Integrated Planning for Poultry Production at Sadia

Author(s): Miguel Taube-Netto

Source: Interfaces, Vol. 26, No. 1, Franz Edelman Award Papers (Jan. - Feb., 1996), pp. 3853
Published by: INFORMS
Stable URL: http://www.jstor.org/stable/25062084
Accessed: 05-09-2016 20:11 UTC
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms

INFORMS is collaborating with JSTOR to digitize, preserve and extend access to Interfaces

This content downloaded from 148.228.161.3 on Mon, 05 Sep 2016 20:11:29 UTC
All use subject to http://about.jstor.org/terms

Integrated Planning for Poultry Production at

Sadia

Miguel Taube-Netto

UniSoma Matem?tica para Produtividade SA

Rua Jos? Paulino 2236-13013-002-Campinasf SP-Brazil

and

IMECC-Departamento de Matem?tica Aplicada

Universidade Estadual de Campinas (UNICAMP)
Caixa Postal 6065-13081-970-Campinas, SP-Brazil

Sadia Concordia SA, the largest poultry producer in Brazil, pro

cessing over 300 million chickens and 11 million turkeys a year,
has increasingly used mathematical models since 1990 to im
prove decision making throughout its production chain. It has
saved more than $50 million over a three-year period, as a re
sult of (1) better conversion of feed to live bird weight; (2) im
proved utilization of birds to produce more than 300 products
classified by weight range, taking into account weight varia
tions between and within flocks; (3) almost 100 percent fulfill
ment of daily production plans with increased output of higher
value products; (4) greater flexibility and reduced lead time in
meeting market demand; and (5) timely and wide ranging stud
ies of different price and demand scenarios. This original and
pioneering use of operations research and management science
approaches in the poultry industry is also adaptable to diverse
production practices and applicable to other animal-processing
industries.

Sadia was founded in 1944 in Concor

dia, a town in the west of the state of

Santa Catarina in southern Brazil. Today

the Sadia group comprises 19 companies
Copyright (c 1996, Institute for Operations Research
and the Management Sciences

with 24 industrial plants spread across the

country. The group employs over 30,000
workers, has an annual income of more
than US$ 2.5 billion and exports its pro
INDUSTRIES?AGRICULTURE/FOOD
PROGRAMMING?LINEAR, APPLICATIONS

0092-2102/96/2601/0038$01.25

INTERFACES 26: 1 January-February 1996 (pp. 38-53)

This content downloaded from 148.228.161.3 on Mon, 05 Sep 2016 20:11:29 UTC
All use subject to http://about.jstor.org/terms

SADIA
duction to around 40 countries. Sadia is

each flock until a slaughter date decided

also the largest Brazilian producer of poul

by the company. The integrator is paid for

try, processed meat, pork, and beef and

each flock according to previously negoti

the second largest processor of soy beans.

ated criteria of performance and efficiency.

UniSoma is a management sciences and

operations research company, specializing

Although a simple concept, the integra

tion system involves operational problems

in industrial production planning, particu

of great complexity and magnitude that

larly in the agricultural business.

Sadia must deal with daily.

The work I describe below was carried

The company currently has seven

out by a UniSoma development and imple chicken plants (Table 1), which processed
mentation team of about 20 people in part more than 300 million chickens in 1994,
nership with Sadia's poultry business area,
which represents about 30 percent of
group income. The description emphasizes
the chicken sector.
In the context of US poultry production

and a turkey plant in Chapec?, which pro

cessed over 11 million birds in the same

year. Chicks are produced in seven hatch

eries near the plants.
Each plant tends to specialize in certain

[Thornton 1995], Sadia would be the sixth

families of products, each of which needs

largest producer in terms of the number of

birds in specific weight ranges. For exam

birds processed, with a production equiva

ple, the Concordia plant caters mainly for

lent to 21 percent of Tyson Foods, the larg the Middle East market and so processes

est US producer, and 42 percent of Gold

Kist, the second largest.

Chicken Production

Chicken production at Sadia began in

the 1950s with a very straightforward pro

cess. Birds were raised in farms owned by

the company. At slaughter time, the indus

smaller chickens with an average weight of

1.714 kg in 1994. The Chapec? plant, on

the other hand, produces for the domestic

market, which prefers larger chickens with

an average weight of 2.309 kg in 1994.

Sadia's chicken-derived product lines in

clude many kinds of whole chickens, dif

trial process was limited to simply pluck

ferent types of cuts, sausages, burgers, and

ing, eviscerating, and packing. The decisive

other processed items, totaling more than

change came in 1961 with the beginning

300 products. This number is much larger

of the "integration" structure, a system al

if one counts different packages of identi

ready being applied in the USA.

