Operations Research and Management Science at P&G in the 20th Century
The Procter & Gamble Company (P&G) was founded in 1837 and grew into a multinational leader in consumer goods by the 20th century. Following WWII, the Cincinnati-based company saw rapid expansion into international markets and a growing portfolio of new products, such as Tide laundry detergent, Crest toothpaste, and Charmin toilet paper. With the addition of these brands (and a rapid increase in consumer demand) efficient production and inventory planning was necessary. P&G had to implement novel solutions in operations research and management science, helping, in turn, advance the fields’ growth in private industry.
The management science group at P&G evolved from a single division in 1947 into a “matrix” organization by 1972, comprising multiple divisions for each operating function of the company. The Industrial Engineering Division was established to handle cost control, work measurement, and methods improvement for the company’s manufacturing section. Initial quantitative analysis from the group was done under such names as “management control systems” and “operations research services”. One such effort that began as a study on the static aspects of warehouse logistics (e.g. optimal pallet sizes and patterns) grew to include more complex storage usage. The division had a driving slogan that was continually asked by everyone in manufacturing and logistics: “if it were not for what basic cause?”. Inventory control scheduling (ICS) at P&G came from a 1957 study on production scheduling and inventory control. ICS was completely innovative at the time and implemented across the entire company. The system attempted to keep production stable so long as inventory levels were within control limits. Work on production scheduling and inventory control led the OR/MS group to P&G marketing. This was a unique and pioneering relationship, resulting in several members of the management science team moving to corporate headquarters in downtown Cincinnati. Ed Ignall, who was involved in the development and implementation of ICS for the detergent line of P&G’s business, was one of the practitioners who made that significant move from production to marketing. The Data Processing Systems Division was created in 1954 “to introduce electronic digital computing, but the orientation was primarily toward the commercial areas of business.” The evolution toward management information systems and the overlapping function of these two divisions led to the creation of a joint Management Systems Division in 1971 (Montillon 1972).
As the initial OR/MS projects of the Industrial Engineering and Data Processing Systems Divisions were underway, Proctor & Gamble economists were making headway in the realm of business forecasting. Wilson Wright, whose work was featured in Business Forecasting in Practice: Principles and Case, edited by Abramson and Mack, was an important figure. He wrote an analysis of American business outlook in the Post War period, (1945-1949), noting the “high and unstable” level of industrial production and price increases that resulted from wartime inflation. Considering these factors, in addition to the idea that federal expenditure may be “determined by international political developments beyond the control of [the US]”, Wright proposed a boom and recession cycle occurring in three to four year fluctuations and that this forecast could be used for determining inventory policy and production schedules (Wright 1956). This prediction largely held true, though not at as severely as some forecasters believed. There was a brief recession from November 1948 to October 1949 and a short but substantial economic drop following the Korean War (July 1953-May 1954). Wright’s three to four-year cycles held through the Recession of 1958 but ended in the early 1960s, which brought a near decade of sustained economic growth.
Wright played a role in encouraging the federal adoption of OR/MS practices and participated in hearings before the Subcommittee on Fiscal Policy of the Joint Economic Committee of the United States Congress. He was the only expert from industry, joining economists from the University of Chicago, the University of Minnesota, Penn, and RAND. In his testimony, Wright called for the establishment of task forces capable of managing government expenditure like a business, believing “groups commissioned for the performance of this task can be used by Congress in the way that managers or directors of a corporation sometimes employ the professional service of firms specializing in operations research and management engineering” (Wright 1957).
P&G staff were among the earlier members of the Operations Research Society of America (ORSA). George D. Montillon, the head of the company’s OR/MS group, was the first P&G employee listed as a member of the society in 1955 (ORSA 1955), and was shortly after followed by Donald W. Grace and Norman R. Smith in 1957 (ORSA 1957). Prior to joining P&G, Grace was a combat engineer responsible for managing the supply side of Army operations during World War II. At P&G, he set up one of the first mainframe computers and taught at nearby Xavier University. The company eventually sent him to Stanford, where he earned graduate degrees in computational analysis prior to returning to Cincinnati (“Obituary” 2019). Grace was not the only P&G staffer affiliated with Xavier. Keppel “Duane” Wait, a research executive with the company, also taught part time at the university (“Obituary” 2010). Just as Montillon, Grace, and Smith were members of ORSA, Wait was a member of The Institute of Management Science (TIMS) and played an active role in that society’s Miami Valley Chapter. In 1973, he led a joint meeting on technical computing with the American Institute of Industrial Engineers (Juggenheimer 1973). P&G’s John Bieda led a Miami Valley Chapter meeting on multi-dimension scaling applications a year prior (Juggenheimer 1972). P&G involvement in TIMS was larger than the participation of a few employees. In the early 1970s, the company was part of TIMS’ Institutional Membership program. From 1972 through 1976, this partners program included companies such as General Electric, IBM, Ernst & Ernst, and Mobil Oil.
