12.7.

Aggregate Planning for Multi-Period Problems

As discussed in Sec. 12.2, inventory is another means to increase supply. Specifically, a firm can produce more than demand in
one period, keep the items in inventory, and use them as supply in future periods. This option can be utilized when aggregate
planning has planning horizon of more than a single period, that is, multi-period problems. Hence, in addition to overtime,
subcontracting, and setting up new resources, aggregate planning in multi-period problems can also use inventory to match its
supply with demand.

12.7.1. Single Aggregate Resource and Single Aggregate

Product
The most basic multi-period aggregate planning is when all products are aggregated into one aggregate product and all
resources are also aggregated into one aggregate resource. The following example presents such a case.

Example

Example 12.5

Delish Soup is a firm that produces about 27 different types of chicken, beef, seafood, and vegetable soups. Their
production system consists of two departments: kitchen and packaging. In the kitchen department, ingredients are washed
and prepared for different recipes and are cooked in large pots. The soups are then sent to the packaging department
where they are filled into cans, sealed, and cooked for another half an hour. The cans are then cooled down for 10 minutes
and labeled. While almost all of the operations in the packaging department are automated, the main tasks in the kitchen
department are done manually by workers.

Demand for Aggregate Product: The firm is in the process of planning its production and workforce for the next 6 months.
Since all soups are sold in cans of the same size (i.e., cans that contain 14.75 ounces of soup), managers have aggregated
all of their 27 different products into one, which we call "Soup." They also decided to use 100 cans as the unit of the
aggregate product, that is, one unit of the aggregate product is 100 cans of soups. The forecasted demand for the
aggregate product in the next 6 months is presented in Table 12.2.

Table 12.2 Forecasted Demand for the Aggregate Product (Soup) in Example 12.5

Month 1 2 3 4 5 6

Demand 60,000 240,000 192,000 90,000 210,000 120,000

Profit and Costs: The net profit of each aggregate unit, not considering worker wages and inventory holding costs, is about
$50. The firm pays workers an hourly wage of $8 during regular working hours and $12 during overtime. Because of the
change in demand from month to month, the firm often hires temporary workers to use for 1 or more months. Hired workers
start their jobs in the beginning of a month. A hired worker is fired at the end of a month, if the worker's service is not
needed in the next month. A hired worker will work for at least 1 month, and if needed can continue working for another
month and so on. The costs of hiring and firing a worker are $3000 and $4000, respectively. The cost of holding one unit of
the aggregate product (i.e., 100 cans) in inventory for 1 month is about $5.

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
 Resources: The main resource used in the production is labor, which is the bottleneck. Equipment (e.g., can fillers and
sealers) used in the process are semiautomated and have ample capacity to satisfy demand. A time study has shown that
6 minutes of a worker's time (mainly in the kitchen department) are required to process one aggregate product. The firm
has 50 permanent workers who work 8 hours a day and 25 days a month. The firm does not want to fire its 50 permanent
workers and wants to return to the workforce level of 50 at the end of the 6 months. Also due to space and equipment
limitations, the number of workers that can work in a month cannot be more than 70 workers.

To plan its workforce in the next 6 months, the firm needs to know if it needs to hire temporary workers, or use overtime
instead (or do both). Furthermore, it would like to know how many cans of soup it must produce each month and how much
inventory it should carry from month to month. The goal is to maximize the total profit in the next 6 months.

The problem that Delish Soup is facing is a standard multi-period aggregate planning problem that involves only one aggregate
product (i.e., batch of 100 canned soups) and one aggregate resource (i.e., labor who mainly work in the kitchen department).
The following analytics illustrates how this problem can be solved using mathematical programming models.

ANALYTICS

Multi-Period Aggregate Planning—Single Resource and Single Product

Periods and Products: Consider a process that uses one pool of "aggregate resource" (i.e., workers) to produce several products, which are all
aggregated into one "aggregate product." The aggregate planning is for T periods. Also, suppose

ℓ = Number of regular working hours in a day

nt = Number of working days in Period t

Resources: The process has one main aggregate resource—pool of w workers—that processes the product, where

T eff = Effective mean process time (measured in hours) of a worker to process a product

Temporary workers can be hired in the beginning of a period to work for at least one period. A temporary worker, if not needed, is fired at the end
of a period.