Used with growing success, integrated

cal products. The large product line stems

mainly from the need to cater to many dis

poultry production is a very simple con

tinct markets. For example, a whole

cept. An integrator is generally a small

chicken exported to Saudi Arabia must

farmer who grows a few basic crops. The

company supplies the integrator with a

weigh when ready for cooking between

0.975 kg and 1.025 kg. This is a small

flock of chicks, feed, and all necessary

chicken, with a very tight weight range.

technical support. The integrator's respon

However, the Japanese market likes cuts

that are not in demand elsewhere. Unlike

sibilities are to provide the growing houses

at his or her own expense and look after

other large global poultry processors that

January-February 1996 39
This content downloaded from 148.228.161.3 on Mon, 05 Sep 2016 20:11:29 UTC
All use subject to http://about.jstor.org/terms

TAUBE-NETTO
Number of
Plant Integrated (

Concordia

Number of Birds Bird Mean

Processed in 1994 Weight in 1994
(millions) (kg)

1,204
1,267
1,119

Chapec?

Dois Vizinhos

Toledo

Francisco Beltrao
Am?rico Brasiliense

Agroavicola

Total

1,061
644
414
350

6,059

309

64 1.714
31 2.309
62 1.715
68 1.802
31 2.301
29 2.007
24 2.100

Table 1: Sadia's processing plants.

have the simpler task of producing chick

mortality rate. The metabolism of the

ens around a fixed mean weight, Sadia

chickens is such that they gain weight very

quickly. The weight gain varies consider

faces a more complex challenge of plan

ning the growth and processing of chick
ens of a wide span of ages from 33 to 52
days, corresponding to a weight range of

ably with temperature and air humidity,

for example, leading to strong seasonal dif

Flock weight distributions range over

ferences in weight gain in unclimatized

houses. Nevertheless and very roughly
speaking, around two kg of feed leads to a

different products, causing strong inter

gain of one kg in weight, given a plentiful

dependence between products.

supply of water for chickens that are at

about 1.300 to 2.700 kg (Figure 1).

At any time in the life of a flock, bird

least 42 days old, an age at which the bird

weight is distributed approximately nor

is already large enough to be slaughtered

mally with a coefficient of variation be

for most products. However, this feed-to

, tween nine and 14 percent (Figure 2). The

coefficient of variation is to some extent

age. Consequently, when a flock is not

dependent on flock sex, the individual

grower, and the season of the year.

version of feed into weight the following

live-weight ratio increases rapidly with

slaughtered on a determined day, the con

The part of the growth curve that is of

day will be less efficient. Similarly, the

interest for production flocks is the practi

mortality rate of each flock not only in

cally linear section from 33 to 52 days.

Each flock behaves differently depending

on its breed, sex, physical construction of

its house, competence and dedication of

creases with age but also varies seasonally

and with several other factors.

The variability of bird weight within a

flock means that the flock supplies chick

the grower, and other factors.

ens for products in several weight ranges.

Throughout the growth of each flock,

the relevant variables that must be ob

This means that when planning the supply

served and controlled are, besides mean

must also consider the supply of birds in

weight, total feed consumption and the

of chickens in a certain weight range, we

neighboring ranges and the proportion of

INTERFACES 26:1 40
This content downloaded from 148.228.161.3 on Mon, 05 Sep 2016 20:11:29 UTC
All use subject to http://about.jstor.org/terms

SADIA
Live Weight Ranges (kg)

Products
W
h

1.400

1.900

2.400

2.900

WB1

WB2

WB3

0.900

WB4
WB5

WB6
WB7
WB8
BR1
BR2

BR3

BR4

e
a
s

BR5

BR6
BR7
BR8
BR9
LG1

L
e

LG2
LG3
LG4
LG5

LG6

T
h
i

TG1

TG2
TG3

TG4

TG5

Figure 1: Various live weight ranges are needed for certain Sadia products. Within their
ranges, the products can be further classified in narrower ranges. Some products are packed in
boxes of a standard weight, which causes an additional effort of combining products of differ
ent weights to make up a box with the least variation from the standard.

birds in each flock that will be declassified

vantage of their parts. Even perfect birds

(that is, not good enough for sale as whole

carcasses) for various reasons. Some of the

can be cut up. For example, the growing

demand for white meat means that an in

declassified birds can be cut up to take ad

creasing proportion of perfect carcasses are

January-February 1996 41
This content downloaded from 148.228.161.3 on Mon, 05 Sep 2016 20:11:29 UTC
All use subject to http://about.jstor.org/terms

TAUBE-NETTO
4.000 y

0 5 10 15 20 25 30 35 40 45 50 55 60

Age (days)

Figure 2: A flock growth curve shows weight variability at each age.

cut up for breast-related products, which

sell for a better price. The different de

mand with the greatest possible profit. Fur

thermore, it must decide which new flocks

mands for higher quality chicken parts

must be housed with integrated growers to

generate a supply of lesser quality parts.

supply sufficient raw material to meet the

Product complementarity shows up in sev

eral ways in each market. We thus see the

demand forecast for the following months.

