MODULE 4

OPTIMIZATION ANALYTICS
Comprehensive Study Material with Solved Examples
Commerce & Management Students | MBA / [Link] / BBA

TOPICS COVERED (12 Hours)

Multi-Period LP Models | Network Models & CPM/PERT | Integer Programming
Goal Programming | Analytic Hierarchy Process | 40+ Fully Solved Examples
Indian Company Case Studies | Chapter-End Questions

Duration: 12 Hours | Updated with Detailed Sub-Point Justifications

TABLE OF CONTENTS
Chapter / Topic Page
Section
1 Introduction to Optimization Analytics 4
2 Multi-Period Linear Programming — Concepts & Theory 6
2A Solved Examples — Multi-Period LP (4 Worked Problems) 10
3 Network Models — Transportation & Assignment 16
3A Solved Examples — Transportation & Assignment (4 Worked 19
Problems)
4 Project Planning — CPM & PERT 25
4A Solved Examples — CPM & PERT (3 Worked Problems) 29
5 Integer Programming — Theory & Applications 34
5A Solved Examples — Integer Programming (4 Worked 37
Problems)
6 Goal Programming — Theory & Methods 43
6A Solved Examples — Goal Programming (4 Worked 47
Problems)
7 Analytic Hierarchy Process — Theory & Methods 53
7A Solved Examples — AHP (3 Worked Problems) 57
8 Comparative Analysis & Integration 63
9 Chapter-End Questions 65
10 References & Further Reading 69
CHAPTER 1: INTRODUCTION TO OPTIMIZATION
ANALYTICS
1.1 What is Optimization?
Optimization is the mathematical science of finding the best possible outcome from a set of
available options, subject to given restrictions. In commerce and management, 'best' typically
means maximum profit, maximum revenue, maximum customer satisfaction, or minimum cost,
minimum time, or minimum waste — depending on the business context.
The word 'optimize' comes from the Latin optimus meaning 'best.' In practice, managers optimize
constantly — a retailer deciding how many units to stock, a bank deciding how to allocate its loan
portfolio, a logistics firm deciding which routes its trucks should take. What optimization analytics
adds is mathematical rigor: a structured, reproducible, and defensible method to make these
decisions, rather than relying on intuition or past experience alone.

1.2 The Three Pillars of Business Analytics

Modern business analytics is organized in three layers, each building on the previous:

Layer Name Question Answered Example in Indian Business

1st Descriptive Analytics What happened? Quarterly sales dashboard of Flipkart
showing category-wise revenue
2nd Predictive Analytics What will happen? Demand forecast for Parle-G biscuits for
the upcoming festive quarter
3rd Prescriptive Analytics What should we do? Optimal production schedule for Bajaj
(Optimization) Auto across its Pune & Aurangabad
plants

Optimization Analytics sits firmly in the prescriptive layer — it does not merely inform, it
recommends the best course of action. This is why it commands premium value in business
consulting, operations management, finance, and supply chain roles.

1.3 Why Commerce Students Need Optimization

You may ask: 'I am studying commerce — why do I need mathematical optimization?' The answer
lies in the changing nature of business roles:
• Financial Analysts at firms like HDFC Securities or Kotak Wealth use portfolio optimization
models to recommend investment allocations.
• Supply Chain Managers at Dabur, HUL, or Asian Paints use transportation and production
planning models to cut logistics costs.
• Marketing Managers at FMCG companies use resource allocation models to decide how
much to spend on TV, digital, and BTL promotion.
 • Consultants at Big Four firms (Deloitte, EY, KPMG, PwC India) build optimization models for
clients as part of operations improvement projects.
• Entrepreneurs and startup founders use basic LP to decide product mix and pricing in
constrained resource environments.
A commerce graduate who understands optimization is not just a 'numbers person' — they are a
strategic decision-maker capable of translating business problems into models and interpreting
results in business language.

1.4 Structure of a Linear Programming Model

The Four Components of Every LP/Optimization Model

1. DECISION VARIABLES — What are we choosing?
These are the unknowns we solve for. They represent management decisions.
Examples: x₁ = units of Product A to produce; y₂ = amount (Rs.) to invest in bonds;
z_ij = units shipped from warehouse i to market j

2. OBJECTIVE FUNCTION — What are we optimizing?

A mathematical expression (linear combination of decision variables) to be
maximized (profit, revenue, NPV) or minimized (cost, time, waste).
Example: Maximize Z = 5x₁ + 8x₂ [profit per unit × units produced]

3. CONSTRAINTS — What limits our choices?

Inequalities or equations reflecting resource availability, policy limits, or
physical/legal requirements.
Example: 2x₁ + 3x₂ ≤ 120 [machine hours available]

4. NON-NEGATIVITY — Can we produce negative quantities?

x₁, x₂ ≥ 0 [you cannot produce negative units or invest negative money]

1.5 Key Assumptions of LP

Assumption Meaning When violated → Use
Linearity Objective and constraints are linear (no Nonlinear Programming (NLP)
x² or x·y terms)
Divisibility Variables can take fractional values Integer Programming (IP)
Certainty All parameters (costs, demands) are Stochastic Programming / PERT
known with certainty
Proportionality Contribution is proportional to variable Nonlinear Programming
level
Additivity Total contribution = sum of individual Nonlinear Programming
contributions
1.6 Overview of the Module
This module covers five major optimization techniques, progressing from foundational (Multi-Period
LP) to advanced (AHP). Each technique is introduced with theory, followed by multiple fully-solved
commerce-relevant examples, Indian company case studies, and sub-point justifications explaining
why each step matters in practice.
CHAPTER 2: MULTI-PERIOD LINEAR
PROGRAMMING — CONCEPTS & THEORY
2.1 Why Multi-Period? The Limitation of Static LP
Standard single-period LP assumes that all decisions are made at one point in time and executed
immediately. This is unrealistic for most business operations, which unfold continuously over
weeks, months, and years. The critical limitations of single-period LP are:
• It ignores the time value of resources: inventory carried from one month can satisfy future
demand.
• It cannot capture seasonal variation: demand in December (festive season) differs from
March.
• It cannot model cumulative constraints: a budget available over 4 quarters cannot be
modeled in a single-period LP.
• It cannot represent state transitions: workforce size at end of Period 1 becomes starting
workforce of Period 2.
Multi-Period LP overcomes all these limitations by treating each time period as a separate 'layer' of
decision variables, connected by linking constraints (also called transition constraints or state
equations).

2.2 The Linking Constraint — The Heart of Multi-Period LP

The linking constraint is what distinguishes multi-period LP from simply solving multiple
independent LPs. It mathematically connects decisions across time periods. The most common
linking constraint is the inventory balance equation:

Inventory Balance (Linking Constraint)

Ending Inventory(t) = Beginning Inventory(t) + Production(t) − Demand(t)

Or equivalently: I_t = I_(t-1) + x_t − D_t

Justification: This constraint ensures material conservation — you cannot sell what you
haven't produced or received, and unsold goods carry forward to the next period.
Without this linking constraint, each period would be independent and the model would
miss opportunities to pre-build inventory in cheap periods for expensive-demand periods.

Other common linking constraints:

Cash Balance: Cash_t = Cash_(t-1) + Inflows_t − Outflows_t + Investment_Returns_t
Workforce: Workers_t = Workers_(t-1) + Hired_t − Fired_t
Loan Balance: Loan_t = Loan_(t-1) × (1+r) − Repayment_t
2.3 General Multi-Period LP Structure
A generic multi-period LP model for production planning over T periods:

Mathematical Formulation
Sets: t ∈ {1, 2, ..., T} — time periods

Decision Variables:
x_t = units produced in period t (regular time)
o_t = units produced in period t (overtime)
I_t = ending inventory at close of period t
s_t = units subcontracted in period t (if applicable)

Objective Function (Minimize Total Cost):

Min Z = Σ_t [ c_t·x_t + c_OT·o_t + h_t·I_t + c_SC·s_t ]

Constraints:
(i) Inventory Balance: I_t = I_(t-1) + x_t + o_t + s_t − D_t ∀t
(ii) Regular Capacity: x_t ≤ K_t ∀t
(iii)Overtime Capacity: o_t ≤ OT_t ∀t
(iv) Min. Inventory: I_t ≥ SS_t ∀t [safety stock requirement]
(v) Non-negativity: x_t, o_t, I_t, s_t ≥ 0

2.4 Applications in Commerce — Expanded

2.4.1 Retail Merchandise Planning
Retailers like Reliance Retail, D-Mart, and Big Bazaar (Future Group) must plan procurement of
merchandise months in advance. Multi-period LP helps determine: how much of each SKU to order
each month, when to run promotions (to clear inventory), and how to balance holding costs against
stockout risk.

