at first uncertain about whether to rebuild.  In the aftermath, Greg along with Joe, an analyst in Medford, began a project to explore the possibilities.  The optimization modeling application I’d developed turned out to be the perfect tool for simulating and evaluating the various alternative mill configurations.  In the Pacific Northwest, Boise Cascade had an existing array of mills but with a major piece now missing.  The question was what if anything, should be built, and what capabilities should it have?  Although I supported the effort, it was Greg and Joe my clients, who conducted the analysis for this major capital decision, and they did this with software I developed.  In the end Boise Cascade decided to rebuild a $75 million facility with similar but different capabilities and which restored jobs for about 300 employees.

Joseph Lilly is the President of J.M. Lilly, Inc. and has been a consultant for over years. His business to apply quantitative methods to solve industrial problems and in particular, to help clients optimize their resources.
Copyright ©2009 JM Lilly  All rights reserved worldwide.