Each plant processes between 10 and 20

importance of a method of integrated plan

flocks of 12,000 chickens daily. Sadia

ning that takes into account the relation

houses similar number of chicks every day

among the 350 to 1,300 growers around

ships between the supply of chicken prod

ucts and markets.
Initially, from a tactical point of view,

Sadia must allocate products to plants, and

then it must plan daily production for each
of the seven plants, taking the demand of
the domestic and export markets into ac

each plant.
Grandparents, Breeders and Broilers
The process of producing chicks to be
distributed to the integrated growers be
gins with the purchase of chicks from spe

count. Next it must select flocks to provide

cialist poultry genetics companies. Male

line (ML) chicks can be of either the male

the raw material they need to implement

or female sex, but produce better male

the daily plans. In making these three deci

sions, it must take into account all the

chicks. They are purchased and housed at

technical constraints, such as plant slaugh

82 percent female chicks. Similarly, female

the proportion of 18 percent male chicks to

tering and evisceration capacity, the avail

line (FL) produces better female chicks.

ability of facilities needed to produce cer

They are purchased and housed at the pro

tain products, process yields, different pro

portion of 15 percent male chicks to 85

duction costs, prices of each product in

percent female chicks. The total ML and

each market, always seeking to meet de

FL chicks are purchased at the proportion

INTERFACES 26:1 42
This content downloaded from 148.228.161.3 on Mon, 05 Sep 2016 20:11:29 UTC
All use subject to http://about.jstor.org/terms

SADIA
of 15 percent of ML to 85 percent of FL.

Both ML and FL are called grandparents.

The eggs laid by the grandparents (first

ment in the global conversion of feed to

live weight, that is, in the amount of feed

needed to produce a given quantity of live

generation) are incubated to produce

weight.

breeder chicks (second generation) which

Processing

are sorted by sex and, after more rigorous

After deciding which flocks will be pro

male selection, housed in a 1:12 male/fe

cessed on a given day, Sadia must deter

mine the pickup time from each grower

Excess ML and FL grandparent chicks and concerned, taking into account the weight
distribution of each flock and the travel
male breeder chicks are housed with the

male proportion, again in Sadia hatcheries.

integrated growers along with the broiler

time from the grower to the plant. In de

chicks (third generation) that hatch from

termining pickup times, we seek to avoid

the eggs laid by breeders. The housing

having large numbers of chickens waiting

(that is, placement of birds in a flock) and

on the arrival platform before being hung

discarding dates of the grandparent and

on the processing lines, each of which is

breeder flocks are scheduled according to

adjusted for different ranges of live weight.

the demand for broiler chicks, taking into

After hanging, the chickens undergo a

account the laying rate of each flock, feed

sequence of operations: stunning shock,

costs, available houses, and the minimum

bleeding, plucking, evisceration, inspec

waiting time needed for cleaning and sani

tizing the houses between flocks.

In the grandparent and breeder produc

tion, carcass-quality classification, chilling,

rehanging after chilling, weighing, sorting

by quality and weight to the whole

tion cycles, several activities are responsi

chicken lines and cuts lines with optional

ble for the productivity and quality of the

deboning, packing, freezing or chilling,

chicks produced. Nutritional and sanitary

finished stock, and shipping.

conditions very much influence the laying

Good planning of the flocks to be col

rate, egg size, proportion of broken eggs,

lected and processed daily, synchronized

and hatching rate. These parameters also

with the sequence of operations and pro

depend on the age of the grandparents and

cess capacity, leads to daily production

breeders so that housing and discarding

plans with improved performance.

decisions strongly affect the supply of

good quality chicks.

The treatment that chickens for pro
cessing (broilers) receive in the form of
temperature, humidity, lighting, ammonia
concentration, and feed and water con
sumption is crucial for flock performance.

In unautomated houses, as in Sadia's case,

The plants store the products in freezer

chambers for later shipping or ship them
directly in refrigerated trucks. Sadia out
sources a fleet of about 1,300 trucks that
transport products directly to such clients

as supermarkets or to regional distribution

centers. At these centers, an additional
fleet of 300 owned trucks distributes

the grower makes a great impact. Selecting products to small customers.

flocks to meet certain performance param
eters brings about a substantial improve

Integrated Planning
The processes involved in growing

January-February 1996 43
This content downloaded from 148.228.161.3 on Mon, 05 Sep 2016 20:11:29 UTC
All use subject to http://about.jstor.org/terms

TAUBE-NETTO
broiler chickens have been described by

mathematical programming techniques.

North and Bell [1994]. Several authors

(Pesti [1992] for example) discuss the eco

The system is routinely supported by sta

nomic issues concerning these processes,

suggesting the use of computers and sim

ple optimization approaches. I do not

know of any optimized integrated poultry
planning systems with the breadth of the
system installed at Sadia, named PIPA, the
Portuguese acronym for Integrated Poultry

tistical studies and is complemented by a

feed formulation module.
The strategic module plans the overall
material flow of eggs to incubators, chicks
to growers, broilers to plants, and products

to markets, as well as matching resource

capacities.
This module permits what-if investment
analysis; for instance, planners can evalu

Production Planning.
The PIPA system seeks to optimize deci

ate the effects of opening a new plant, en

sions throughout the production stages. It

tering a new market, introducing new

supports planning and control activities all

products, specializing plants to a particular

along this decision chain, answering such

questions as these:
?How many grandparent chicks should
Sadia purchase and when?
?When should Sadia discard and replace

set of products, or associating products to

markets.