Demand and Costs:

dt = Forecasted demand for the aggregate product in Periodt

r = Net profit of one product (not including resource hourly cost and inventory costs)
cI = Inventory holding cost of one unit of aggregate product in a period
cR = Hourly cost of one resource used in regular working hours
cO = Hourly cost of one resource used in overtime hours
cH = Cost of hiring one worker
cF = Cost of firing one worker

Note that in this model (as opposed to the single-period model), net profit r does not include the resources' (regular) hourly cost. Also, c O is the
total cost of one overtime hour of a resource (not just the additional cost beyond regular hourly cost).

Decision Variables: Decisions are the production, inventory, and workforce levels in each Period t = 1, 2, … , T , which are represented by the
following decision variables:

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
 ANALYTICS
Production and Sales :
−−−−−−−−−−−−−−−−−−−−
Xt = Number of products produced in Period t
St = Number of products sold in Period t
Inventory:
−−−−−−−−−
It = Number of products in inventory at the beginning of Period t
Workforce Level:
−−−−−−−−−−−−−−−
Ht = Number of workers hired in the beginning of Period t
Ft = Number of workers fired at the end of Period t
Resource Work Schedule:
−−−−−−−−−−−−−−−−−−−−−−
Wt = Number of workers in Period t
Ot = Total number of overtime worker-hours in Period t
O max = Maximum allowed overtime hours for a worker in a day

Mathematical Programming Model: The mathematical programming model that finds the optimal production, overtime, inventory, and workforce
levels in order to maximize the total profit in the next T periods is the following linear programming:

T
∑
Maximize Z= t=1 rSt − c IIt − c R ℓ nt Wt − c OOt − c HHt − c FFt
subject to :

T effXt − ℓ ntWt − Ot ≤ 0 for all t = 1, 2, … , T (Resource time constraints)

St ≤ dt for all t = 1, 2, … , T (Sales constraints)
Ot − ntO maxWt ≤ 0 for all t = 1, 2, … , T (Maximum overtime constraints)
Wt − Wt−1 + Ht − Ft−1 = 0 for all t = 1, 2, … , T (Workforce conservation constraints)
It − It+1 + Xt − St = 0 for all t = 1, 2, … , T (Production conservation constraints)

Xt, St, It, Ht, Ft, Wt, Ot ≥ 0 for all t = 1, 2, … , T

where W0 = w is the current number of workers at the beginning of the planning horizon and F0 = 0 . Also, I 1 is the available inventory of the
aggregate product in the beginning of the planning horizon—beginning of Period 1, and the inventory at the end of planning horizon is zero, that is,
I T +1 = 0 .

The period in the Delish Soup multi-period aggregate planning inExample 12.5 is a month and the firm is facing an aggregate
planning for a 6-month planning horizon (i.e., T = 6). Using the decision variables defined in the above analytics, we develop
the objective function and constraints as follows.

Objective Function

In the Delish Soup example, we have net profit of selling a unit of the aggregate product (i.e., 100 cans of soup) to ber = $50.
The costs include inventory holding cost of cI = $5 per month, cost of regular hour cR = $8 per worker per hour, cost of
overtime cO = $12 per worker per hour, cost of hiring cH = 3000 per worker, and cost of firing cF = $4000 per worker. Thus,
the objective function of the aggregate planning that maximizes the total profit in the 6 months would be

Maximize Z = 50(S1 + … + S6) − 5(I1 + … + I6) − 8(25 × 8)(W1 + … + W6)

−12(O1 + … + O6) − 3000(H1 + … + H6) − 4000(F1 + … + F6)

Resource Time Constraint

=6 = 0.1
© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
Recall from the example that it takes a worker to process a unit of aggregate product aboutTeff = 6 minutes = 0.1 hours.
Also, the firm works ℓ = 8 regular hours a day and n t = 25 days a month. Thus, the total number of regular working hours in a
month is ℓn t = 8(25) = 200 hours. Using our analytics, the resource (i.e., labor) time constraints would be

0.1 X1 − 200 W1 − O1 ≤ 0
0.1 X2 − 200 W2 − O2 ≤ 0
0.1 X3 − 200 W3 − O3 ≤ 0
0.1 X4 − 200 W4 − O4 ≤ 0
0.1 X5 − 200 W5 − O5 ≤ 0
0.1 X6 − 200 W6 − O6 ≤ 0

Each constraint is similar to what we had in the single-period aggregate planning problem with overtime. However, we have six
of them, one for each of the 6 months.