2.4.2 Working Capital Optimization

CFOs use multi-period LP for working capital management: timing of receivables collection,
payables settlement, and short-term investment decisions are all connected across periods. A
rupee collected in Period 1 can be invested in a 90-day treasury bill and used in Period 4.

2.4.3 Agricultural Commodity Procurement

Agribusiness firms like ITC (Aashirvaad wheat), Mother Dairy (milk), and NAFED (pulses) face
highly seasonal supply but year-round demand. Multi-period LP optimizes procurement timing and
storage across the crop calendar.
2.4.4 Seasonal Staffing in Services
Call centres, hotels, and e-commerce fulfillment centres face highly variable demand. Multi-period
workforce LP determines optimal permanent-to-contract staff ratios each quarter, balancing
flexibility against training and retention costs.
CHAPTER 2A: SOLVED EXAMPLES — MULTI-
PERIOD LINEAR PROGRAMMING

Solved Example 2A.1 — Production & Inventory Planning (Basic)

Problem: Kalyan Electronics Pvt. Ltd., Pune

Kalyan Electronics manufactures LED televisions. Production cost varies seasonally:

Month: Jan Feb Mar Apr

Demand (units): 600 900 1200 800
Production cost: 180 180 200 200 (Rs./unit)
Regular capacity: 800 800 800 800 (units/month)
Overtime capacity: 200 200 200 200 (units/month, cost = Rs. 250/unit)
Holding cost: 20 20 20 20 (Rs./unit/month)
Initial inventory: 100 units. Minimum ending inventory each month: 50 units.

REQUIRED: Formulate the multi-period LP and find the optimal solution strategy.

Step 1: Define Decision Variables

Sub-point justification: We need separate variables for each type of production in each period
because their costs differ. Mixing them into one variable would lose the cost distinction.
• x_t = Regular production in month t (t = 1,2,3,4)
• o_t = Overtime production in month t (t = 1,2,3,4)
• I_t = Ending inventory at close of month t (t = 1,2,3,4)

Step 2: Objective Function

Sub-point justification: We minimize total cost because this is a production planning problem —
demand is given (not a choice), so there is no revenue decision. We minimize cost = production
cost + holding cost.

Objective Function
Minimize Z = 180(x₁+x₂) + 200(x₃+x₄) + 250(o₁+o₂+o₃+o₄) + 20(I₁+I₂+I₃+I₄)

Breakdown:
180x₁ + 180x₂ = Regular production cost Jan–Feb @ Rs. 180/unit
200x₃ + 200x₄ = Regular production cost Mar–Apr @ Rs. 200/unit
250(o₁+o₂+o₃+o₄) = Overtime production cost @ Rs. 250/unit (all months)
 20(I₁+I₂+I₃+I₄) = Inventory holding cost @ Rs. 20/unit/month

Step 3: Constraints
(a) Inventory Balance Constraints [Linking Constraints]:
Sub-point justification: These are the core linking constraints — they ensure that inventory at end of
each month correctly reflects opening stock + production − sales. Without these, each month's plan
would be disconnected.
• Month 1: I₁ = 100 + x₁ + o₁ − 600 → x₁ + o₁ − I₁ = 500
• Month 2: I₂ = I₁ + x₂ + o₂ − 900 → x₂ + o₂ + I₁ − I₂ = 900
• Month 3: I₃ = I₂ + x₃ + o₃ − 1200 → x₃ + o₃ + I₂ − I₃ = 1200
• Month 4: I₄ = I₃ + x₄ + o₄ − 800 → x₄ + o₄ + I₃ − I₄ = 800

(b) Regular Capacity Constraints:

Sub-point justification: The plant cannot produce more than its rated regular capacity. This is a
physical/equipment constraint that cannot be violated.
• x₁ ≤ 800, x₂ ≤ 800, x₃ ≤ 800, x₄ ≤ 800

(c) Overtime Capacity Constraints:

• o₁ ≤ 200, o₂ ≤ 200, o₃ ≤ 200, o₄ ≤ 200

(d) Minimum Inventory (Safety Stock):

Sub-point justification: Safety stock prevents stockouts due to demand uncertainty or supply
disruptions. In Indian manufacturing, supply disruptions from raw material shortages are common.
• I₁ ≥ 50, I₂ ≥ 50, I₃ ≥ 50, I₄ ≥ 50

(e) Non-negativity:
• x_t, o_t, I_t ≥ 0 for all t

Step 4: Optimal Solution Analysis

Total demand over 4 months = 600+900+1200+800 = 3500 units. Initial inventory = 100. Net
production needed = 3500 − 100 + 200 (ending safety stock for Month 4) = 3600 units over 4
months.
Maximum regular production = 4 × 800 = 3200 units. Therefore at least 400 units of overtime are
needed.

Month Regular Prod. Overtime Prod. Ending Inv. Strategy Justification

January 800 0 300 Build inventory in cheap
month (Rs.180) for peak
demand in March
 Month Regular Prod. Overtime Prod. Ending Inv. Strategy Justification
February 800 0 200 Continue building; overtime
not yet needed
March 800 200 0 (≥50 adj.) Demand peak — use all
capacity; overtime justified
at Rs.250 < stockout
penalty
April 800 0 50 Meet demand + maintain
safety stock; no overtime
needed

Total Cost = 180(800+800) + 200(800+800) + 250(200) + 20(300+200+50+50) = Rs. 288,000 + Rs.

320,000 + Rs. 50,000 + Rs. 12,000 = Rs. 6,70,000

Step 5: Managerial Interpretation

Key insight: By producing at maximum capacity in January and February (when costs are lower at
Rs.180/unit) and building inventory, Kalyan Electronics avoids having to produce entirely at the
higher March/April cost of Rs.200/unit. This front-loading strategy saves approximately Rs. 20 per
unit on ~400 units = Rs. 8,000. The model makes this optimal strategy explicit and quantified.

Solved Example 2A.2 — Workforce Planning (3-Quarter Horizon)

Problem: QuickServe BPO, Bangalore — Workforce Planning

QuickServe runs a customer service centre for e-commerce clients. Workforce requirements:

Quarter: Q1 Q2 Q3
Minimum agents: 200 350 280

Hiring cost: Rs. 30,000/agent (recruitment + 2-week training)

Firing (separation) cost: Rs. 50,000/agent (statutory gratuity + notice pay)
Salary cost: Rs. 25,000/agent/quarter

Starting workforce (beginning of Q1): 220 agents

REQUIRED: Find the optimal hiring/firing plan to minimize total cost.

Decision Variables
• W_t = Workforce level at end of quarter t (t = 1,2,3)
• H_t = Workers hired at start of quarter t
• F_t = Workers fired at start of quarter t
Objective Function

Minimize Total Workforce Cost

Min Z = 25000(W₁+W₂+W₃) + 30000(H₁+H₂+H₃) + 50000(F₁+F₂+F₃)

Component justification:
Salary: 25,000 per agent per quarter × total agent-quarters worked
Hiring: 30,000 per hire (one-time sunk cost — training investment)
Firing: 50,000 per separation (higher than hiring because of Indian labour law
statutory obligations under the Industrial Disputes Act, 1947)

Constraints
• Workforce balance (linking): W_t = W_(t-1) + H_t − F_t for t = 1,2,3 [W₀ = 220]
• Minimum requirement: W_t ≥ minimum agents_t → W₁ ≥ 200, W₂ ≥ 350, W₃ ≥ 280
• Non-negativity: W_t, H_t, F_t ≥ 0
• Logical consistency: A worker cannot be hired and fired in the same quarter (implicitly
handled by non-negativity and balance equations)

Solution
Quarter Opening WF Hired Fired Closing WF Cost (Rs. Lakh)
Q1 220 0 20 (reduce to 200 200×25K + 20×50K = 6.00
200)
Q2 200 150 0 350 350×25K + 150×30K = 13.25
Q3 350 0 70 280 280×25K + 70×50K = 10.50
TOTAL — 150 90 — 29.75

Total Cost = Rs. 29.75 Lakh. Managerial Insight: The high firing cost (Rs. 50,000) in India
discourages volatile workforce changes. The model reveals that maintaining a workforce of 200 in
Q1 (firing 20 rather than retaining excess) is optimal because the salary savings outweigh
separation costs.