To allocate products to plants over time,

the system must represent transient condi
tions, such as the current state of flocks at

current flocks of grandparents and parents? each site and product demand over time.

?When should it house a flock of broiler

The tactical and operational modules per

chicks at a particular grower? (This is a key form this dynamic planning and control.
The first of the two tactical modules,

decision of the PIPA system since housing

is a commitment of resources to meet fu

Chick Planning, simultaneously synchro

ture demand.)

nizes the laying cycles of grandparent and

?When should it slaughter each flock?

parent chickens with the requirements for

(This is also a critical decision as flocks

broiler chicks. It determines the optimal re

should be slaughtered in line with con

placement policy for grandparent and par

firmed demand.)

ent cycles, minimizing the total cost of

?How much of each product should it

allocate to each plant quarterly, monthly,

and weekly?
?How can it match flocks with slaughter
ing and production capacity every day?

?How can it synchronize flock pickup

broiler chick production.

A central team in Chapec? coordinates

the whole complex of sites, each of them
consisting of growers and a plant. It inter
acts with local teams to provide production
guidelines to each site. In doing this tacti

with hanging to provide a proper weight

cal planning, it considers estimated global

distribution during daily production?

demand for families of products over a ho

rizon of 12 to 18 months in order to allo

The PIPA system is composed of several

interacting modules, arranged at three lev

cate production to each plant according to

els (Figure 3): strategic, tactical, and opera

overall grower capacity and plant slaugh

tional. All the modules are optimized using

tering capacity at each site. Each product is

INTERFACES 26:1 44
This content downloaded from 148.228.161.3 on Mon, 05 Sep 2016 20:11:29 UTC
All use subject to http://about.jstor.org/terms

SADIA

PIPA SYSTEM

Integrated

Global
Planning

Statistical

Support

Site
Planning

Chick
Planning

Feed
Formulation

Flock

Planning & Control

Plant
Planning & Control

Shift

Planning & Contr

Figure 3: The PIPA system (Integrated Planning for Poultry

structure, organized on three levels: strategic, tactical, and o

characterized by the rangeperforms

of bird weight
the actual housing t

ular grower
from which it can be produced
and bywith the support

Planning
and Control
modul
yield coefficients relating its
final weight
to
ule
also
determines
the
indivi
carcass weight (or parts weight).

slaughtered
every day base
Following the productionbe
guidelines,
the
rent
estimated performance o
second tactical module, Site
Planning,

ously housed.
The Plant Plan
determines housing, slaughtering,
and

Control
takes this slau
production schedules for each
site module
based

schedule
its consequent d
on the current status of flocks
and and
daily

weight
distribution
into accou
slaughtering capacities, the
estimated
de

mining
the
site's
mand for families of products,
the
prices
ofproduction s

the costs,
next seven
to 15 days, ba
products, and feed and chick
aiming
firmed
at maximizing a site's margin.
At demand.
this

more
formal description of
point, the main decision is A
how
many

and
plant
planning and contro
chicks will be housed every
day
to provide
Using distribution
representative growth
the proper amount and weight

of flocks
of birds for the followinggroups
months.
Sadia at different c

January-February 1996 45
This content downloaded from 148.228.161.3 on Mon, 05 Sep 2016 20:11:29 UTC
All use subject to http://about.jstor.org/terms

TAUBE-NETTO
of growth, specified by sex and lineage, the distribution will determine the best pro

Site Planning module decides the slaugh

tering age of each group of flocks in order

duction schedule for the day.

The Flock Planning and Control module

to supply enough birds to meet demand

identifies the flocks to slaughter on a par

for families of products, respecting slaugh

ticular day, based on information about

tering capacities. In each group of flocks,

the flock groups from the Site Planning

the variable Y(g, j) denotes the percentage

module. It also identifies the growers to

of birds in the group g to be slaughtered

house on a particular day based on the to

on day ;. This percentage is adjusted later

tal number of chicks to be housed that day

to correspond to the slaughtering of indi

from the Site Planning module.

vidual whole flocks when day j actually

On the frontier between tactical plan

occurs. The variable X(b, i, j) determines

ning and operational control, because of

the number of chicks of type b to be

the variable and fast growth of flocks, Sa

housed on day / for slaughtering on day j.

