20th Century Management Science at PPG

Pittsburgh Plate Glass Company (renamed “PPG Industries” in 1968) was founded in 1883 as the first significant producer of flat plate glass. Over the first half of the 20th century, PPG’s product selection expanded to include paints, coatings, and industrial chemicals. Efficient management of a widespread production pipeline across multiple verticals became an obvious need. This necessity was met with the adoption and advent of novel management science techniques that soon after found applicability in other industries. Though the company did not play as a strong a role in directly influencing professional operations research and management science societies as other firms of the mid-century, their collaboration with academic practitioners of MS contributed to the field. Prior to the formation of an official OR/MS group at the company, PPG was an important player in the advancement of many MS techniques, namely in production management and inventory planning.

            PPG’s close relationship with the nearby Carnegie Institute of Technology was a differentiating factor in their adoption of advantageous management science techniques. In the late 1950s, the school received a “large grant from the U.S. government for research directed to developing computer-oriented mathematical-modeling approaches to management” (Cooper 2004). Herbert Simon, a researcher at the university’s Graduate School of Industrial Administration (GSIA) and future Nobel laureate, played a key role in the project. With funding from the Office of Naval Research (ONR), Simon led research into the scheduling of paint production at PPG’s plant in Springdale, PA (Augier and March 2004). In addition to PPG, the researchers used data from the Westinghouse Electric Corporation, Wearever Inc., and a few other companies in the region.

This study led to the development of mathematical and statistical tools that could improve inventory-control systems and production planning. Simon and others devised a decision-making process within the context of aggregates, such as total labor force and total value of monthly production and inventories (as in PPG’s paints). He and his colleagues looked at five years’ data of bi-monthly sales of paint – one package-unit of one color. Using historical data, they were able to successfully forecast the actual sales, one month in advance. Carnegie researcher Charles Holt, and his student, Peter Winters incorporated seasonal trends using exponentially weighted moving averages (Holt et. al 1960). Exponential smoothing for demand forecasting was earlier suggested by Brown(1956). Though the analysis was implemented at PPG, subsequent researcher visits to the firm indicated that certain elements were not initially followed, namely because the company failed to fire workers when the model indicated they should be fired (Nahmias and Olsen 2015).

Nevertheless, these methods became more widely used following the 1960 publication of Planning Production, Inventories, and Work Force by Simon and his GSIA colleagues, Holt, Franco Modigliani, and John Muth (“HMMS”). Their techniques, such as the Linear Decision Rule (LDR), were among the many in the mid-20th century that migrated MS tools out of academic spaces into the realm of decision makers in business and industry. By the late 1970s, however, no company reported to be using the LDR (Schwartz & Johnson 1978).  This is not to say PPG paint model wasn’t a forefather to the multi-period planning used in present day. In fact, it served as the launching point for new approaches to the problem. Papers on aggregate planning problems first appeared in the mid-1950s and the HMMS text marked a significant concentration of these ideas. Linear programming formulations of production planning problems, like the one at PPG, were developed by Fred Hanssmann and future TIMS President, Sidney W. Hess. The Holt/Winters exponential smoothing method became a widely used element in forecasting statistics.  Holt’s initial work in exponential smoothing, published by the ONR, may have been a foundational piece for later more general forecasting models, such as ARIMA, by Box and Jenkins ( 1970).

In addition to working with Carnegie, PPG had a long-standing affiliation with the Mellon Institute of Industrial Research. For example, the company sponsored multiple faculty fellowships at Mellon in 1938 (Mellon Institute 1938). When the Mellon Institute and Carnegie Institute merged into Carnegie-Mellon University in 1968, PPG continued their respective relationships with persons of the unified institution. In the 1960s and 1970s, PPG continued to use leading inventory and production models of the era.  One such example was SIMPLAN, a planning and budgeting language “to facilitate the development and programming of corporate planning models” developed by Social Systems Inc. (SSI 1975).

