Professional Documents
Culture Documents
Interfaces
Publication details, including instructions for authors and subscription information:
http://pubsonline.informs.org
This article may be used only for the purposes of research, teaching, and/or private study. Commercial use
or systematic downloading (by robots or other automatic processes) is prohibited without explicit Publisher
approval, unless otherwise noted. For more information, contact permissions@informs.org.
The Publisher does not warrant or guarantee the article’s accuracy, completeness, merchantability, fitness
for a particular purpose, or non-infringement. Descriptions of, or references to, products or publications, or
inclusion of an advertisement in this article, neither constitutes nor implies a guarantee, endorsement, or
support of claims made of that product, publication, or service.
© 1998 INFORMS
INFORMS is the largest professional society in the world for professionals in the fields of operations research, management
science, and analytics.
For more information on INFORMS, its publications, membership, or meetings visit http://www.informs.org
Raw Material Management at Welch’s, Inc.
INTERFACES 28:5 14
WELCH’S
press raw grapes into juice. Welch’s pas- centrated juice. Concentration adds an-
teurizes the fresh juice and stores it in re- other step beyond the settling process.
frigerated tank farms. Packaging opera- From the time of harvest, it takes several
tions at each plant run at a steady pace all months to obtain concentrated juice (Fig-
year long and draw juice from the tank ure 1).
farms as needed to support production. The annual grape harvest varies a great
The value of the grape juice stored in the deal in size depending on the spring and
Downloaded from informs.org by [190.156.183.33] on 02 October 2017, at 08:48 . For personal use only, all rights reserved.
tank farms often exceeds $50 million per summer weather. Welch’s takes all grapes
year based on cash market value. grown on acreage contracted to NGCA.
Time must pass before the grape juice is Because weather has an impact on the
ready for conversion into finished jams, yield of grapes per acre, the exact size of
jellies, juices, and concentrates. Raw grape the grape harvest is never known until the
juice contains insolubles that slowly settle end of harvest when all grapes are picked.
to the bottom of the tank. After this set- Although each grower is subject to rigid
tling process, the juice is ready for use in quality standards, differences in sweet-
the production of finished goods. ness, color, and acidity do exist among
Some of Welch’s products require con- growing areas. Weather also plays an im-
Figure 1: In processing grapes into grape juice and end products, Welch’s goes through several
steps: (1) unprocessed grapes are pressed into single strength juice that is unsettled; (2) after a
period of time passes, insolubles settle to the bottom of the tank and settled juice results; both
settled and unsettled juice is used directly as ingredients; (3) settled juice is either concentrated
or undergoes further processing; concentrate is used as an ingredient or is shipped as a finished
product to bulk customers.
September–October 1998 15
SCHUSTER AND ALLEN
national consistency, Welch’s often trans- understand the #FS unit of measure for in-
fers juice for blending between plants. ventory and demand of grape juice. By
The Common Planning Unit converting #FS into tons, Welch’s achieves
Welch’s uses “tons” as the base unit of a universal physical unit of measure that
aggregate planning in materials manage- everyone understands. Members of NGCA
ment. A “ton” represents the amount of receive payment based on how many tons
grape juice obtained from pressing one ton (or #FS) of grapes they deliver to Welch’s.
of grapes. Different growing areas may In turn, Welch’s determines the profit per
produce grape juice that has different lev- ton (or #FS) sold as finished product in a
els of sweetness. To calculate a common given month.