Sales Constraints

The forecasts for demand in each month are shown inTable 12.2. The sales in each month t (i.e., St) cannot be larger than the
demand in that month (i.e., dt); in other words, St ≤ dt. Therefore, following the analytics, the sales constraints corresponding
to the aggregate planning model of Delish Soup for the 6 months would be:

S1 ≤ 60,000 S2 ≤ 240,000 S3 ≤ 192,000

S4 ≤ 90,000 S5 ≤ 210,000 S6 ≤ 120,000

Overtime Constraints

The maximum allowed overtime for a worker in a day is O max hours. Hence, the maximum allowed overtime for a worker in a
month is n t O max. If there are Wt workers in month t, then the maximum total worker-hours of overtime for all workers will be
nt O max Wt. Hence, when deciding for the total overtime worker-hours Ot for month t, we must have it less than the maximum
allowed; specifically, Ot ≤ n t O max Wt or Ot − n t O max Wt ≤ 0, as shown in the analytics. For Delish Soup we have n t = 25
days and O max = 3 hours, then n t O max = 25(3) = 75. The corresponding maximum overtime constraints are therefore,

O1 − 75 W1 ≤ 0
O2 − 75 W2 ≤ 0
O3 − 75 W3 ≤ 0
O4 − 75 W4 ≤ 0
O5 − 75 W5 ≤ 0
O6 − 75 W6 ≤ 0

Workforce Conservation Constraints

The main idea behind workforce conservation constraints is to model how hiring and firing of workers in a period impact the
available workforce in the next period. Recall that Delish Soup hires workers to start in the beginning of a month and fires them,
if needed, at the end of a month. Suppose that the current month is montht. The workforce conservation constraint implies
that the number of workers that will be working this month (i.e., Wt) is equal to the number of workers that were working last
month (i.e., Wt− 1) minus the number of workers who were fired at the end of last month (i.e.,Ft− 1) plus the number of workers
that are hired in the beginning of this month (i.e., Ht). In other words,

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
Wt = Wt− 1 + Ht − Ft− 1 ⇒ Wt − Wt− 1 − Ht + Ft− 1 = 0

Considering that Delish Soup currently has W0 = 50 permanent workers and since F0 = 0, the above constraint for the 6
months are as follows:

W1 − H1 = 50
W2 − W1 − H2 + F1 = 0
W3 − W2 − H3 + F2 = 0
W4 − W3 − H4 + F3 = 0
W5 − W4 − H5 + F4 = 0
W6 − W5 − H6 + F5 = 0

The firm wants to keep its current 50 permanent workers throughout the planning horizon. Therefore, we must add the
following constraints to make sure that the workforce level in each period does not go below 50:

W1 ≥ 50 W2 ≥ 50 W3 ≥ 50
W4 ≥ 50 W5 ≥ 50 W6 ≥ 50

On the other hand, as stated in the example, because of the limitation in space and equipment, the number of workers working
in a month cannot exceed 70. This is modeled with the following constraints:

W1 ≤ 70 W2 ≤ 70 W3 ≤ 70
W4 ≤ 70 W5 ≤ 70 W6 ≤ 70

Finally, since the firm would like to end the planning horizon with the same 50 workers, we must have the following constraint:

W6 − F6 = 50

The above constraint is obtained by writing the workforce conservation constraint for period (i.e., month) 7. Note that the
number of workers at the end of period 6 is equal to the number of workers at the beginning of period 7 (since no one is hired in
period 7). Hence, the firm wants to have W7 = 50. For period 7, the workforce conservation constraint is
W7 = W6 + H7 − F6 = 50, which also guarantees that the number of workers at the beginning of period 7 is 50. Considering
that the firm does not hire any workers in period 7, that is, H7 = 0, then we get W6 − F6 = 50.

Production Conservation Constraints

The production conservation constraints imply that a unit produced in a period will either be sold in that period, or will be carried
as inventory to be sold in future periods. Consequently, there is a relationship among the inventory at the beginning of a period
and production quantity and sales in the previous period. Suppose that the current month is month t + 1. The inventory at the
beginning of this month (i.e., It+ 1) is equal to the inventory at the beginning of the last month (i.e., It) plus what was produced
last month (i.e., Xt) minus what was sold last month (i.e., St); therefore, the production conservation constraint for month
t + 1 will be

It+ 1 = It + Xt − St ⇒ It − It+ 1 + Xt − St = 0

=0
© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
Considering that the firm does not currently have any inventory (i.e.,I1 = 0) and that the firm does not want to have inventory
at the end of the planning horizon (i.e., I7 = 0), we have the following production conservation constraints for Delish Soup:

−I2 + X1 − S1 = 0
I2 − I3 + X2 − S2 = 0
I3 − I4 + X3 − S3 = 0
I4 − I5 + X4 − S4 = 0
I5 − I6 + X5 − S5 = 0
I6 + X6 − S6 = 0

Finally, we should not forget important constraints Xt ≥ 0, St ≥ 0, It ≥ 0, Ht ≥ 0, Ft ≥ 0, Wt ≥ 0, Ot ≥ 0 for all

t = 1,2,… ,6.
Optimal Solution

Using Excel Solver we find the optimal solution for Delish Soup's multi-period aggregate planning problem as shown inFig.
12.3.

From the optimal solution in Table 12.3 we conclude the following:

Production and inventory: In each month, the firm should produce equal to the demand in that month, except for the first and
fourth months, in which the firm should produce more than the demand. Specifically, in the first month the firm should
produce 47,500 aggregate products more than the demand in that month and should carry them as inventory to satisfy
demand in the second month. Also, in the fourth month, the firm should produce an additional 17,500 units more than
demand in that month and use them to satisfy the demand in the fifth month.

Workforce: To achieve the above production plan, the firm should hire 4 and 16 temporary workers in the first and the
second month, respectively—a total of 20 workers. It should then fire 10 of those workers at the end of the fifth month and
fire the rest at the end of the sixth month.

Overtime: The aggregate plan calls for overtime in the second, third, and the fifth month. Specifically, in the second month
the firm must have 5250 worker-hours of overtime. Having W2∗ = 70 workers working in the second month, and n 2 = 25
working days in a month, this translates into O∗2/(n 2W2∗) = 5250/(25 × 70) = 3 hours of overtime for each worker in
that month. This number is 2.97 hours and 3 hours for the third and the fifth months.

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
Table 12.3 Optimal Production, Inventory, and Workforce Plans for Delish Soup in Example 12.5

Optimal solution Month 1 Month 2 Month 3 Month 4 Month 5 Month 6

Sales (St∗ ) 60,000 240,000 192,000 90,000 210,000 120,000

Production (Xt∗ ) 107,500 192,500 192,000 107,500 192,500 120,000

Inventory (I t∗ ) 0 47,500 0 0 17,500 0

Workforce level (Wt∗ ) 54 70 70 70 70 60

Overtime hours (O∗t ) 0 5250 5200 0 5250 0

Hiring (Ht∗ ) 4 16 0 0 0 0

Firing (Ft∗ ) 0 0 0 0 10 10

The above plan results in a total profit of $44,316,200 in the 6 months.

One point worth mentioning here is the relationship between Ht and Ft− 1 in the optimal solution. As Table 12.3 shows, in all
months we either have (Ht∗ ≠ 0 and Ft∗− 1 = 0), or (Ht∗ = 0 and Ft∗− 1 ≠ 0), or we have (Ht∗ = 0 and Ft∗− 1 = 0). In the optimal
solution we will never have (Ht∗ ≠ 0 and Ft∗− 1 ≠ 0). The reason becomes obvious when we consider an example of (
H3 = 5 ≠ 0 and F2 = 3 ≠ 0). This implies that the firm should fire three workers at the end of month 2, and then immediately
hire five workers in the beginning of month 3. This will increase the workforce level by two workers from month 2 to month 3.
Under this hiring/firing plan, the firm will pay firing cost for three workers and hiring cost for five workers. However, a cheaper
way to increase workforce level by two workers is not to fire anyone at the end of month 2 (i.e., F2 = 0) and just hire two
workers in the beginning of month 3 (i.e., H3 = 2). Under this plan, the firm only pays the hiring cost for two workers. Hence,
the optimal solution of an aggregate planning always results in Ht∗ × Ft∗− 1 = 0.

12.7.2. Multiple Aggregate Resources and Multiple

Aggregate Products
The Delish Soup case is an example of a basic multi-period aggregate planning problem with a single aggregate product and
single aggregate resource. In this section, we extend our analytics to cases in which products are aggregated to more than one
aggregate product and there is more than one aggregate resource that processes these products.

ANALYTICS

Multi-Period Aggregate Planning with Multiple Resources and Multiple Products

Consider the basic multi-period model and suppose that all products are aggregated into K different aggregate products processed by I
different pools of resources. Aggregate Resource Pool i is a pool of wi resources (i.e., workers) who work nit days in Period t . Also, let

( )
T effk (i) = Effective mean process time of a worker in Resource i to process a product of type k

( )
where T effk (i) is measured in hours.