Among many other OR/MS practice areas, Procter and Gamble was a leader in statistical applications. In 1963, P&G’s Satya D. Dubey wrote a letter to the editor of Operations Research, concerning a theorem on a ratio of random, stochastically independent variables, X/Y and Y. He believed “in many statistical applications, a ratio of two random variables is relevant. It is not a very uncommon practice among many applied statisticians to consider that the expectation of the ratio of two random variables is the ratio of their expectations – a result of rare validity” (Dubey 1965). His short paper provided important details to this useful connection. Before moving to the Ford Motor Company, Dubey also authored a technical report on statistical inferences for laws concerning the Weibull distribution and presented his findings at the annual meeting of the American Statistical Association (ASA) in 1962 (Dubey 1966). In 1971, Bieda presented “A multi-brand, multi-segment model of consumer behavior: on estimating the parameter” at the Social Statistics Section of the ASA. In that paper, he reported on the findings from “an empirical study of the Minimum Chi Square (sic) estimation procedure when this estimation procedure is used for estimating the parameters of a heterogeneous linear learning model in the context of consumer purchasing behavior.” Bieda created models that looked at a one segment market, a two segment market, and a three segment market. Though he and his team were able to approach the global minima, some cases were far enough away from the known minima to cast doubt on the search routine (Bieda 1971). Furthermore, a number of OR/MS award recipients started their careers at P&G. Lanchester Prize recipient Peter Kolesar began his career as a member of the P&G team that worked on “Inventory Controlled Production Scheduling System”. And INFORMS Fellow Richard E. Nance was a member of the company’s management staff in the two years between his undergraduate and graduate degrees. Kolesar credits his time at the company as having a profound impact on him, inspiring his decision to go back to graduate school.
In 1962, P&G ran a traveling salesman problem contest asking for the shortest tour through thirty-three cities across twenty-two American states, beginning and ending in Chicago. Fifty-four $1000 prizes and one $10,000 grand prize were offered. Robert Karg and Gerald Thompson of the Carnegie Institute of Technology found the optimal route using a Bendix G-20 electronic computer. Their results were published in Management Science two years later and have been cited in over three hundred fifty subsequent papers (Karg and Thompson 1964). Additional research areas at P&G included feed-back based inventory, econometrics, and accelerated computing in Markov chains.
P&G was ahead of the curve when it came recruiting talented OR/MS practitioners from academia to industry. In addition to those mentioned previously, some team members returned to universities and led meaningful careers as research professors. This included Richard L. Francis (Professor in the Industrial and Systems Engineering Department at the University of Florida), Edward J. Ignall (Engineering and Applied Sciences Professor at Columbia University), George P. H. Styan (mathematical statistician at McGill University), Jeremy Schapiro (operations research group at MIT), and John H. Ramberg (statistician at University of Arizona). In the publishing space, Harry Smith, Jr. (Dubey’s supervisor at P&G) co-authored the thorough textbook on regression, Applied Regression Analysis (1966), which contained many distinguished examples from P&G.
In the 1970s, P&G began a close working relationship with the University of Cincinnati. Dennis Sweeney, a professor at the university’s Operations, Business, Analytics, and Information Systems (OBAIS) Department, decided to “get out of the ivory tower” and spend his sabbatical in industry. He began working with P&G’s Mike Doherty on developing new solution procedures for mixed integer programming. They applied and tested their solutions on real P&G problems. This resulted in the publication of multiple research papers, leading Doherty to serve as an adjunct faculty member at UC for nine years, teaching evening courses (INFORMS 2019). The relationship was rekindled in the early 1990s when a joint research project between the two organizations took on redesigning the company’s North American supply chain (Camm et. al 1997). This project help launched P&G’s Center of Excellence in Supply Chain Analytics and was named a finalist of the 1995 Franz Edelman Award. When the UC Center for Business Analytics was established in 2012, P&G was its first corporate member.
Procter & Gamble researchers played a role in the early development of artificial intelligence in advisory systems. In 1980, Franz Dill was asked by P&G chairman Owen “Brad” Butler to work with data pertinent to sales district costs, financial forecasts, and regional profitability. Dill and Butler collaboratively coded the first ever executive information system. (Dill 2008) P&G’s C-Suite Advisory Development effort from 1989-1991 included three systems: Automated CEO, New Initiative Advisor, and Major Capital Appropriation Screener. P&G also endeavored to build a decision support system that linked sales information and retail scanner data. (Power 2003). In the 1990s, the company used Prolog style systems, like M1, and developed systems such as the Coffee Blending advisory system, saving millions of dollars. Other application areas included a neural net-based induction system for advertising campaigns and a system that “learned” supply chain solutions from traffic and inventory data.