inventory unit for grape juices of different Welch’s has a refined cost-accounting
sweetness, Welch’s uses instruments called system that calculates requirements for
refractometers [Gould 1977, p. 78] to find Concord and Niagara grapes by month in
the percent of water soluble solids (%WSS) of #FS and tons. The system accounts for the
recovery loss and the cost of converting
The plants process nearly Concord and Niagara grapes into finished
300,000 tons of grapes per product. In June 1996, the company imple-
year. mented a fully integrated material require-
ments planning (MRP) system that calcu-
different lots of grape juice. Using this lates time-phased requirements for all
measurement, it calculates pounds of fruit components needed to manufacture fin-
solids (#FS) using a standard table that ished products. The new MRP system
converts volume (gallons) and %WSS into takes advantage of relational-database-
#FS. It is #FS that becomes the base inven- information technology and operates in
tory unit for Welch’s accounting systems. real time rather than in batch mode. The
As juice is concentrated and reconstituted, new, minicomputer-based cost-accounting
%WSS may increase or decrease, but #FS and MRP systems also allow for extraction
remains the same for an initial amount of of data to computer spreadsheets. Yet,
grape juice. The Food and Drug Adminis- both these systems have two major draw-
tration (FDA) sets forth rules called the backs: (1) they assume infinite capacity
standard of identity that ensure juice is re- and do not consider operational con-
constituted to a set level of %WSS. This straints in MRP calculations, and (2) they
protects consumers by preventing compa- do not provide optimal cost solutions for
nies from adding excess water to blending juices.
INTERFACES 28:5 16
WELCH’S
Since most large-scale MRP systems do truck) to use for transfers of juice between
not consider capacity or material con- plants.
straints, it is hard to obtain even nonopti- The JAC makes decisions based on infor-
mal solutions to blending and logistics mation from existing cost-accounting and
problems. The Welch’s MRP system uses MRP systems. Both of these systems need
regenerative MRP logic. To make even mi- the input of a set of recipes to calculate
nor changes to bills of material, Welch’s grape requirements. Welch’s has a staff of
Downloaded from informs.org by [190.156.183.33] on 02 October 2017, at 08:48 . For personal use only, all rights reserved.
must make a complete run of the MRP research scientists trained in food technol-
system to obtain new net requirements for ogy who find the best recipe to satisfy
grape juice. The Welch’s MRP system consumer taste. In most situations, a range
takes about six hours per run. This virtu- of recipes will be acceptable to the con-
ally eliminates the possibility of using it to sumer. Some recipes may contain grape
interactively find nonoptimal solutions. juice bought outside the cooperative. Juice
The Juice-Availability Committee from other varieties of grapes is often
Deciding how to use the grape crop is a blended with Concord or Niagara juice.
complex task given changing demand and This gives Welch’s additional flexibility in
uncertain crop quality and quantity. To balancing supply and demand when there
deal with this situation, in 1984 Welch’s is a short grape crop. Given a group of
established a cross-functional committee feasible recipes, the JAC must choose the
that would be responsible for making de- set of recipes for each plant that minimizes
cisions on crop usage, the juice-availability total cost and meets operational
committee (JAC). Its members come from constraints.
staff and line departments within Welch’s. Historically, the JAC made its recipe de-
The JAC has representatives from cost ac- cisions after intense discussion. Trial-and-
counting, quality assurance, research and error methods guided thinking about the
development, engineering, plant opera- best set of recipes to use. Computer
tions, and logistics. Every month the com- spreadsheets often served an important
mittee meets to discuss crop-usage strat- role in what-if analysis. However, it would
egy with the goal of minimizing total take many iterations to arrive at a feasible
system cost within constraints set forth by plan by simultaneously adjusting recipes
plant operations. and the amount of grape juice to be trans-
Typical decisions concern ferred between plants.
(1) What recipes to use for major product Welch’s lacked a formal system for opti-
groups, mizing raw material movement and the
(2) The transfer of grape juice between recipes used for production. The opera-
plants, tions management literature shows several
(3) The control of carryover to ensure solutions for this problem [de Matta and
enough room for the next crop, Miller 1993; Liberatore and Miller 1985;
(4) The financial impact of recipe and Markland and Newett 1976]. Several of
transfer decisions, and these solutions build on the models of hi-
(5) The mode of transportation (rail or erarchical integration set forth by Hax and
September–October 1998 17
SCHUSTER AND ALLEN
Meal [1975]. The work of these authors be- sioned a recursive solution method in
comes critical to planning in the process which the existing cost-accounting system
industries because of their treatment of ca- initially acts as a database, providing infor-
pacity. Taylor, Seward, and Bolander mation on grape juice demand to the third
[1981] note that process industries tend to model. In turn, the third model calculates
schedule capacity first and then materials. optimal recipes and interplant transfer
The existing cost-accounting and MRP sys- schedules based on operational constraints
Downloaded from informs.org by [190.156.183.33] on 02 October 2017, at 08:48 . For personal use only, all rights reserved.