These numbers of chicks to be housed will

dia must plan the daily matching of weight

be adjusted when day / actually occurs to

Each plant processes between

10 and 20 flocks of 12,000
chickens daily.

house individual flocks (usually containing

about 12,000 chicks). The variable W(f, j)
is the amount of tons of the product family
/ to be produced on day /. These three sets
of variables are related in order to meet

distribution with production requirements

future demand for all products over time,

slaughtering, and pickup. This module bal

every day.
Sadia begins matching birds and pro
cessing during a production shift by sched

ances the activities, compensating for low

uling flock pickups from growers during

respecting daily capacities of housing,

or zero production on weekends and holi

in advance but adjust it operationally

the day. It uses a deterministic simulation

days. It also compensates for variations in

module that is integrated with the control

growth performance over time and market

of the temporary stock on the bird arrival

seasonalities. The module provides infor

mation for flock planning, from which Sa
dia can schedule individual flocks for both

weight distributions for the production

shifts that will be considered for the Shift

housing and slaughtering. It is also an im

Production Planning and Control module.

portant tool for setting tactical sales com

This operational module takes into account

mitments.

platform. The simulator also provides the

product specifications; bird weight distri

On a particular day, Sadia must know

bution for each shift; daily demand, which

the overall weight distribution of the birds

is determined by the previous module, and

planned to arrive at the plant, which de

production priorities in case a schedule is

pends on the characteristics of each group

of flocks. For each weight interval, the
model totals the number of birds consid
ered for slaughter on that day. This overall

spoiled by unforseen events; capacity and

weight range calibration of the whole
chicken lines and the cutting lines; and
breast and deboning capacity. The output

INTERFACES 26:1 46
This content downloaded from 148.228.161.3 on Mon, 05 Sep 2016 20:11:29 UTC
All use subject to http://about.jstor.org/terms

SADIA
of the module includes schedules for the

whole chicken lines and cutting lines; de

boning schedules; the level of production
for each shift; and the level of utilization
of slaughter house resources.
The tactical and operational planning

tained the heaviest chickens mainly from

male flocks (flocks in which all the chicks
are males) and the lightest from female
flocks because of the processing rules then
in force.
The managers in charge of planning re

horizons overlap to permit continuous ad

sisted breaking this paradigm, as was natu

justments. Each site has its own timing for

ral, but later came to appreciate the bene

planning and control, but the hierarchy

structure of the modules is the same at all

fits of the new methodology, which in

the sites. The Toledo site, for example,

supplies, better adaptability to market fluc

tuations, and more efficient utilization of

plans housing every week with a time ho

cluded improved predictability of chicken

rizon of six to 12 months; it also plans

processing capacity. This predictability per

flock slaughtering weekly with a time hori

zon of 60 to 120 days; it plans pickups

mitted them to shut down the plant for

one or more shifts or to schedule weekend

daily until 8:30 am, three to four days

production during a planning horizon.

ahead; it plans production based on al

They also insisted on another paradigm

ready determined flock slaughtering daily

here, the full use of processing capacity,

until 3:00 pm, seven days ahead; it plans

but soon came to view it more subtlely

shifts until 6:00 pm for the next day.

when they understood the mathematical

Implementation

model's power to show the interdepen

An early module developed by UniSoma dence of flock growth with market condi
for determining housing and processing

tions. Many decisions have until recently

dates was the first one Sadia implemented,

been based on product margins obtained

from historical accounting data. These de
cisions are now made using sensitivity
analysis and scenario studies based on

in December 1990 at the Concordia plant.

This module did not use detailed represen
tation of products and processes, later in
troduced in the Site Planning module. It
considered only the live chicken profile

product mix alternatives and other con

straint modifications, using better defined

that would best satisfy weekly production

objectives, including price-quantity elastic

needs, as judged by the production plan

ning team.

ity considerations.

This early version, and subsequently the

The installation of the Plant Planning

and Control module increased the plan

Site Planning module, improved the feed

ning level of detail. The differences be

to-live-weight conversion significantly for

tween planned and executed schedules di

minished. Sadia necessarily increased the

male, female, and mixed flocks (appendix).

These modules treated each flock individu

level of daily control and synchronized

ally through its growth, feed consumption,

bird flows in accordance with their weight

and mortality curves, to determine optimal

and with the pickup, hanging, and cutting

processing ages with the explicit objective

operations. Sadia introduced pickup plan

of minimizing cost. Previously, Sadia ob

ning at the same time in order to meet

January-February 1996 47
This content downloaded from 148.228.161.3 on Mon, 05 Sep 2016 20:11:29 UTC
All use subject to http://about.jstor.org/terms

TAUBE-NETTO
some of these needs.

modules is mathematical programming.

More recently, in June 1994, it imple

mented the Shift Planning and Control

UniSoma developed most of the mod

ules in the C programming language on a

module. This module has already out

Unix platform and used IBM's OSL mathe

grown the PIPA system, setting a chal

matical programming software [IBM 1991]

lenge for future improvements in the auto

mation and control of the plants.

Sadia began installing the Chick Plan

ning module for grandparents and breed
ers in November 1994. Its immediate func
tion is to indicate chick supply alternatives
that better use capacity in order to mini

mize supplies from third parties. These al

ternatives demand new data and adminis
trative routines.

except for the Shift Production Planning

and Control module and the Integrated

Global Planning module, which UniSoma

developed in GAMS [Brooke, Kendrick,
and Meeraus 1992] with a link to OSL.
The linear programming problems in the
modules are of varying sizes (Table 2).