In the closing years of the twentieth century and into the 2000s, P&G has remained an active member of the OR/MS community and has been recognized for its achievements. The company was a major sponsor ($100,000) of the INFORMS meeting held in Cincinnati in the spring of 1999 and received the 2004 INFORMS Prize “for its integration of OR/MS principles into the decision making structure of the firm, to tackle issues such as global manufacturing site location and sourcing, inventory management, supply chain design, and consumer behavior analysis.” P&G’s analytics team was also among six finalists for the 2010 Franz Edelman Prize for their work on inventory optimization through user adoption of inventory tools.
Compiled by: Reed Devany
Author’s Note: The author would like to give special thanks to Pete Kolesar and Jeff Camm for their contributions to this piece.
Links and References
Bieda, J. C. (1971) A Multi-brand Multi-segment Model of Consumer Behavior: on Estimating the Parameters. in Proceedings of the Social Statistics Section of the American Statistical Association 1971, Goldfield, E. D., ed. American Statistical Association: Washington, D,C. 226-230.
Butler, O. B. (1976) What Marketing Expects from R&D. Research Management, 19(1): 7-9.
Camm, J. D., Chorman, T. E., Dill, F. A., Evans, J. R., Sweeney, D. S., & G.W. Wegryn (1997) Blending OR/MS, Judgment, and GIS: Restructuring P&G’s Supply Chain. Interfaces, 27(1): 128-142.
Dill, Franz (2008, July 28) Early Business Intelligence Needs [Blog post]. http://eponymouspickle.blogspot.com/search?q=brad+butler
Draper, N. R. & Smith, H. (1966) Applied Regression Analysis. Wiley: New York.
Dubey, S. (1965) A Theorem on a Ratio of Random Variables. Operations Research, 13(3): 476-477.
Dubey, S. (1966) On some Statistical Inferences for Weibull laws. Naval Research Logistics Quarterly, 13(3): 227-251.
Farasyn, I., Humair, S., Kahn, J., Neale, J., Rosen, O., Ruark, J., . Willems, S. (2011) Inventory Optimization at Procter & Gamble: Achieving Real Benefits Through User Adoption of Inventory Tools. Interfaces, 41(1): 66-78.
INFORMS (2004) Award Recipients: Procter & Gamble. INFORMS Recognizing Excellence. Web Article. Accessed April 9, 2020. https://www.informs.org/Recognizing-Excellence/Award-Recipients/Procter-Gamble
INFORMS (2019, May, 30) 2019 UPS George D. Smith Prize Winner, University of Cincinnati. [Video file] retrieved from https://www.youtube.com/watch?v=hcQbYc9LUIk.
Juggenheimer, R. W. (1972) Chapter and College News. Interfaces, 2(3): 70-82.
Juggenheimer, R. W. (1973) Chapter and College News. Interfaces, 3(3): 87-101.
Karg, R. L. & G. L. Thompson (1964) A Heuristic Approach to Solving Travelling Salesman Problems. Management Science¸10(2): 223-248.
Smith, A. (2010) [Press Release] Procter & Gamble Among Finalists for INFORMS Edelman Award. INFORMS News Room. Web Article. Published March 26, 2010. Accessed April 9, 2020. https://www.informs.org/About-INFORMS/News-Room/Press-Releases/Procter-Gamble-among-finalists-for-INFORMS-Edelman-Award
Martin, W. S. (1966) Long-Range Planning for a Consumer Industry. Research Management 9(4): 221-228.
Montillon, G. D. (1972) Implementation of Management Science at Procter & Gamble. in Academy of Management Annual Meeting Proceedings, 1972(1):123-125.
“Obituary,” Keppel Wait 1938-1010 (2010, June 19) Retrieved from https://www.legacy.com/obituaries/greenvilleonline/obituary.aspx?n=keppel-wait&pid=143625477. Accessed 4/7/2020.
“Obituary,” Donald Wayne Grace (2019, July 29) Retrieved from https://www.clementsfuneralservice.com/obituaries/Donald-Wayne-Grace?obId=5524447. Accessed 4/7/2020.
ORSA (1955) Members of the Society. Journal of the Operations Research Society of America, 3(4): 11-42.
ORSA (1957) Members of the Society. Journal of the Operations Research Society of America¸ 5(6): 8-66.
Porteus, E. L., & J. C. Totten (1978) Accelerated Computation of the Expected Discounted Return in a Markov Chain. Operations Research, 26(2): 350-358.
Power, D.J (2003) A Brief History of Decision Support Systems. DSSResouces.COM. Web Article. Accessed May 26, 2020. http://dssresources.com/history/dsshistoryv28.html
Wright, W. (1956) An Analysis of the Business Outlook. In Business Forecasting in Practice: Principles and Case by Abramson, A. G., & R. H. Mack, eds. John Wiley & Sons: New York. 255-269.
Wright, W. (1957) Economy and Efficiency in Government Expenditures. Federal Expenditure Policy for Economic Growth and Stability: Papers Submitted by Panelists Appearing Before the Subcommittee on Fiscal Policy, Joint Economic Committee, November 5, 1957. United States Government Printing Office: Washington. 258-263.
Associated Historic Individuals
Kolesar, Peter J.Nance, Richard E.