tems at Welch’s lacked the ability to con- and cost. After it completes this calculation,
sider such elements of capacity as rates of the optimal recipes are input into both the
concentration and capability of transport- cost-accounting and MRP systems. The
ing grape juice between plants. Within the next output of both these systems will then
context of capacity, Welch’s needed sys- reflect an optimal cost plan and a set of rec-
tems that would consider a wide set of ipes that meets operational constraints. All
downstream users of the cost-accounting
Welch’s faced a difficult and MRP systems, other than the JAC, also
problem in juice logistics. receive benefits from having system out-
puts that reflect an optimal plan for grape
feasible recipes and take into account the juice logistics.
timing of the availability of juice from the When choosing a recursive method to
new crop. find solutions, it is important that the in-
To improve the cost-accounting and puts and outputs passing between steps in
MRP systems, we developed a third model the calculation be accurate. The cost-
that works independently but draws data accounting and MRP systems at Welch’s
from the cost-accounting system. Both the use a single, detailed bill of material
cost-accounting and MRP systems calcu- (BOM) for each recipe. Welch’s audits criti-
lated requirements for grape juice cor- cal BOMs once per month. This provided
rectly, but each lacked the ability to calcu- a high degree of confidence that the initial
late optimal-cost recipes and interplant output concerning grape juice demand
transfer schedules based on the opera- from the cost-accounting system would
tional constraints at each plant. Without not provide false data to the third model.
consideration of operational constraints, As we began to think of a mathematical
the output from the cost-accounting and formulation of the third model, one diffi-
MRP systems was of little practical use to culty remained. The managers at Welch’s
the JAC in making decisions on how best had little desire for mathematical models.
to use the grape crop. Wider use of inaccu- A tradition existed that decision making
rate outputs from the cost-accounting and should not rely on models but rather on
MRP systems caused additional problems consensus of the management group. Few
in forecasting financial performance and active models for decision making existed
managing the purchase of juice outside the at Welch’s.
cooperative. The First Solution
In employing a third model, we envi- In the fall of 1992, Welch’s faced a diffi-
INTERFACES 28:5 18
WELCH’S
cult problem in juice logistics. The Eastern processing industry demand is elastic for
1992 crop was high in acid and low in finished goods sold to customers. Even
yield. It was not possible to meet product small increases in price have drastic im-
specifications for acidity using the materi- pacts on demand. Complicating matters,
als at hand, so the JAC faced the complex retailers often drop items soon after manu-
logistics of blending large amounts of juice facturers announce a price increase. To re-
at several plants to maintain quality. The gain distribution, retailers require pay-
Downloaded from informs.org by [190.156.183.33] on 02 October 2017, at 08:48 . For personal use only, all rights reserved.
September–October 1998 19
SCHUSTER AND ALLEN
JLM (8,000 variables) was much too large Matta and Miller 1993, p. 119]. A firm may
for a spreadsheet application. The version have excesses or shortages of inventory at
of What’sBest! that we purchased handled the end-item level while appearing to have
4,000 variables and 2,000 constraints. adequate inventory at the aggregate level.
Though larger versions of What’sBest! ex- Inventory excesses or shortages at the end-
ist, we thought that managing 8,000 deci- item level often balance each other out
sion variables in a spreadsheet with con- when viewed in aggregate terms. This
Downloaded from informs.org by [190.156.183.33] on 02 October 2017, at 08:48 . For personal use only, all rights reserved.
INTERFACES 28:5 20
WELCH’S
(see constraint 2). In between harvests, the than the previous methods of analysis.
variable cost of storing grape juice at each This reduced the time it takes to find the
plant is approximately the same. The geo- best recipes. By incorporating constraints
graphical separation between Welch’s bulk dealing with plant capacity (such as con-
storage facilities is such that the cost of straint 6), the model produced plans that
match actual conditions at the plants.