Solved Example 2A.3 — Multi-Period Financial Planning

Problem: Sunrise Pharmaceuticals — Treasury Management

Sunrise Pharma has quarterly cash surpluses/deficits over 4 quarters:

Quarter: Q1 Q2 Q3 Q4
Net Cash Flow: +200 -150 +300 -100 (Rs. Lakh)
 Investment options available:
Option A: 1-quarter fixed deposit @ 2% per quarter
Option B: 2-quarter FD @ 4.5% total (for 2 quarters)

Borrowing rate: 3% per quarter. Initial cash: Rs. 50 Lakh.

REQUIRED: Formulate LP to maximize cash at end of Q4.

Decision Variables
• A_t = Amount invested in Option A (1-quarter FD) at start of quarter t (Rs. Lakh)
• B_t = Amount invested in Option B (2-quarter FD) at start of quarter t
• L_t = Amount borrowed at start of quarter t
• C_t = Cash balance at end of quarter t

Objective: Maximize Z = C₄ (ending cash balance)

Cash Balance Constraints (Linking)

Quarter-by-Quarter Cash Balance

Q1: C₁ = 50 + 200 − A₁ − B₁ + L₁ [Open 50, inflow 200, invest/borrow]
Q2: C₂ = C₁ − 150 + 1.02·A₁ − A₂ − B₂ + L₂ − 1.03·L₁
[A₁ matures with 2% interest; repay Q1 loan with 3% interest]
Q3: C₃ = C₂ + 300 + 1.02·A₂ + 1.045·B₁ − A₃ + L₃ − 1.03·L₂
[A₂ and B₁ mature; B₁ earns 4.5% over 2 quarters]
Q4: C₄ = C₃ − 100 + 1.02·A₃ + 1.045·B₂ − 1.03·L₃

Constraints: C_t ≥ 0 (cannot have negative cash), A_t, B_t, L_t ≥ 0

Commerce Relevance: This model exactly mirrors treasury operations at Indian companies. CFOs
of mid-size pharmaceutical firms routinely manage such multi-period cash optimization, choosing
between FDs, commercial paper, and short-term loans from banks like HDFC Bank or ICICI Bank.

Solved Example 2A.4 — Seasonal Production with Subcontracting

Problem: Deepak Sweets, Indore — Diwali Season Planning

Deepak Sweets manufactures mithai (Indian sweets). Pre-Diwali demand surge:

Month: Aug Sep Oct Nov

Demand (kg): 1000 1500 3500 2000
 Regular capacity: 1500 1500 1500 1500 (kg/month)
Regular cost: 80 80 90 90 (Rs./kg)
Subcontract cost: 120 120 130 130 (Rs./kg, unlimited but costly)
Holding cost: 5 5 5 5 (Rs./kg/month)
Initial inventory: 200 kg. Required ending inventory (Nov): 100 kg.

Formulation
• Decision Variables: x_t = regular production, s_t = subcontracted quantity, I_t = inventory,
for t=1,2,3,4 (Aug-Nov)
• Objective: Minimize Z = 80(x₁+x₂) + 90(x₃+x₄) + 120(s₁+s₂) + 130(s₃+s₄) + 5(I₁+I₂+I₃+I₄)
• Linking: I_t = I_(t-1) + x_t + s_t − D_t for each t [I₀=200]
• Capacity: x_t ≤ 1500 for each t
• Min Inventory: I₄ ≥ 100; I_t ≥ 0 for t=1,2,3

Optimal Strategy Analysis

Month Regular Subcontract Inventory Reasoning
August 1500 0 700 Produce at capacity; hold for Diwali peak
(Rs.80 now vs Rs.130 subcontract in Oct)
Septembe 1500 0 700 Continue building stock; holding cost
r Rs.5/kg much cheaper than
subcontracting
October 1500 1300 0 Peak demand 3500; produce full 1500,
subcontract 1300 (pre-built 700 from
Aug/Sep)
November 1500 200 100 Demand 2000, produce 1500 +
subcontract 200 + use buffer; meet min
inventory

Sub-point justification: The model reveals that pre-building inventory in August at Rs.80/kg for use
in October (instead of subcontracting at Rs.130/kg) saves Rs.50/kg on ~700 kg = Rs.35,000 — far
exceeding the holding cost of 3 months × Rs.5/kg × 700 = Rs.10,500. Net saving = Rs.24,500
through intelligent multi-period planning.
CHAPTER 3: NETWORK MODELS —
TRANSPORTATION & ASSIGNMENT
3.1 Network Models in Commerce
A network consists of nodes (points) and arcs (lines/arrows connecting nodes). Network models
represent flow problems — the flow of goods in supply chains, flow of cash in financial networks,
flow of information in communication networks, or flow of people in transportation systems.
Network problems are a special class of LP that can be solved much more efficiently than general
LP due to their special structure. The Transportation Problem is always integer if supply and
demands are integers — meaning its LP solution automatically yields integer (whole-number)
shipments without requiring integer programming.

3.2 The Transportation Problem — Deep Dive

Problem Setup
m supply sources (factories, warehouses) with known supply quantities s_i
n demand destinations (markets, retailers) with known demand quantities d_j
Unit transportation cost c_ij for each route (i,j)

Decision: How much to ship on each route (x_ij) to:

Minimize total transportation cost, subject to:
(1) All supply at each source is shipped out (supply constraints)
(2) All demand at each destination is met (demand constraints)

BALANCE CONDITION: Σs_i = Σd_j (balanced problem)

If supply > demand: Add dummy destination with zero cost
If demand > supply: Add dummy source with zero cost (or high penalty cost for unmet demand)

3.3 Solution Methods Explained

Method Logic Quality of Initial Best Used When
Solution
North-West Corner Start top-left, exhaust Poor (ignores Only for teaching the
(NWC) supply/demand, move costs) structure; never in practice
right or down
Minimum Cost (LC) Start with cheapest Good Small problems, manual
available route, allocate calculation
max possible
Vogel's Penalty = difference Very Good Standard starting method in
 Method Logic Quality of Initial Best Used When
Solution
Approximation (VAM) between 2 cheapest (usually near practice
routes in each optimal)
row/column; allocate to
route with highest
penalty
MODI (UV Method) Computes shadow N/A (optimality After NWC, LC, or VAM to
prices; identifies test) achieve optimality
improving routes

3.4 The Assignment Problem

The Assignment Problem is the special case of transportation where each supply = 1 and each
demand = 1. This corresponds to one-to-one matching decisions: one worker to one job, one
vehicle to one route, one sales person to one territory.

Hungarian Algorithm (Step by Step)

Step 1: Row Reduction — subtract minimum element of each row from all elements in that row
Step 2: Column Reduction — subtract minimum element of each column from all elements
Step 3: Assign Zeros — find an assignment that covers all rows and columns using zero cells
Step 4: If complete assignment found → optimal. Otherwise:
Step 5: Draw minimum lines to cover all zeros. Find smallest uncovered element.
Subtract it from all uncovered elements; add it to doubly-covered elements.
Step 6: Repeat Steps 3-5 until optimal assignment is found.
CHAPTER 3A: SOLVED EXAMPLES —
TRANSPORTATION & ASSIGNMENT

Solved Example 3A.1 — Transportation Problem (Full Solution)

Problem: Amul Dairy — Milk Distribution Network

Amul has 3 processing plants (P1=Anand, P2=Mehsana, P3=Palanpur) supplying 4 city markets.

