Case: Procter & Gambles Supply Chain RedesignSource Blending ORIMS, Judgment, and GIS: Restructuring P&G's Supply Chain, by J.D. Camm ,T. Chorman, F. Dill, J. Evans, D. Sweeney, G. Wegryn, Interfaces 27: 1 January-February 1997 (pp. 128-142)

Snapshot of P&G,1990sWorldwide market leader in laundry detergents, diapers, feminine protection pads, shampoos, facial moisturizers, acne teen skin care products, and fabric softeners300 brands of consumer goods Sales in140 countries Operating units (plants, divisions, facilities) in 58 countries P&G had worldwide sales of $33.5 B in fiscal 1995 and earnings of $2.64 B.

Strengthening Global Effectiveness InitiativePurposesstreamline work processes drive out non-value-added costs eliminate duplicationrationalize manufacturing and distributionScope at outsethundreds of suppliers over 50 product lines 60 plants 15 distribution centers hundreds of customer zones representing thousands of corporate customers

North American Product Supply StudyKey part of SGE InitiativeSought reengineering ofProduct sourcing Distribution network Principal toolsBusiness AnalyticsIT network optimization models geographical information system (GIS)

Five Factors Motivating Supply Chain RedesignDeregulation of the trucking industry had lowered transportation costsProduct compaction (detergents in concentrated form, compact packaging of diapers) more product per truckloadP&G's focus on TQM improved reliability and increased throughput at every plantdecrease in product life cycles from 3-5 yr to 18-24 mo required plants to update equipment more frequentlyCorporate acquisitions gave P&G excess capacity

Scope of North American Product Supply StudyFive Factors motivated update of product sourcing decisions, i.e. choosing best location for production, and scale of operations for each productConstraints and trade-offsscope of production at any particular site is limited to products that rely on similar technologies producing too many products at a site increases the complexity of operationslarge, single-product plants exposes a firm to riskPlant locations affect the costs of supplying raw materials and distributing finished products design of the distribution system must be considered

Planning and Organization30 major multifunctional product-strategy teams for developing product sourcing options aligned with P&G's major business categories: detergents, diapers, etc.composed of individuals from various functional areas: finance, manufacturing, distribution, purchasing, R&D, plant operations organized around product category groups that shared similar technology and could therefore be produced at the same manufacturing site One separate distribution and customer service team charged with developing options for DC locations, assigning customers to DCs, and making transportation decisions

3-*

SourcingDistribution

Business Analytics to the RescueBusiness Analytics can identify a small set of the most promising alternative designs out of an astronomical number of possibilitiesEnables product-strategy teams to collect and analyze appropriate data in order to generate detailed risk-adjusted cash flows for a reasonable number of scenariosMore important reason for sound analysis: potential impact of the project on peopleP&G Analytics group partnered with University of Cincinnati's Center for Productivity Improvement

Objectives of Business Analytics TeamSourcing: to provide mathematical models and decision support for the product-strategy teamsDistribution: to provide support to a team of experts in transportation and distribution who were concentrating on warehousing, distribution, and customer allocation problemsPutting the pieces together: to ensure that the composition of a complete-supply-chain solution across product-strategy and distribution teams was the best possible

Decision Support P&Gs legacy system: mainframe-based comprehensive logistics optimization model to support sourcing decisions for multiple product categories and multiple echelons, requiring long turnaround times for each model runNew Target: simple interactive PC-based tool that would allow product-strategy teams to quickly evaluate options (choices of plant locations and capacities), make revisions, evaluate the new options, employ a GIS, and guide users to better options in an evolutionary fashion

Modeling StrategyTo decompose the overall supply-chain problem into two easily solved subproblems: a distribution-location problem a product-sourcing problem for each product category ReasoningManagement's organization of the strategic-planning process into a distribution team and product-category teams implied a natural decomposition across echelons of the supply chain and across product categories. Business Analytics team determined that manufacturing and raw-material costs dominated distribution costs by a very large margin, suggesting that product-sourcing decisions were not highly sensitive to the downstream distribution-system design. Direct plant-to-customer shipments accounted for the large majority of plant shipments, suggesting that sourcing decisions were more sensitive to customer locations than to DC locations.

Modeling AssumptionsFor each customer zone, the proportion of demand satisfied by direct shipments as well as the proportion satisfied by shipments through DC is a constant for each product categoryDC locations could be chosen independently of the plant locations, due torelatively small volume (10 to 20 percent) shipped through DCssmall number of DCs (five to eight) needed to support that volume fact that manufacturing costs dwarfed distribution costs

DC - Customer Optimization ModelAggregation of trade-customer demand into 150 customer zones, which provided sufficient granularityMajor considerations on the choice of DC locations customer location customer services sole sourcingproximity to customer zones to maintain current levels of customer serviceEmployed uncapacitated facility-location model to find optimal DC locations and to assign customers to DCs. For a fixed number of DCs, the model finds optimal locations, while ensuring that each customer zone is assigned to a single DC. The objective is to minimize the cost of all DC-customer zone assignments.