The JLM proved an invaluable The output of the JLM acts as a commu-
tool in planning between 1994 nication tool for the JAC. An optimal solu-
and 1996. tion, presented in standard format, be-
comes the official plan of action. We found
transporting grape juice between plants is that the JAC accepted the JLM only after
greater than the variable cost of storage. we incorporated the model into meetings
This means juice seldom moves between as a major discussion point. By shifting the
plants unless an individual plant has a JLM from being a special project to becom-
shortage. ing an operational tool, the chair of the
Tons shipped is the expression for de- JAC gained gradual acceptance from other
mand. The research and development de- members of the committee. Our experi-
partment sets maximum and minimum ences very much parallel those reported
amounts of patron (NGCA) sourced grape previously [de Matta and Miller 1993,
juice as part of each recipe. In the JLM, p. 120].
given parameters a and b serve as the As crop or demand conditions change,
maximum and minimum percentages of the JAC reruns the JLM to find a new opti-
patron grape juice content (see constraints mal solution. Our experience is that run
3 and 4). times are very small (less than one minute
The material balance (constraint 2) is a on a 486, 66 MHz microcomputer). Welch’s
key part of the JLM. This constraint deals runs the JLM about 10 times per month.
with inventory, tons shipped, and in-out The JLM proved an invaluable tool in
transfers for a plant. The transfer balance planning between 1994 and 1996. In July
(constraint 7) links raw-material move- 1994, Welch’s had a chance to purchase
ment between plants. With the recipe pa- some raw grapes on the open market. This
rameters a and b providing a range of pos- opportunity occurs every year. The sup-
sible recipes, the JLM chooses a single plier of the grapes had sold to Welch’s for
recipe for each product group that mini- several years. The JLM quickly showed
mizes the cost of transfer, the recipe, and that Welch’s did not need the extra grapes.
September–October 1998 21
SCHUSTER AND ALLEN
crop in 1996, Welch’s used the JLM to challenging problem has become of great
model over 100 scenarios of different reci- importance to Welch’s. We also plan to in-
pes and logistics costs in an effort to keep vestigate different ways to assign costs and
grape carryover under control. The JLM profit within the JLM. As we accumulate
was very useful in generating new optimal more experience with the JLM, new ques-
solutions based on the assumption that the tions arise concerning the best ways to
1996 grape crop would be large. However, minimize cost and also to maximize profit.
crop forecasts change very quickly as a re- Taking a broader view, the development
sult of weather conditions. A harsh winter of models like the JLM raises interesting
and a late spring frost diminished the an- questions about the role of cost-accounting
ticipated 1996 grape harvest to the lowest and MRP systems within the process in-
level in five years. Again the JLM pro- dustries. Virtually every MRP system cur-
vided a quick way to model the impact of rently installed assumes infinite capacity
the sharp loss in the grape crop. This al- when calculating net requirements and lot
lowed Welch’s to adjust recipes rapidly so sizes for materials. In multiplant situa-
that it could meet demand without fear of tions, similar to Welch’s, logistics costs and
running out of stock. capacity constraints remain important fac-
Conclusion tors for consideration in MRP systems.
The JLM is a classic application of LP to Most process-oriented firms recognize the
an aggregate planning problem. It pro- need for production-planning systems
vides important information for a key that consider capacity constraints; how-
decision-making committee with the re- ever, few have experimented with tech-
sponsibility of managing a large asset. We niques to impose constraints on MRP us-
met several snags to implementation but ing a recursive, third model approach.
were able to overcome these problems by Our experience with the JLM, and its in-
reducing the problem size and using terplay with existing cost-accounting and
spreadsheet optimization. By formulating MRP systems, leads us to conclude that
the JLM as a simple LP within a spread- third models can play an important role
sheet optimizer, we succeeded in explain- in planning and decision making. We be-
ing the model to people with different lev- lieve this a practical area for future
els of mathematical understanding. In research.
addition, those within Welch’s gradually Acknowledgment
accepted mathematical-modeling princi- We thank the two referees, associate edi-
ples as they used the JLM as a daily tor, managing editor, and editor for con-
INTERFACES 28:5 22
WELCH’S
September–October 1998 23
SCHUSTER AND ALLEN
Management Journal, Vol. 38, No. 4, pp. 39–47. MBAs at Indiana University,” Interfaces, Vol.