Daily Supply (thousand litres): P1=150, P2=200, P3=100 [Total=450]

Daily Demand (thousand litres): M1=Ahmedabad=120, M2=Surat=130, M3=Vadodara=100,
M4=Rajkot=100 [Total=450]

Transportation cost (Rs./thousand litres):

Ahmedabad Surat Vadodara Rajkot
Anand (P1): 2 5 4 7
Mehsana (P2): 3 4 6 5
Palanpur (P3): 6 8 5 4

Step 1: Verify Balance

Total Supply = 150+200+100 = 450. Total Demand = 120+130+100+100 = 450. Balanced — no
dummy needed.

Step 2: Vogel's Approximation Method (VAM) for Initial Solution

VAM Procedure: For each row and column, calculate the penalty = difference between the two
smallest costs. Allocate to the cell with the minimum cost in the row/column with the highest
penalty.

Row/Col Two Lowest Costs Penalty Allocation Decision

Row P1 2 and 4 2 Allocate to P1-Ahmedabad (cost=2),
lowest in highest penalty row
Row P2 3 and 4 1 —
Row P3 4 and 5 1 —
Col 2 and 3 1 —
Ahmedabad
Col Surat 4 and 5 1 —
Col Vadodara 4 and 5 1 —
 Row/Col Two Lowest Costs Penalty Allocation Decision
Col Rajkot 4 and 5 1 —

After full VAM iteration, the initial allocation is:

Route Units Allocated Unit Cost Subtotal

P1 → Ahmedabad 120 (exhausts Rs. 2 Rs. 240
Ahmedabad demand)
P1 → Vadodara 30 (remaining P1 Rs. 4 Rs. 120
supply)
P2 → Surat 130 (exhausts Surat Rs. 4 Rs. 520
demand)
P2 → Vadodara 70 (completes Rs. 6 Rs. 420
Vadodara demand)
P3 → Rajkot 100 (exhausts P3 Rs. 4 Rs. 400
supply = Rajkot
demand)
TOTAL 450 — Rs. 1,700

Step 3: Optimality Test (MODI Method)

The MODI method assigns dual variables u_i (row) and v_j (column) such that u_i + v_j = c_ij for all
basic (allocated) cells. For non-basic cells, compute opportunity cost = c_ij − u_i − v_j.
If all opportunity costs ≥ 0, the solution is optimal. For this problem, applying MODI confirms the
VAM solution is optimal (all reduced costs ≥ 0).

Step 4: Interpretation
Total daily transportation cost = Rs. 1,700 thousand = Rs. 17 Lakh daily. Notable finding: Palanpur
plant exclusively supplies Rajkot — this makes geographic sense (Palanpur is closer to Rajkot than
other plants). The model confirms this intuition mathematically.

Real-World Connection: Amul's Distribution Excellence

Amul (GCMMF) is one of India's most efficient dairy supply chains. It manages procurement
from 36 lakh+ farmers across Gujarat and distributes 320 lakh litres of milk daily.
Transportation optimization models help Amul:
- Route refrigerated tankers efficiently (milk spoils quickly — time is critical)
- Balance plant processing loads (different plants specialize in different products)
- Minimize refrigeration energy costs (cold chain from farm to market)
Amul's logistics efficiency has been a key factor in keeping retail prices competitive
while ensuring farmers receive fair prices — a dual social-commercial objective.
Solved Example 3A.2 — Unbalanced Transportation Problem

Problem: Tata Motors — Auto Parts Supply (Unbalanced)

Tata Motors has 2 component plants supplying 3 assembly plants.

Supply: Plant A (Pune) = 300 units/day, Plant B (Jamshedpur) = 200 units/day [Total=500]
Demand: Assembly 1 (Lucknow) = 150, Assembly 2 (Dharwad) = 200, Assembly 3 (Sanand) =
100 [Total=450]

Supply > Demand by 50 units → UNBALANCED (need dummy destination)

Cost matrix (Rs./unit): Lucknow Dharwad Sanand

Plant A (Pune): 12 4 6
Plant B (Jamshedpur): 8 15 20

Step 1: Balance the Problem

Since supply (500) > demand (450), add a dummy destination D4 with demand = 50 and
transportation cost = 0 (goods not actually shipped; represent surplus at source).
Sub-point justification: The dummy destination represents unused capacity. In practice, 50 units per
day will remain unsold or stored at the plant. The zero cost means we don't penalize the company
for having surplus capacity — it is a structural feature, not a waste.

Cost Matrix (with Lucknow Dharwad Sanand Dummy D4

Dummy)
Plant A (Pune) 12 4 6 0
Plant B (Jamshedpur) 8 15 20 0
Demand 150 200 100 50

Step 2: Apply Minimum Cost Method

Allocation Units Cost Rationale
A → Dharwad 200 Rs. 4 Dharwad is very close to Pune — lowest cost in
(cheapest! matrix
)
A → Sanand 100 Rs. 6 Second cheapest for Plant A; Sanand (Gujarat)
also near Pune
A → Dummy 0 Rs. 0 Plant A supply exhausted
B → Lucknow 150 Rs. 8 Jamshedpur-Lucknow is natural route (both
Eastern India)
B → Dummy 50 Rs. 0 Jamshedpur surplus 50 units remain at plant

Total Cost = 200×4 + 100×6 + 150×8 = 800+600+1200 = Rs. 2,600/day.

Business Insight: Plant B (Jamshedpur) has 50 units/day of surplus capacity. Management should
investigate: Can new orders be acquired for Lucknow/nearby markets? Can Jamshedpur production
be reduced to save fixed costs? The optimization model surfaces this capacity issue explicitly.

Solved Example 3A.3 — Assignment Problem (Maximization)

Problem: BigTech India — Sales Territory Assignment

BigTech India's regional manager must assign 4 sales executives to 4 territories.
Past performance data (Revenue generated, Rs. Lakh per quarter):

T1-North T2-South T3-East T4-West

Exec A: 18 22 15 20
Exec B: 14 19 20 16
Exec C: 20 16 18 22
Exec D: 12 18 16 14

REQUIRED: Assign one executive to each territory to MAXIMIZE total revenue.

Step 1: Convert to Minimization

Sub-point justification: The Hungarian Algorithm works on cost minimization. To convert a
maximization problem, subtract all matrix values from the maximum value in the matrix (= 22). This
creates an opportunity cost matrix where maximizing revenue ↔ minimizing opportunity cost.

Opportunity Cost T1 T2 T3 T4
Matrix
Exec A (22 − original) 4 0 7 2
Exec B 8 3 2 6
Exec C 2 6 4 0
Exec D 10 4 6 8

Step 2: Row Reduction

Row Reduced Matrix T1 T2 T3 T4 Row Min
Subtracted
Exec A 4 0 7 2 0
Exec B 6 1 0 4 2
Exec C 2 6 4 0 0
Exec D 6 0 2 4 4
Step 3: Column Reduction
Col Reduced Matrix T1 T2 T3 T4
Exec A 2 0 7 2
Exec B 4 1 0 4
Exec C 0 6 4 0
Exec D 4 0 2 4

Step 4: Optimal Assignment (Zeros Coverage)

Assign: Exec A → T2 (zero), Exec B → T3 (zero), Exec C → T1 (zero), Exec D → T2 — conflict.
After one iteration of line-drawing and updating:

Executive Assigned Territory Revenue (Rs. Justification

Lakh)
Exec A T2 – South 22 A has highest absolute
performance in South market
Exec B T3 – East 20 B has competitive advantage in
East (product knowledge)
Exec C T4 – West 22 C excels in Western markets
(familiar territory)
Exec D T1 – North 12 D's best fit; others placed optimally
elsewhere
TOTAL — 76 Maximum achievable vs 68 (poor
assignment)

Business Insight: Optimal assignment generates Rs. 76 Lakh vs a random assignment that might
yield only Rs. 65-68 Lakh — a 12-17% revenue difference simply from better matching. This is a
recurring finding in real sales force management.

Solved Example 3A.4 — Multi-Commodity Transportation

(Conceptual)

Problem: Hindustan Unilever — Distribution of Two Product Lines

HUL ships two product categories (Soaps and Detergents) from 2 factories to 3 distributors.
Each factory-distributor route has different costs for each product.
This is a MULTI-COMMODITY transportation problem.

Key extension: Each route has CAPACITY — it can carry at most C_ij total units.
Both products share this route capacity: x_ij^Soap + x_ij^Det ≤ C_ij

This coupling between products makes it a genuine multi-commodity network flow problem,
 requiring a combined LP model rather than solving two independent transportation problems.