The PIPA system has eight dedicated

IBM RS/6000 model 370 workstations,
seven of which are located in the pro

The Integrated Global Planning module

cessing plants and the eighth in SadiData

is a tool that enables Sadia to perform

in S?o Paulo city where part of UniSoma's

studies of the production process as a

development team is based.

whole. It performs such studies more fre

quently as production increases. The mod

el's intended original function of determin
ing master production plans is now per
formed by a recently installed multisite

planning module.
The modules of the PIPA system make

use of various mathematical resources. The

statistical support, for example, uses
econometric models and multivariate sta

Benefits

The direct benefits Sadia obtained with

the PIPA system can be divided into four

categories: (1) feed conversion improve

ment, (2) more high value products, (3)
faster response to market fluctuations, and
(4) greater sensitivity to market opportuni

ties. It also realized other less tangible

benefits.

tistics. The pickup schedule planning is

One of the most relevant performance

indicators for the poultry industry is the

simulation based. Generally speaking,

so-called feed consumption conversion ra

however, the technique most used in the

tio (or the ratio, as we will call it).

Typical Values of Problems Sizes

Module Number of Variables Number of Constraints Nonzero Elements

Site Planning 20,000 5,000 100,000

Flock Planning and Control 8,000 2,000 180,000

Plant Planning and Control 11,000 4,000 60,000
Shift Planning and Control 9,000 3,000 30,000

Chick Planning 130,000 35,000 250,000

Integrated Global Planning 5,000 2,000 90,000

Table 2: Large linear program sizes typically occur in modules of the PIPA system.

INTERFACES 26:1

48

This content downloaded from 148.228.161.3 on Mon, 05 Sep 2016 20:11:29 UTC
All use subject to http://about.jstor.org/terms

SADIA
The ratio is defined as the quotient of
the total feed consumption and the weight

ings somewhere in the range of $20 mil

lion (that is, 0.032 X 105,000 X 165 X 36)

of a bird. For a fixed final weight, a ratio as

to $24 million (that is, 0.039 X 105,000

small as possible is desirable. It is known

that the ratio varies with several factors,

X 165 X 36). Similar studies point to a

such as sex, breed, and climate, making it a

totaling a gain somewhere in the range of

reduction of another $5 million for turkeys,

very seasonal quotient. But the main

$25 to $29 million over a three-year

source of variation is bird weight (or its

period.

proxy, age). It so happens that as age in

Sadia achieved another direct benefit

creases, feed consumption and live weight

from the PIPA system by raising produc

curves diverge, such that the ratio in

tion of the so-called higher value products.

In the past three years, the

The implementation of the PIPA system

caused a shift in the average percentage of

benefits add up to over US$50

production of each product. Data show

creases continuously with age. Clearly, Sa

dia needed to search for a slaughtering

an average rise of 11 percent in weight of

production. Considering an average in

policy suitable to the production plan. De

come of $420 million in the past three

million.

that after the implementation of the PIPA

system, those higher value products show

ciding to put off slaughtering a flock at a

years, and that average net margin of

certain age, for instance, leaving it in the

higher value products is 15 percent higher

than other products (the last two estimates

field one more day, implies raising the av

erage weight of the flock at the expense of

a feed consumption higher than that ob

served until then.

Sadia conducted a study to compare the

ratio in each of its plants before and after
the implementation of the modules of the

PIPA system designed to support the deci

sions of housing and picking up the birds

(appendix).
We found that the mean decrease in the

are both quite conservative ones), we have

an increase in revenue of $6 million, lead
ing to estimates of a total of $18 million
over the past three years.

The process of housing is based on sales

forecasts. When actual sales differ too
much from forecasts, the profile of birds at
the growers conflicts with market needs.

With the PIPA system, Sadia can find, in

the fastest way, the best possible match

ratio ranges between 3.2 to 3.9 percent.

considering both new market needs and

Considering that (1) the feed consumption

the profile of birds at growers, avoiding

of chickens processed by Sadia is today

loss of income or taking advantage of bet

about 105,000 tons/month, (2) the cost of

ter opportunities.

a ton of feed is about $165, and (3) the av

erage length of time since the implementa

Such a benefit is difficult to quantify, ba

sically because it varies with the nature of
the disturbance that occurred in the mar

tion of the PIPA system in each plant is 36

months, then we have, as a result of the

ket. As an illustration, however, I can

gain in feed to live weight conversion, sav

mention that in the second half of 1994, as

January-February 1996 49
This content downloaded from 148.228.161.3 on Mon, 05 Sep 2016 20:11:29 UTC
All use subject to http://about.jstor.org/terms

TAUBE-NETTO
a consequence of fast moves programmed

administrative structure, the poultry sector

by the PIPA system after the implementa

belongs to a business unit, which is led by

tion of an economic plan in Brazil called

Piano Real, a rise in income of $4 million

a single executive, in contrast with the de

was observed.

quantitative representation of the decision

Several strategic analyses become possi

ble with the PIPA system. For instance,

one can look backward and see the im

provement in gross margin that could have

been reached in a given and fixed past,
keeping the exact scenario of birds housed
and considering actual market prices.
By varying processing age and sales mix

centralization of the processing plants. The

processes in the poultry sector by the PIPA
system is helping this new administration
make faster decisions with greater objectiv
ity and efficiency.