Demand and Costs:

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
 ANALYTICS
( )
dt k = Forecasted demand for aggregate Product k in Period t
r( k ) = Net profit of one Productk (not including resource hourly cost and inventory costs)
cI = Inventory holding cost of one unit of aggregate product in a period
c iR = Hourly cost of one resource in aggregate Resourcei used in regular working hours
c iO = Hourly cost of one resource in aggregate Resourcei used in overtime hours
c iH = Cost of hiring one worker for Resourcei
c iF = Cost of firing one worker of Resourcei

Decision Variables: Decisions are production, overtime, inventory, and workforce levels in each Period t = 1, 2, … , T , for aggregate Resources
i = 1, 2, … , I and aggregate Products k = 1, 2, … , K , which are represented by the following decision variables:

Production and Sales :

−−−−−−−−−−−−−−−−−−−−
Xkt = Number of Products k produced in Period t
Skt = Number of Products k sold in Period t
Inventory :
−−−−−−−−−−
Ikt = Number of Products k in inventory at the beginning of Period t
W orkforce Level :
−−−−−−−−−−−−−−−−
Hit = Number of workers hired for Resource i in the beginning of Period t
Fit = Number of workers fired from Resource i at the end of Period t
Resource W ork Schedule :
−−−−−−−−−−−−−−−−−−−−−−−
Wit = Number of workers in Resource i in Period t
Oit = Total number of overtime worker-hours of Resource i in Period t
O max = Maximum allowed overtime hours for a worker in a day

Mathematical Programming Model: The mathematical programming model that finds the optimal production, overtime, inventory, and workforce
levels in order to maximize the total profit in T periods is

T K K I
∑ ∑ ∑ ∑
Maximize Z = t=1 ( k=1 r( k) Skt − k=1 c IIkt − i=1 [ c iR ℓ nit Wit + c iOOit + c iH Hit + c iFFit])

subject to :

K
∑
( )
k=1 T effk (i) Xkt − ℓ nit Wit − Oit ≤ 0 for all t = 1, 2, … , T , i = 1, 2, … , I
Skt ≤ dkt for all t = 1, 2, … , T , k = 1, 2, … , K
Oit − nit O max ≤ 0
Wit for all t = 1, 2, … , T , i = 1, 2, … , I
Wit − Wi, t−1 + Hit − Fi, t−1 = 0 for all t = 1, 2, … , T , i = 1, 2, … , I
Ikt − Ik, t+1 + Xkt − Skt = 0 for all t = 1, 2, … , T , k = 1, 2, … , K

Xkt, Skt, Ikt, Hit, Fit, Wit, Oit ≥ 0 for all k = 1, 2, … , K , i = 1, 2, … , I , t = 1, 2, … , T

Note that Wi0 is the number of workers in aggregate Resource Pool i in the beginning of the planning horizon (time t = 0 ) and Fi0 = 0 . Also I k1
and I k ,T +1 are inventory of aggregate Product k in the beginning and at the end of planning horizon, respectively.

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
The above aggregate planning model is essentially the same as the previous model with one aggregate product and one
aggregate resource, except that: (i) each aggregate resource has its own set of available regular time, overtime, and workforce
conservation constraints, and (ii) each aggregate product has its own set of sales and production conservation constraints.

12.7.3. Multi-Period Aggregate Planning with Lost Sales and

Backorders
In some cases when the total demand is higher than supply, then different aggregate products compete for resources, as in
product-mix problem. Hence, it is possible that in some periods, the process is not capable of satisfying the demand for one or
some of the aggregate products. In these situations, the main question is: Which demand should the firm satisfy fully, and
which demand should be partially satisfied? The answer is not easy to find, since it depends on several factors such as net
profit of the products, the costs of unsatisfied demand of each type of product, and the amount of resources needed to produce
each type. However, in this section we show how the analytics for aggregate planning can be extended to incorporate shortage
of products and their corresponding costs. When there is a shortage of a product in a period, the unsatisfied demand in the
period is either lost or backordered.

Incorporating Lost Sales

Lost sales correspond to situations where the customers whose demands are not satisfied in a period are not willing to wait to
receive their products in future periods. They buy the product from competitors. This, in addition to the immediate loss of profit,
has long-term cost of losing customers' future business. The following analytics shows how lost sales can be incorporated to
the aggregate planning models.