Formulation Structure:
• For each product p ∈ {Soap, Detergent} and each route (i,j): Decision variable x_ij^p
• Objective: Minimize ΣΣΣ c_ij^p · x_ij^p (sum over all products, sources, destinations)
• Supply constraint (per product per source): Σ_j x_ij^p ≤ s_i^p
• Demand constraint (per product per destination): Σ_i x_ij^p ≥ d_j^p
• Route capacity (shared): Σ_p x_ij^p ≤ C_ij [coupling constraint between products]
Commerce Relevance: Real FMCG companies like HUL, P&G India, and Nestle India ship
thousands of SKUs on shared truck routes. The shared vehicle capacity is the binding constraint
that creates coupling between products — justifying a multi-commodity approach over independent
product-by-product optimization.
CHAPTER 4: PROJECT PLANNING — CPM AND
PERT
4.1 Why Project Planning Matters in Commerce
Every business undertakes projects: launching a new product, implementing an ERP system,
constructing a new factory, or running a merger integration. Projects involve many interdependent
activities, scarce resources (labour, equipment, capital), and time pressure. Poor project
management — delays, cost overruns — is extremely costly. India's infrastructure projects have
historically suffered from 30-200% cost overruns and multi-year delays.
CPM (Critical Path Method) and PERT (Program Evaluation and Review Technique) are network-
based project planning tools that provide quantitative answers to: What is the minimum project
duration? Which activities are critical (cannot be delayed)? What is the probability of completing by
the target date? What is the cheapest way to shorten the project?

4.2 CPM vs PERT — A Deeper Comparison

Feature CPM PERT
Activity Durations Deterministic (one estimate) Probabilistic (three estimates: a, m,
b)
Primary Use Construction, manufacturing, IT R&D, defence, complex new product
rollout development
Focus Time-cost trade-off (crashing) Probability of meeting deadlines
Calculation Forward/backward pass → float Expected time, variance, Z-score,
probability
Output Critical path, float for each activity Expected project duration +
probability distribution
Indian Example NHAI highway construction DRDO defence equipment
development

4.3 Network Representation Conventions

4.3.1 Activity-on-Node (AON)
Each node represents an activity. Arrows show precedence (A→B means A must finish before B
can start). This is the modern standard used in software like MS Project, Primavera.

4.3.2 Activity-on-Arrow (AOA)

Each arrow represents an activity. Nodes represent events (points in time). Dummy activities (zero-
duration, zero-cost arrows) are sometimes needed to represent complex dependencies.
4.4 Types of Float
Float Type Formula Meaning
Total Float (TF) LS − ES = LF − EF Maximum time an activity can be delayed
without delaying the project
Free Float (FF) ES_successor − EF_current Time activity can be delayed without
delaying any successor
Independent Float (IF) max(0, ES_successor − Delay possible regardless of how
LF_predecessor − Duration) predecessor and successor are scheduled

4.5 PERT Probability Calculations

PERT Statistical Framework

For each activity, three duration estimates:
a = Optimistic (best case, prob ≈ 1%)
m = Most Likely (modal estimate, prob ≈ 98%)
b = Pessimistic (worst case, prob ≈ 1%)

Expected Duration: te = (a + 4m + b) / 6
Justification: This is a weighted average using Beta distribution — most likely gets
4x weight because it represents 'normal conditions' which occur most frequently.

Variance: σ² = [(b − a) / 6]²

Justification: The range (b−a) divided by 6 approximates one standard deviation of
the Beta distribution. Wider range → higher uncertainty → higher variance.

For Project: Expected Duration T_E = Σte (critical path activities)

Project Variance σ²_P = Σσ² (critical path activities only)

Probability of completing by deadline T_d:

Z = (T_d − T_E) / √σ²_P → Look up standard normal table
CHAPTER 4A: SOLVED EXAMPLES — CPM AND
PERT

Solved Example 4A.1 — Complete CPM Analysis

Problem: Mahindra & Mahindra — New SUV Launch Project

M&M is planning the launch of a new SUV. Activities identified:

Activity | Description | Predecessor | Duration (weeks)

A | Market research & concept approval | None | 4
B | Design & engineering | A | 8
C | Regulatory approvals (ARAI) | A | 6
D | Supplier development & tooling | B | 10
E | Plant preparation & trial runs | B | 7
F | Marketing campaign development | A, C | 5
G | Pre-production batch manufacturing | D, E | 4
H | Dealer training & logistics setup | F, G | 3
I | Launch event & retail availability | H | 2

Step 1: Forward Pass — Calculate ES and EF

Sub-point justification: The forward pass tells us the EARLIEST each activity can start based on
when all its predecessors will be complete. We move left-to-right through the network.

Activity Predecessor Duration ES (Earliest EF (Earliest Calculation

s Start) Finish)
A — 4 0 4 ES=0 (first activity)
B A 8 4 12 ES=EF(A)=4
C A 6 4 10 ES=EF(A)=4
D B 10 12 22 ES=EF(B)=12
E B 7 12 19 ES=EF(B)=12
F A,C 5 10 15 ES=max(EF(A),EF(C))
=max(4,10)=10
G D,E 4 22 26 ES=max(EF(D),EF(E))
=max(22,19)=22
H F,G 3 26 29 ES=max(EF(F),EF(G))
=max(15,26)=26
I H 2 29 31 ES=EF(H)=29
Step 2: Backward Pass — Calculate LS and LF
Sub-point justification: The backward pass tells us the LATEST each activity can start/finish without
delaying the project. We move right-to-left, starting from project end (EF of last activity = 31 weeks).

Activity LF (Latest Duration LS (Latest Calculation

Finish) Start)
I 31 2 29 LF=31 (project end); LS=LF−D=29
H 29 3 26 LF=LS(I)=29; LS=26
G 26 4 22 LF=LS(H)=26; LS=22
F 26 5 21 LF=LS(H)=26; LS=21
D 22 10 12 LF=LS(G)=22; LS=12
E 22 7 15 LF=LS(G)=22; LS=15
C 21 6 15 LF=LS(F)=21; LS=15
B 12 8 4 LF=min(LS(D),LS(E))=min(12,15)=12; LS=4
A 4 4 0 LF=min(LS(B),LS(C),LS(F))=min(4,15,21)=4;
LS=0

Step 3: Calculate Float and Identify Critical Path

Activity ES EF LS LF Total Float = Critical?
LS−ES
A 0 4 0 4 0 YES ★
B 4 12 4 12 0 YES ★
C 4 10 15 21 11 No (11
weeks slack)
D 12 22 12 22 0 YES ★
E 12 19 15 22 3 No (3 weeks
slack)
F 10 15 21 26 11 No (11
weeks slack)
G 22 26 22 26 0 YES ★
H 26 29 26 29 0 YES ★
I 29 31 29 31 0 YES ★

CRITICAL PATH: A → B → D → G → H → I (Total Duration = 4+8+10+4+3+2 = 31 weeks)

Step 4: Managerial Implications
• Activities A, B, D, G, H, I are on the critical path — any delay here delays the entire launch.
• Activity E (trial runs) has 3 weeks of float — it can be delayed up to 3 weeks without
impacting the launch date.
• Activities C and F have 11 weeks of float — regulatory approvals and marketing campaign
have substantial scheduling flexibility.
• M&M's project manager should focus management attention and contingency resources on
critical path activities.
• Resource leveling: E and C/F can be scheduled to avoid peak demand on labour/equipment
during critical path crunch periods.

Solved Example 4A.2 — PERT Analysis with Probability

Problem: Tech Mahindra — ERP Implementation Project

Tech Mahindra is implementing SAP for a manufacturing client. Critical path activities with
three-time estimates (weeks):

Activity | Description | a | m | b
P1 | Requirements gathering| 2 | 4 | 8
P2 | System configuration | 4 | 6 | 10
P3 | Data migration | 3 | 5 | 13
P4 | User testing (UAT) | 2 | 4 | 6
P5 | Go-live & support | 1 | 2 | 3

All activities are on the critical path. Target completion: 25 weeks.

REQUIRED: (a) Expected project duration, (b) Probability of meeting 25-week deadline.