In the technological and investment

sphere, Sadia analyzes decisions in an inte

grated context. For example, PIPA's repre
sentation of the production processes and

in 10 percent more or less of each product,

their relationship with the supply of chick

one can increase gross margin by up to 14

ens has stimulated planners to perform

percent [Elias and Lopes 1994]. This kind

more detailed studies about the specializa

of analysis points the way to improvement,

showing the path to raising gross margin

tion of processing plants, the redefinition

of distribution centers, the introduction of

through a more appropriate, and still feasi

new breeds, the instrumentation of houses

ble, sales mix. Elias and Lopes estimated

to control temperature, humidity, and feed,

potential gains, in this area, above $30 mil

lion per year.

and the automation of plants.

In summary, in the past three years, the

direct benefits of the PIPA system add up

to over US$50 million. This figure is ex

The complete representation through

mathematical modeling of the decision

chain over different time horizons and the

integration of key production capacities

pected to increase in the next years as a re

sult of continuous improvements in the use

with the market environment proved to be

of the PIPA system.

profitability.

Other benefits include the following: (1)

a way towards increased efficiency and

Final Comments

duction plans are implemented as planned,

Today, Sadia sees the mathematical for

malization of the whole chain in the

which is particularly important for export

chicken and turkey sectors as essential to

Practically 100 percent of the critical pro

shipments. (2) Waiting time on the arrival

platforms has been reduced by 50 percent,
resulting in less weight loss and lower
mortality and more profitable use of chick

ens as Sadia takes advantage of short-term

market opportunities.

The PIPA system also helped Sadia to

consolidate organizational changes it has
carried out since April 1994. In the new

its utilization of technical and managerial

information in a manner consistent with its
objectives and with its physical, economic,
and administrative constraints. In fact,

managers better understand and accept

their responsibilities when they can see the
interdependencies between production
capacity and the roles of each part of the

decision chain.

INTERFACES 26:1 50
This content downloaded from 148.228.161.3 on Mon, 05 Sep 2016 20:11:29 UTC
All use subject to http://about.jstor.org/terms

SADIA
The treatment of the quantitative aspects Acknowledgments
of planning, via mathematics and statistics,
I thank Sadia Concordia SA for accept
complements other initiatives by Sadia, as
ing the challenge of putting together its
in the reengineering of its administrative

managerial capabilities with the more for

structure and its quality program. Sadia

mal framework of mathematical modeling

sees the techniques and attitudes in this

and for understanding the scope of this in

treatment as comprising a technology in it

teraction long before the results were ap

self, namely decision technologies. This

view facilitates interaction with other tech

parent. In particular, I thank ?lvio Flores,

nologies, such as information technology

this work started, for his early insight into

and automation.

Recently Sadia widened the scope of its

the director for animal production when

the potential benefits for Sadia; Alberto
Stringhini, the current animal production

partnership with UniSoma to include the

director, for his continuing encouragement;

development and implementation of the

and also Cloraci Sartori, central production

PIPRA system (Integrated Feed Production

planning manager, for his enthusiastic and

Planning) over the next three years. This

professional day-to-day support. In addi

system extends the idea of optimal feed

tion, I thank Walter Fontana Filho, presi

formulation over a planning horizon of

dent of the Sadia Group, for supporting

many time periods, encompassing the pur

chase of ingredients. The system will en
able Sadia to save between US$10 million

this work and its publication. I also grate

fully acknowledge the assistance received
from the UniSoma technical staff who

and US$30 million a year, according to

simulations carried out with a one-year

mando Milioni for the many discussions

planning horizon and seasonal price fluc

which resulted in a better paper and pre

tuations of ingredients.

We widened the PIPA system itself in

scope to develop a production supervision

fully share the merit of this work, Ar

sentation at the INFORMS meeting in Los

Angeles in April 1995, Alistair Clark for
suggesting clarifications in the paper, and

system in the processing plants that relies

Tarcisio Lopes and Marcelo Elias for their

on new technologies to dynamically weigh

modeling and implementation

the birds and improve the utilization of the

contributions.

freezing tunnels.

APPENDIX

The development of the PIPA system

makes a reality forecasts made by Cremers
[1994] in a special 10th anniversary issue
of World Poultry magazine: "The next step
will be to use all of this information in an
optimal way. This is a challenge for manu
facturers to make management software
programs which can run the whole prod
uct process based on pre-set but still

flexible goals."