*DC - Customer Optimization ModelMin CijXiji Xij =1, j Ji Yi = k Xij Yi, i I, j JXij = 1 if customer j assigned to DC iYi = 1 if DC i is chosenMixed Integer Linear Program

Product Sourcing ModelTransportation model for each product category: product-strategy teams specify the plant location and capacity options to be evaluatedArc costs are the sum of manufacturing, warehousing at the plant, and transportation costs. Manufacturing costs were the most important consideration in the product-sourcing decision, so team made careful estimates. DC costs were composed of actual per-unit storage and handling costs at each of the company's existing DCs, and appropriate estimates at new DC locationsTransportation costs estimates were based onnegotiated rates P&G was already paying for shipments between locations, along with rate tables; orFixed cost per linear function of distance between 2 points

*Product Sourcing ModelMin CijXijj Xij = ai, i I i Xij =dj, j JXij 0, i I, j JXij = shipment from plant i to DC jLinear Program

*Integration with GISUsed to display results of optimizations and manipulate data through a menu-driven system for sensitivity analysis and further model runs and evaluationAdvantageseasy to understand - visualization provides insightquery details by point-and-clickfacilitates acceptance of analytical techniqueshighlighted database errors

*

*Solution Composition and VerificationDC Location and Customer AssignmentOption GenerationOption SelectionVerificationDistribution Team: Facility Location Model30 Sourcing Teams: Product Sourcing Model and@RISK NPV ModelSteering Team: Logistics Modeling System

Project Results & BenefitsIntegrated solution called for plant consolidations: by mid-1996, P&G had closed 12 sites and written off over a billion dollars worth of assets and people transition costsOver 6,000 people impacted, but treated fairly through early retirement, relocation, or retraining and placement As of 1997, annual savings were well over $250 million (before tax) largest portion is due to lower manufacturing expenses, operating fewer plants with less staff some savings in packing materials and ingredients but with fewer DCs delivery expenses actually increased Securities and Exchange Commission closely monitored and verified the savings

*SOLVED WITH LINDO ON A 486 33MHZ; MODELS HAD ROUGHLY 2000 VARIABLES AND 2200 CONSTRAINTS, ONLY 17 BINARY

LOT OF WORK INVOLVED IN DEVELOPING COST COEFFICIENTS: MATERIAL HANDLING, INVENTORY, TRANSPORTATION, CROSS-BORDER DUTIES.

FIXED COSTS NOT RELEVANT BECAUSE MOST WERE RENTED; FOCUS ON TRADEOFF IN REDUCTION IN TRANSPORTATION COSTS AND IMPROVEMENTS TO CUSTOMER SERIVCE PROVIDED BY ANOTHER WAREHOUSE

*SOLVED WITH LINDO ON A 486 33MHZ; MODELS HAD ROUGHLY 2000 VARIABLES AND 2200 CONSTRAINTS, ONLY 17 BINARY

LOT OF WORK INVOLVED IN DEVELOPING COST COEFFICIENTS: MATERIAL HANDLING, INVENTORY, TRANSPORTATION, CROSS-BORDER DUTIES.

FIXED COSTS NOT RELEVANT BECAUSE MOST WERE RENTED; FOCUS ON TRADEOFF IN REDUCTION IN TRANSPORTATION COSTS AND IMPROVEMENTS TO CUSTOMER SERIVCE PROVIDED BY ANOTHER WAREHOUSE

***EACH CATEGORY TEAM DEVELOPED FROM 2 TO 12 OPTIONS IN THE FINAL ANALYSIS. EACH OPTION SUBJECTED TO THOROUGH FINANCIAL ANALYSISAND RISK ASSESSMENT.

VERIFICATION: FIX PLANT LOCATIONS, AND SOLVED AN mip TO OPTIMIZE THE REST OF THE SUPPLY CHAIN, CHOOSING DISTRIBUTION CENTER LOCATIONS AND PRODUCT FLOWS.

SOLUTION WAS NOT SIGNIFICANTLY DIFFERENT FROM THE COMPOSED SOLUTION.

FINAL STEP - CLEAN SHEET STUDY TO DETERMINE IDEAL LOCATIONS FOR PLANTS AND DCs WITHOUT REGARD TO EXISTING LOCATIONS. LOWER BOUND ON ANNUAL OPERATING COSTS. USED CURRENT COST AS UPPER BOUND. TOTAL COST OF RECOMMENDED SOLUTION WAS ACTUALLY LOWER WHEN CAPITAL COSTS WERE FACTORED IN.