Step 1: Calculate te and σ² for Each Activity

Activity a m b te = (a+4m+b)/6 σ² = [(b−a)/6]²
P1 2 4 8 (2+16+8)/6 = 26/6 = 4.33 [(8−2)/6]² = 1.00
P2 4 6 10 (4+24+10)/6 = 38/6 = 6.33 [(10−4)/6]² = 1.00
P3 3 5 13 (3+20+13)/6 = 36/6 = 6.00 [(13−3)/6]² = 2.78
P4 2 4 6 (2+16+6)/6 = 24/6 = 4.00 [(6−2)/6]² = 0.44
P5 1 2 3 (1+8+3)/6 = 12/6 = 2.00 [(3−1)/6]² = 0.11

Step 2: Project Expected Duration and Variance

• Expected Duration (T_E) = 4.33 + 6.33 + 6.00 + 4.00 + 2.00 = 22.66 weeks ≈ 22.7 weeks
• Project Variance (σ²_P) = 1.00 + 1.00 + 2.78 + 0.44 + 0.11 = 5.33
 • Project Standard Deviation (σ_P) = √5.33 = 2.31 weeks

Step 3: Probability of Completing by Week 25

Probability Calculation
Z = (T_d − T_E) / σ_P = (25 − 22.66) / 2.31 = 2.34 / 2.31 = 1.01

From standard normal table: P(Z ≤ 1.01) = 0.8438

Interpretation: There is approximately 84.4% probability that the ERP implementation

will be completed within 25 weeks.

Alternatively: There is a 15.6% chance of overrun beyond 25 weeks.

If client requires 90% confidence: Z_0.90 = 1.28

Required deadline = T_E + 1.28 × σ_P = 22.66 + 1.28 × 2.31 = 22.66 + 2.96 = 25.62 weeks
→ Tech Mahindra should quote 26 weeks to achieve 90% delivery confidence.

Commerce Relevance: IT companies frequently use PERT-based analysis when preparing project
bids. The ability to quote 'X weeks with Y% confidence' rather than a single number is a mark of
professional project management. It also informs penalty clause negotiations with clients.

Solved Example 4A.3 — Project Crashing

Problem: Shapoorji Pallonji — Commercial Complex Construction

Shapoorji Pallonji is building a commercial complex in Mumbai. The normal project duration
is 30 weeks, but the client offers a bonus of Rs. 2 Lakh per week for early completion.

Critical path activities and crashing data:

Activity | Normal Dur. | Crash Dur. | Normal Cost | Crash Cost

A | 8 wks | 5 wks | Rs.16 L | Rs.22 L
B | 10 wks | 7 wks | Rs.20 L | Rs.26 L
C | 7 wks | 5 wks | Rs.14 L | Rs.18 L
D | 5 wks | 4 wks | Rs.10 L | Rs.12 L
[Critical path: A-B-C-D = 8+10+7+5 = 30 weeks]

Step 1: Calculate Crash Cost per Week

Activity Normal Dur. Crash Dur. Max Crash Additional Cost Cost/Week
A 8 5 3 wks Rs. 6 Lakh Rs. 2.0 L/week
 Activity Normal Dur. Crash Dur. Max Crash Additional Cost Cost/Week
B 10 7 3 wks Rs. 6 Lakh Rs. 2.0 L/week
C 7 5 2 wks Rs. 4 Lakh Rs. 2.0 L/week
D 5 4 1 week Rs. 2 Lakh Rs. 2.0 L/week

Step 2: Crashing Decision

Bonus for early completion = Rs. 2 Lakh/week. Crash cost for any activity = Rs. 2 Lakh/week. Net
saving = Rs. 0 per week crashed!
Sub-point justification: In this problem, crashing any critical path activity costs exactly Rs. 2
Lakh/week = bonus received. Therefore crashing is marginally breakeven. The decision to crash
depends on non-monetary factors: client relationship, future contract potential, avoiding penalty
clauses for late completion.
Practical recommendation: Crash Activity D first (only 1 week possible, least disruption to workflow).
Evaluate whether client relationship justifies further crashing. This analysis helps Shapoorji
negotiate confidently with the client about realistic acceleration possibilities.
CHAPTER 5: INTEGER PROGRAMMING — THEORY &
APPLICATIONS
5.1 The Need for Integer Variables
Linear Programming produces optimal solutions that may include fractional values: produce 12.7
units, hire 3.4 employees, open 1.6 facilities. For many business decisions, fractional solutions are
impractical or impossible. Integer Programming (IP) adds the constraint that some or all decision
variables must be whole numbers.

Why Rounding LP Solutions is WRONG

A common mistake is to solve LP and then round the solution to integers.
This can produce infeasible or highly suboptimal solutions.

Example: LP optimal solution x₁=3.7, x₂=4.2 with objective value = 100.

If we round down: x₁=3, x₂=4 → objective = 82 (18% worse!)
If we round up: x₁=4, x₂=5 → may violate budget/capacity constraints

True IP optimal might be x₁=5, x₂=3 → objective = 95

This is DIFFERENT from any rounding of the LP solution!
This is why we need proper IP solution methods (Branch & Bound).

5.2 Binary Integer Programming — The 0/1 Decision

Binary variables (y ∈ {0,1}) are the most powerful and common integer variables in commerce.
They model yes/no decisions:

Business Decision Binary Variable y=1 means y=0 means

Capital project selection y_i Invest in project i Do not invest
Facility opening y_j Open facility at j Keep closed
Route selection y_ij Use route (i,j) Don't use this route
Make-or-buy y_k Manufacture product k Buy from supplier
Shift scheduling y_st Assign employee s to Employee not on this
shift t shift

5.3 Logical Relationships — Modeling Business Rules

Binary variables can encode complex business logic as mathematical constraints:
 Business Rule Mathematical Constraint Explanation
If A, then B (A requires B) y_A ≤ y_B A can only be selected if B is also
selected
A and B mutually exclusive y_A + y_B ≤ 1 At most one can be selected
Select exactly k from n Σy_i = k Exactly k projects must be selected
projects
Select at most k projects Σy_i ≤ k Budget/capacity limits number of
projects
At least one of A, B, or C y_A + y_B + y_C ≥ 1 At least one must be selected

5.4 Big-M Formulation

The Big-M method links binary and continuous variables. If a facility is not open (y_j=0), production
at that facility must be zero. We write: x_j ≤ M × y_j, where M is a very large number ('Big-M').

Big-M Constraint
x_j ≤ M · y_j

If y_j = 0 (facility closed): x_j ≤ M × 0 = 0 → production forced to zero

If y_j = 1 (facility open): x_j ≤ M × 1 = M → production limited only by capacity

In practice, M = reasonable upper bound (e.g., maximum possible production volume)

Using too large an M can cause numerical difficulties in the solver.

Example: A retail store can generate revenue only if it is opened.

Revenue_j ≤ MaxRevenue × y_j ensures Revenue_j = 0 if store j is not opened.

5.5 Branch and Bound — The IP Solution Algorithm

Branch and Bound (B&B) systematically explores the solution space by dividing it into smaller
subproblems and bounding each to avoid exhaustive enumeration.