Feed-to-Live-Weight Conversion
Improvement
The feed-to-live-weight conversion fac
tor (ratio) is one of the most relevant
performance indicators for the poultry in
dustry. It is a quotient that expresses the
relationship between the quantity of feed
consumed by a chicken or flock and its
weight. Naturally, for birds of identical
weight, a low ratio is better.
This indicator is very useful for assessing

January-February 1996 51
This content downloaded from 148.228.161.3 on Mon, 05 Sep 2016 20:11:29 UTC
All use subject to http://about.jstor.org/terms

TAUBE-NETTO
the benefit of the Flock Planning and Con
trol module and its predecessors, as all of
them optimally select the flocks to be col
lected to execute the specified production

plan.

Table 3 shows a comparison of the mean

ratio on the day of processing before and

after the implementation of the above
modules. The comparison is for each plant.

We present the mean and standard devia

tion of the mean monthly ratio of all the
flocks processed in the plant. The number
of lots processed per month in each plant
is around 500. Thus, 20 months of obser

vation corresponds to about 10,000 flocks.

The analysis is shown for only five of the
seven plants, as there are no records for
the other two (Francisco Beltr?o and
Agroavicola) before the PIPA system was
implemented.
At a one percent significance level, there
is enough evidence in four of the five
plants to reject the null hypothesis of equal

means and conclude that the ratio was re

duced after the systems were imple

mented. Similarly, we can conclude that
there was no change in the standard de
viation in four of the five plants. The only
significant change was in the Toledo plant,

which explains the impossibility of show

ing a reduction of the ratio in that plant.

We conclude from Table 3 that the mean

decrease in the feed conversion factor be

longs to a range that varies from 3.2 per
cent (if all plants are considered) to 3.9
percent (if we omit the Toledo plant,
where the variability of the standard de
viation harms the analysis).

One can argue that the simple compari

son between the ratio before and after the

installation of the systems cannot be justi
fied. In fact, a very significant alteration in
the profile of chickens processed in a cer
tain plant could influence the indicator

being observed.
The effect of processing age can be fil
tered out if one repeats the exercise de
scribed above, constraining the comparison
to flocks processed at identical ages. Such

an analysis was conducted by Zullo and

Milioni [1994], who showed that in these
circumstances the decrease in the ratio is

even more significant. Furthermore, they

showed that when the comparisons are
limited to the same ages, a second benefit
is noticeable, namely a systematic decrease
in the standard deviation of the ratio, indi

cating a less (and so desirable) process

Mean Monthly Ratio

Plant

Period

Concordia

before
after

Chapec?

before

Dois Vizinhos

before

Toledo

before

Am?rico

before

after
after
after
after

Standard

Number of Months

Observed

21

45
23
39
27
36
33
33
33
24

Mean

Deviation

2.00
1.94
2.12
2.03
2.00
1.91
1.97
1.96
2.17
2.09

0.033
0.036
0.049
0.050
0.045
0.049
0.028
0.039
0.074

Table 3: The feed-to-live-weight conversion factor (ratio) improved at five of the seven plants
with the installation of the PIPA system.

INTERFACES 26:1

52

This content downloaded from 148.228.161.3 on Mon, 05 Sep 2016 20:11:29 UTC
All use subject to http://about.jstor.org/terms

0.069

SADIA

Jan May Sept Jan May Sept Jan May Sept Jan May Sept Jan May Sept Jan May Sept
Period

Figure 4: The feed conversion rate decreased after the implementation of the PIPA system at

the Concordia plant.

Pesti, Gene and Miller, Bill R. 1993, Animal

variability.
Finally, Figure 4 pragmatically ends the Feed Formulation, Van Nostrand Reinhold,
New York.
discussion about the nature of the gain ob
served in the ratio. It shows that the gainsThornton, Gary 1995, "Nation's broiler indus
due to the introduction of the optimization try," Broiler Industry, Vol. 58, No. 1,
pp. 27-30.
systems are immediately noticeable, char
Zullo, Sergio and Milioni, Armando 1994, "Ra
acterizing a discontinuity in the process.
tio improvement with the use of the PIPA
This indicates that the gains are due more system," internal report, UniSoma,
to the introduction of these systems than Campinas, SP, Brazil.
to a possible genetic improvement.

References

Brooke, A.; Kendrick, D.; and Meeraus, A.

1992, "GAMS?A user's guide, release 2.25,"

The Scientific Press, South San Francisco,

California.

Cremers, Jan 1994, "Flexibility will be the key,"

World Poultry, Vol. 10, No. 1/2, pp. 71-73.

Elias, Marcelo and Lopes, Luciano 1994, "The

PIPA system and market sensitivity," internal
report, UniSoma, Campinas, SP, Brazil.
IBM 1991, "Optimization subroutine library

(OSL) user's guide, release 2."

North, M. and Bell, Donald D. 1994, Commer

cial Chicken Production Manual, fourth edi
tion, Chapman and Hall, New York.
Pesti, Gene 1992, "Computer applications in
poultry management," course notes, Depart
ment of Poultry Science, The University of
Georgia, Athens, Georgia.

January-February 1996 53
This content downloaded from 148.228.161.3 on Mon, 05 Sep 2016 20:11:29 UTC
All use subject to http://about.jstor.org/terms