Allen, S. J. and Schuster, E. W. 1994, “Practical 26, No. 2, pp. 105–111.
production scheduling with capacity con-
straints and dynamic demand: Family plan- Donald F. Biggs, Director of Logistics,
ning and disaggregation,” Production and In- Welch’s, Inc., 3 Concord Farms, 555 Vir-
ventory Management Journal, Vol. 35, No. 4, ginia Road, Concord, Massachusetts
pp. 15–21.
de Matta, R. and Miller, T. C. 1993, “A note on 01742-9101, writes: “Ed and Stuart worked
the growth of a production planning system: together to develop the juice logistics
Downloaded from informs.org by [190.156.183.33] on 02 October 2017, at 08:48 . For personal use only, all rights reserved.
A case study in evolution,” Interfaces, Vol. 23, model as an improvement to our juice
No. 4, pp. 116–122.
management process.
Excel 5.0, Microsoft Corporation, Redmond,
Washington. “Since Welch’s is a farm cooperative,
Finch, B. J. and Cox, J. F. 1987, Planning and management of raw materials (grapes) en-
Control System Design: Principles and Cases for trusted to us by our growers takes on
Process Industries, American Production and
Inventory Control Society, Inc., Falls Church,
great importance. We constantly deal with
Virginia. changes in supply and demand, and must
Gould, W. A. 1977, Food Quality Assurance, AVI know how these changes affect grape in-
Publishing Company, Westport, Connecticut. ventory. The model provides an excellent
Hax, A. C. and Meal, H. C. 1975, “Hierarchical
integration of production planning and method to obtain quick answers to impor-
scheduling,” in Studies in the Management Sci- tant questions asked by senior manage-
ences, Vol. 1, Logistics, ed. M.A. Geisler, North ment. During the past several years of
Holland, Amsterdam.
operation, the model provided keen in-
Leon, L.; Przasnyski, Z; and Seal, K.C. 1996,
“Spreadsheets and OR/MS models: An end- sights to several significant problems. De-
user perspective,” Interfaces, Vol. 26, No. 2, cisions made after analysis of the model
pp. 92–104. results contributed meaningful cost
Liberatore, M. L. and Miller, T. 1985, “A hierar-
chical production planning system,” Inter-
reductions.
faces, Vol. 15, No. 4, pp. 1–11. “A brief example demonstrates the
Markland, R. E. and Newett, R. J. 1976, “Pro- model’s effectiveness. This past week
duction-distribution planning in a large scale Welch’s faced a major change in the ex-
commodity processing network,” Decision
Sciences, Vol. 7, No. 4, pp. 579–594. pected size of the 1996 grape crop. Several
Plane, D. R. 1994, Management Science: A Spread- runs of the model led to a new set of reci-
sheet Approach, Boyd and Fraser, Danvers, pes in response to the new crop estimate.
Massachusetts.
The new set of recipes minimized trans-
Taylor, S. G.; Seward, S. M.; and Bolander, S. F.
1981, “Why the process industries are differ- portation cost while maintaining estab-
ent,” Production and Inventory Management lished quality standards. The model ac-
Journal, Vol. 22, No. 4, pp. 9–24. complished these calculations in a fraction
Schuster, E. W. and Finch, B. J. 1990, “A deter-
ministic spreadsheet simulation model for
of the time previously required.
production scheduling in a lumpy demand “Through the years, the use of applied
environment,” Production and Inventory Man- management science has helped Welch’s
agement Journal,” Vol. 31, No. 1, pp. 39–42. with a number of important decisions. The
What’sBest!, release 2.0, LINDO Systems, Chi-
cago, Illinois. juice logistics model is one of several sys-
Winston, W. L. 1996, “The teachers’ forum: tems that contribute to strategic decision
Management science with spreadsheets for making.”
INTERFACES 28:5 24