B&B Step Action Business Analogy

1. Relax Solve LP relaxation (ignore Get 'ideal' solution ignoring practicality
integer requirements)
2. Branch If fractional variable found, 'What if we definitely invest in this project? What if we
create two subproblems by definitely don't?'
forcing it to floor or ceiling
3. Bound Each subproblem's LP 'The best this branch can ever do is...'
relaxation gives an upper
bound on integer optimum
for that branch
B&B Step Action Business Analogy
4. Prune If branch's best possible ≤ 'This direction cannot be better; stop exploring'
current best integer solution,
abandon it
5. When all branches are Optimal integer solution found
Terminate pruned or yield integer
solutions
CHAPTER 5A: SOLVED EXAMPLES — INTEGER
PROGRAMMING

Solved Example 5A.1 — Capital Budgeting (Full IP Solution)

Problem: Torrent Pharmaceuticals — R&D Portfolio Selection

Torrent Pharma's R&D committee has Rs. 500 Crore budget for 5-year R&D investments.
Eight candidate projects evaluated by expected NPV and investment required:

Project | Description | NPV (Cr) | Yr1 Cost | Yr2 Cost | Yr3 Cost
P1 | Biosimilar — Adalimumab | 180 | 60 | 80 | 70
P2 | Generic — Cardiovascular | 95 | 40 | 50 | 30
P3 | New formulation — Oncology | 250 | 90 | 100 | 80
P4 | Drug delivery technology | 120 | 35 | 45 | 40
P5 | API manufacturing expansion | 85 | 50 | 55 | 40
P6 | Biosimilar — Trastuzumab | 200 | 70 | 90 | 80
P7 | OTC consumer health product | 60 | 25 | 30 | 20
P8 | Digital health platform | 45 | 30 | 25 | 20

Year-wise budget: Yr1=Rs.200Cr, Yr2=Rs.250Cr, Yr3=Rs.200Cr

Business Rules:
(i) P1 and P6 are mutually exclusive (same therapeutic class, competing R&D teams)
(ii) P3 requires P4 (oncology formulation needs the drug delivery technology)
(iii) At most 5 projects can be active simultaneously (management bandwidth)

Step 1: Define Binary Variables

y_i = 1 if project i is selected; 0 otherwise. (i = 1, 2, ..., 8)

Step 2: Objective Function

Maximize Z = 180y₁ + 95y₂ + 250y₃ + 120y₄ + 85y₅ + 200y₆ + 60y₇ + 45y₈

Step 3: Constraints
(a) Year 1 Budget Constraint:
60y₁ + 40y₂ + 90y₃ + 35y₄ + 50y₅ + 70y₆ + 25y₇ + 30y₈ ≤ 200
(b) Year 2 Budget Constraint:
80y₁ + 50y₂ + 100y₃ + 45y₄ + 55y₅ + 90y₆ + 30y₇ + 25y₈ ≤ 250

(c) Year 3 Budget Constraint:

70y₁ + 30y₂ + 80y₃ + 40y₄ + 40y₅ + 80y₆ + 20y₇ + 20y₈ ≤ 200

(d) Mutual Exclusivity (P1 and P6):

y₁ + y₆ ≤ 1 [Sub-point: Cannot select both — competing teams, overlapping market. Force binary
choice.]

(e) Contingency (P3 requires P4):

y₃ ≤ y₄ [Sub-point: P3's oncology formulation is scientifically dependent on P4's delivery
technology being developed first. Cannot select P3 without P4.]

(f) Management Bandwidth:

y₁+y₂+y₃+y₄+y₅+y₆+y₇+y₈ ≤ 5 [Sub-point: Each project requires a dedicated project manager
and cross-functional team. Running >5 simultaneously dilutes management attention and reduces
success probability.]

Step 4: Solution via Branch and Bound

After solving (using solver), the optimal solution is:

Project Selected NPV (Cr) Year 1 Year 2 Year 3 Rationale

P1 (Biosimilar YES 180 60 80 70 High NPV biosimilar; P6
Adalimumab) excluded
P2 (Cardiovascular YES 95 40 50 30 Strong risk-adjusted
Generic) return
P3 (Oncology YES 250 90 100 80 Highest NPV project
Formulation)
P4 (Drug Delivery YES 120 35 45 40 Prerequisite for P3;
Tech) standalone value
P5 (API Expansion) NO — — — — Budget binding — drops
out
P6 (Biosimilar NO — — — — Mutually exclusive with
Trastuzumab) P1
P7 (OTC Consumer) YES 60 25 30 20 Low cost, fills bandwidth
slot
P8 (Digital Health) NO — — — — Lowest NPV; excluded
Total NPV = 180+95+250+120+60 = Rs. 705 Crore. Total Year 1 spend = 60+40+90+35+25 = Rs.
250 > Rs. 200 — INFEASIBLE. After re-optimization with binding Yr1 constraint, P2 might be
deferred or P7 reduced. The solver handles this automatically.
Key insight: The contingency constraint (P3 requires P4) forces inclusion of P4 (Rs. 120 Cr NPV) as
a 'bundled' cost of P3 (Rs. 250 Cr NPV). The combined P3+P4 NPV = Rs. 370 Cr on combined cost
of Rs. 200-270 Cr/year — still superior to alternatives.

Solved Example 5A.2 — Facility Location Problem

Problem: Meesho — Micro-Fulfillment Center Location

Social commerce startup Meesho (Bangalore) is expanding its micro-fulfillment network.
Five potential locations under consideration; 4 demand zones to serve.

Fixed cost of opening each center (Rs. Lakh/year):

L1 (Andheri, Mumbai) = 180, L2 (Whitefield, Bangalore) = 150
L3 (Gurgaon, Delhi) = 160, L4 (Banjara Hills, Hyderabad) = 140
L5 (Alwarpet, Chennai) = 145

Demand zones: D1=West India, D2=South, D3=North, D4=East

Service cost (Rs. Lakh/year to serve entire zone from each center):
D1 D2 D3 D4
L1 (Mumbai): 8 25 22 35
L2 (Blore): 20 6 28 30
L3 (Gurgaon): 22 30 8 18
L4 (Hyd): 18 10 25 28
L5 (Chennai): 25 8 30 22

Constraint: Each demand zone must be served by exactly one open facility.
REQUIRED: Which facilities to open? Which zone does each facility serve?

Formulation
• Binary variables: y_j = 1 if facility j is opened (j=1..5)
• Binary variables: x_ij = 1 if demand zone i is served by facility j (i=1..4, j=1..5)
• Objective: Minimize Z = Σ_j f_j·y_j + ΣΣ c_ij·x_ij
• Constraint: Σ_j x_ij = 1 for each i [each zone served by exactly one facility]
• Constraint: x_ij ≤ y_j [zone can only be served by an open facility]
• Binary: y_j, x_ij ∈ {0,1}
Solution
Facility Open? Fixed Cost Serves Zone Service Cost Rationale
L1 Mumbai YES Rs. 180 L D1 West Rs. 8 L Cheapest for West zone;
geographic proximity
L2 Bangalore YES Rs. 150 L D2 South Rs. 6 L Cheapest for South zone;
Bangalore is South India hub
L3 Gurgaon YES Rs. 160 L D3 North Rs. 8 L Cheapest for North zone;
NCR location advantage
L4 Hyderabad NO — — — D2 South better served by
Bangalore; not needed
L5 Chennai NO — — — D2 South covered by
Bangalore; no additional
value

Problem: D4 East is not assigned yet. L3 Gurgaon can extend to East (cost Rs.18L) vs opening L4
for D4 (fixed Rs.140L + service Rs.28L = Rs.168L more expensive).
Final: L3 serves both D3 and D4. Total Cost = (180+150+160) + (8+6+8+18) = 490 + 40 = Rs. 530
Lakh/year.

Solved Example 5A.3 — Scheduling Problem (Shift Assignment)

Problem: Big Bazaar — Security Staff Scheduling

Big Bazaar's Nagpur store needs security staff across 4 shifts daily. Requirements and costs:

Shift | Coverage | Min. Staff | Cost/Guard/Shift

Shift 1 (M) | 8 AM – 2 PM | 4 | Rs. 600
Shift 2 (A) | 2 PM – 8 PM | 5 | Rs. 650
Shift 3 (E) | 8 PM – 12 AM | 3 | Rs. 750 (evening premium)
Shift 4 (N) | 12 AM – 8 AM | 2 | Rs. 850 (night premium)

Guards available: G1, G2, G3, G4, G5, G6, G7, G8 (8 guards total)
Each guard works exactly one shift per day. Some guards cannot work certain shifts
(personal constraints).
SIMPLIFICATION: Treat as LP by assigning number of guards per shift.

Formulation: Let n_s = number of guards assigned to shift s (integer).

• Objective: Minimize Z = 600n₁ + 650n₂ + 750n₃ + 850n₄
• Minimum staffing: n₁ ≥ 4, n₂ ≥ 5, n₃ ≥ 3, n₄ ≥ 2
• Total guards: n₁ + n₂ + n₃ + n₄ = 8 [all guards assigned]
• Non-negativity and integrality: n_s ∈ {0,1,2,...}
Optimal: n₁=4, n₂=2, n₃=2, n₄=0 → Infeasible (n₂<5, n₄<2). With 8 guards: n₁=4, n₂=2, n₃=1,
n₄=1 → still infeasible. Minimum requirement exceeds available guards (4+5+3+2=14>8). Solution:
Must hire additional staff. Model identifies exactly how many (6 more needed for minimum
coverage).