ANALYTICS

Multi-Period Aggregate Planning—Incorporating Lost Sales

Consider the multi-period aggregate planning model with I aggregate resources and K aggregate products. Suppose that unsatisfied demands
in a period are lost, and let

( )
c Dk = Cost of losing a unit of demand for Productk

and define decision variable

Ukt = Total lost sales for Product k in Period t

where Ukt ≥ 0 . The multi-period aggregate planning model can be revised as follows to incorporate lost sales:

Objective Function: The following term should be added to the objective function to include lost sales costs:

T K
∑∑
( )
t=1 k=1 c Dk Ukt

Constraints: The sales constraints Skt ≤ dkt must be replaced with the following:

Skt + Ukt = dkt for all t = 1, 2, … , T , k = 1, 2, … , K

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
Note that the new sales constraints in the analytics are straightforward, since they state that the demand for Productk in
Period t (i.e., dkt) is either satisfied by sales in that period (i.e., Skt) or is lost (i.e., Ukt).

Incorporating Backorders

Backorders correspond to the cases where customers whose demands are not satisfied in a period are willing to wait to receive
their products in later periods. This often occurs when the product is popular and the firm has no competition. Hence,
customers do not have a comparable option to the products offered by the firm. Nevertheless, there is a cost associated with
postponing a demand from one period to another. This include costs such as transaction costs, price discount, and cost of free
shipping to convince customers to wait for their orders. Other more long-term cost is the cost of unhappy customers
corresponding to the potential loss of customers' future business.

Similar to the case of lost sales, a simple change in the aggregate planning model can incorporate the case of backorders. The
key is to realize that backordered products are nothing but negative inventory. For example, if the demand in the first period is
for 100 units and only 80 products are produced and sold in that period, then the inventory of products at the end of the first
period is −20, representing a backorder (i.e., shortage) of 20 units. In other words, the total of 20 demands are postponed to
the second period.

Hence, at first glance, it seems that all we need to incorporate backorders in our model is to allow the decision variableIkt (i.e.,
the inventory of Product k in the beginning of Period t) to also take negative values. However, there are two issues with this.
First, the corresponding term in the objective function is −cIIkt representing the holding cost of having (positive) inventory of
Product k. If Ikt becomes negative, then the term −cIIkt becomes positive and more shortages will increase the value of
objective function (i.e., increase profit), which does not make sense. Second, the cost of positive inventory (i.e., holding cost) is
different from the cost of negative inventory (i.e., backorder cost). Therefore, cI in the current aggregate planning model cannot
present both.

These imply that we need to define another decision variable to represent backorders, as illustrated in the following analytics.

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
 ANALYTICS

Multi-Period Aggregate Planning—Incorporating Backorders

Consider the multi-period aggregate planning model with I aggregate resources and K aggregate products. Suppose that unsatisfied demands
in a period are backordered, and let

c+
I
= Cost of holding one unit of (positive) inventory of one product forone period
− =
cI Cost of carrying one unit of backorder demand (i.e., negative inventory) forone period

with their corresponding decision variables

+ =
Ikt Inventory of Productk in the beginning of Period t
− =
Ikt Backorders of Productk in the beginning of Period t

+ ≥ 0 and − ≥ 0 . The multi-period aggregate planning model can be revised as follows to incorporate lost sales:
where I kt Ikt
K
Objective Function: The term −(∑ k=1 c II kt) should be replaced by

K
∑
− ( k=1 c + + + − −)
I I kt c I Ikt

Constraints: The following constraints should be added to the set of constraints:

+− −
Ikt = Ikt Ikt for all k = 1, 2, … , K , t = 1, 2, … , T

+ ≥ 0 and − ≥ 0 .
Also, we should omit the non-negativity constraint I kt ≥ 0 and instead add I kt Ikt

+ and backorder I − are non-negative variables,

Note that in an aggregate planning model with backorders, while inventoryIkt kt
variable Ikt= Ikt+ − Ikt− can either be positive or negative. Also, it is clear that in the optimal solution, the inventory in the
+ = 0 and I − = 0), positive (I + > 0 and I − = 0) , or negative (I + = 0 and
beginning of each period is either zero (Ikt kt kt kt kt
Ikt > 0). In other words, under optimal policy, we have Ikt+ × Ikt− = 0.
−

Finally, notice that Ikt is actually Inventory Position, as discussed in Chap. 9. Inventory position is used to make a decision
whether to place an order (or start production) to increase inventory.

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.