The continued growth and use of novel techniques led to the official formation of a PPG Management Sciences Group in the mid-1970s, following an internal primer on OR/MS methods written in 1973. The initial­ vision of the group over the next twenty years continued within the corporate IT organization of PPG, “to help PPG personnel make better, faster decisions that translate to bottom-line results.” (Kandt 2006). Dave Garson was the group’s leader in the early 1990s and initiated a production scheduling software system that was developed and maintained over the following decade and a half. He also oversaw a strong working relationship with Systems Modeling Corporation as his team used ARENA as their primary simulation platform. Following Garson’s move to a more senior position within the IT department at PPG, Kevin Kandt became the team’s manager in 1996. At any given time during Kandt’s tenure (1996 through 2011), the group comprised six to eight members, each responsible for a variety of MS methods including: Terry Parrinell (computer simulation), David Revistky (production schedule and systems analysis), Margaret Foote (advanced optimization and decision support technologies), and Lisa Vincenty (optimization and process improvement in supply chain management).

At the turn of the millennium, the Management Sciences Team was broadly responsible for not just ORMS functions, but also some general industrial engineering functions within the various business units in the PPG corporate structure.  Although the group’s “management science” title was considered a bit dated, they kept the name in honor of their 1970s legacy. Key activities included:

  •  Development of a Microsoft Windows-based production scheduling software package, including full responsibility for system governance and maintenance, that was used in 30 – 35 of the PPG production facilities.
  • Computer simulation to support manufacturing process improvements.
  • Development and maintenance of a real-time decision support system based on neural net technology for allocating the glass ribbon produced in PPG’s float glass plants.
  • Supply chain network optimization modeling.
  • Monte Carlo simulation for decision support.
  • Business process improvement/engineering.
  • Lean Six Sigma, in cooperation with PPG’s Quality Assurance team.
  • ERP/MRP system specification and selection, as the various PPG business units began looking into Oracle and SAP systems.


Compiled by: Reed E. Devany


Author’s Note: The author would like to give special thanks to Kevin Kandt for his contributions to this piece.

Links and References

Augier, M. & J. G. March (2004) Herbert A. Simon, Scientist in Models of Man: Essays in Memory of Herbert A. Simon, Augier, M. & J. G. March, eds. MIT Press: Boston, MA. 3-32

Box, G. and G. Jenkins (1970) Time series analysis: Forecasting and control, San Francisco: Holden-Day.

Brown, Robert G. (1956). Exponential Smoothing for Predicting Demand. Cambridge, Massachusetts: Arthur D. Little Inc. p. 15.

Cooper, W. W. (2004) Memorial to Herbert A. Simon in Models of Man: Essays in Memory of Herbert A. Simon, Augier, M. & J. G. March, eds. MIT Press: Boston, MA. 67-74.

Egidi, M. & L. Marengo (2004) Near-Decomposability, Organization, and Evolution: Some Notes on Herbert Simon’s Contribution. Models of Man: Essays in Memory of Herbert A. Simon, Augier, M. & J. G. March, eds. MIT Press: Boston, MA. 335-350.

Hanssmann, F. & S.W. Hess (1960) A Linear Programming Approach to Production and Employment Scheduling. Management Science 1(1): 46-51.

Holt, C.C., Modigliani, F., & H. A. Simon (1955) A linear decision rule for production and employment scheduling. Management Science, 2(1): 1-30.

Holt, C. C. (1957) Forecasting trends and seasonal by exponentially weighted averages. Office of Naval Research Memorandum 52. ONR: Washington, DC.

Holt, C.C., Modigliani, F., Muth, J. F., & H. A. Simon (1960) Planning Production, Inventories, and Work Force. Prentice-Hall: Englewood Cliffs, NJ.

Kandt, K. F. (2006) Management Science Group Thrives at PPG Industries. ORMS Today, 2006 (4).

Mellon Institute (1938) [announcement] Multiple Fellowship of the Pittsburgh Plate Glass Company at the Mellon Institute. Science, 84(2168): 56.

Nahmias S. and T. L. Olsen (2015) “Chapter 3 – Sales and Operations Planning” in Production and Operations Analysis: Seventh Addition. Waveland Press: Long Grove, IL. 128-168.

Schwartz, L. B. & R. E. Johnson (1978) An appraisal of the empirical performance of the linear decision rule for aggregate planning. Management Science, 24(8): 844-849.

SSI (1975) [advertisement] Operations Research, 21(9): 1090.

Winters, P. R. (1960) Forecasting sales by exponentially weighted moving averages. Management Science (6): 324-342.