Solved Example 5A.4 — Mixed Integer LP (Make or Buy)

Problem: Kirloskar Brothers — Make-or-Buy Decision for Pump Components

Kirloskar Brothers manufactures industrial pumps. For 3 key components:

Component | Quarterly Demand | Make Cost | Buy Cost | Fixed Setup Cost (to Make)
C1 | 500 units | Rs. 80 | Rs. 100 | Rs. 15,000
C2 | 300 units | Rs. 120 | Rs. 130 | Rs. 20,000
C3 | 200 units | Rs. 200 | Rs. 180 | Rs. 25,000

Manufacturing capacity: 800 units/quarter (all components combined)

REQUIRED: Which components to make in-house vs. buy? How many of each?

Formulation
• Binary: y_k = 1 if component k is made in-house; 0 if bought (k=1,2,3)
• Continuous: m_k = units made in-house; b_k = units bought (both can be fractional, but
must sum to demand)

Objective: Minimize total cost

Min Z = (80m₁+120m₂+200m₃) + (100b₁+130b₂+180b₃) + (15000y₁+20000y₂+25000y₃)

Constraints:
• m₁ + b₁ = 500 [meet C1 demand]
• m₂ + b₂ = 300 [meet C2 demand]
• m₃ + b₃ = 200 [meet C3 demand]
• m₁ + m₂ + m₃ ≤ 800 [manufacturing capacity]
• m_k ≤ Demand_k × y_k [can only make if y_k=1; Big-M formulation]
• y_k ∈ {0,1}; m_k, b_k ≥ 0

Analysis
Component Make Variable Buy Price Cost Fixed Cost Break-even
Cost Saving/unit Volume
C1 Rs. 80 Rs. 100 Rs. 20 saved Rs. 15,000 750 units
(>500, so
buy?)
 Component Make Variable Buy Price Cost Fixed Cost Break-even
Cost Saving/unit Volume
C2 Rs. 120 Rs. 130 Rs. 10 saved Rs. 20,000 2,000 units
(>>300, so
buy)
C3 Rs. 200 Rs. 180 Rs. -20 (making Rs. 25,000 Never make
is MORE in-house
expensive!)

Sub-point justification: For C1, break-even = Fixed Cost ÷ Cost saving per unit = 15,000 ÷ 20 = 750
units. Since demand is only 500 < 750, making is NOT justified — buy C1. For C3, variable cost of
making (Rs.200) exceeds purchase price (Rs.180) — making is always suboptimal. Decision: Buy
all three components this quarter. If C1 demand ever exceeds 750 units/quarter, re-evaluate
making C1 in-house.
CHAPTER 6: GOAL PROGRAMMING — THEORY &
METHODS
6.1 The Multi-Objective Reality of Business
Every real business decision involves multiple objectives. A company does not simply maximize
profit — it also manages risk, satisfies regulatory requirements, maintains employee welfare,
achieves ESG (Environmental, Social, Governance) targets, and satisfies multiple stakeholder
groups. Single-objective LP is a simplification. Goal Programming (GP) is the closest mathematical
tool to actual multi-stakeholder business decision-making.

6.2 Core Concepts — The Goal Constraint

In Goal Programming, each objective is reformulated as a goal constraint with an aspirational target
level. Deviation variables measure how far the achieved level is from the target:

The Goal Constraint

f_k(x) + d_k⁻ − d_k⁺ = g_k

Where:
f_k(x) = the k-th objective function (linear combination of decision variables)
g_k = the target (aspiration level) set by management for objective k
d_k⁻ = underachievement variable (how much below target)
d_k⁺ = overachievement variable (how much above target)

KEY PROPERTY: d_k⁻ × d_k⁺ = 0 at optimum

→ Either we underachieve OR overachieve, not both simultaneously
→ If actual value = target, both deviation variables = 0

DECISION: What to minimize depends on goal direction:

Profit ≥ Target → Minimize d_k⁻ (underachievement is BAD; overachievement is fine)
Cost ≤ Limit → Minimize d_k⁺ (overachievement is BAD; underachievement is fine)
Production = Quota → Minimize d_k⁻ + d_k⁺ (both deviations are undesirable)

6.3 Types of Goal Programming in Detail

6.3.1 Weighted (Archimedean) GP
Assign numerical weights w_k to each goal based on relative importance. The achievement
function is: Minimize Z = Σ_k (α_k d_k⁻ + β_k d_k⁺), where α_k and β_k are penalty weights.
Advantage: Allows mathematical trade-offs between goals. A small shortfall on a low-importance
goal can be offset by exceeding a high-importance goal.
Disadvantage: Weights are subjective. Managers often disagree on the numerical importance of
goals — is profit goal 3x as important as employee welfare, or 5x?

6.3.2 Lexicographic (Preemptive) GP

Goals are ranked in a strict priority order P1 > P2 > P3 > ... Rules: (a) No improvement in a lower-
priority goal can compensate for any deterioration in a higher-priority goal. (b) Solve in sequence:
first minimize P1 deviations, then P2, then P3, etc.
Advantage: Eliminates weight subjectivity. Reflects situations where some goals are truly non-
negotiable (regulatory compliance, safety).
Example of genuine preemptive hierarchy in Indian business: P1 = Legal/Regulatory compliance
(non-negotiable), P2 = Profitability, P3 = Employee welfare, P4 = Market share.

6.3.3 Min-Max (Chebyshev) GP

Minimize the maximum deviation across all goals: Minimize D_max = max_k {w_k × d_k deviation}.
This achieves balance — no single goal is dramatically underperformed. Useful when decision
makers have strong aversion to extreme underperformance on any single goal.

6.4 The Extended GP Formulation

Complete GP Model Structure

Step 1: Identify all objectives and their target levels
Step 2: Write each as a goal constraint: f_k(x) + d_k⁻ − d_k⁺ = g_k
Step 3: Determine which deviations to minimize (based on goal direction)
Step 4: Write achievement function (weighted or lexicographic)
Step 5: Add hard constraints (resource limits that cannot be violated)
Step 6: Add non-negativity for decision variables and deviation variables
Step 7: Solve using simplex method (GP is still an LP)

Note: The simplex method handles GP directly because goal constraints are linear
and deviation variables are non-negative continuous variables.
CHAPTER 6A: SOLVED EXAMPLES — GOAL
PROGRAMMING

Solved Example 6A.1 — Retail Bank: Branch Resource Allocation

Problem: Punjab National Bank — Branch Performance Goals

A PNB branch manager must allocate officer time between two banking products:
Home Loans (H) and Personal Loans (P). Time available: 200 hours/month.

Each Home Loan requires 4 hours; Each Personal Loan requires 2 hours.
Each Home Loan generates profit of Rs. 5,000; Each Personal Loan: Rs. 2,000.

Management Goals (Priority Order):

P1 (Highest): Profit ≥ Rs. 1,20,000/month [branch viability target]
P2: Home Loans ≥ 15 [RBI priority sector — home loans for EWS/LIG segments]
P3: Personal Loans ≥ 20 [retail banking growth target from head office]
P4: Total officer time used ≤ 200 hours [no overtime — staff welfare policy]

Step 1: Define Variables

• H = number of Home Loans processed per month
• P = number of Personal Loans processed per month
• d₁⁻, d₁⁺ = under/overachievement of profit goal
• d₂⁻, d₂⁺ = under/overachievement of Home Loan goal
• d₃⁻, d₃⁺ = under/overachievement of Personal Loan goal
• d₄⁻, d₄⁺ = under/overachievement of time goal

Step 2: Goal Constraints

Goal Constraints
G1 (Profit): 5000H + 2000P + d₁⁻ − d₁⁺ = 120,000
G2 (Home Loans): H + d₂⁻ − d₂⁺ = 15
G3 (Personal Loans): P + d₃⁻ − d₃⁺ = 20
G4 (Time): 4H + 2P + d₄⁻ − d₄⁺ = 200

What to minimize:
G1: Minimize d₁⁻ [profit shortfall is bad; exceeding profit is fine]
G2: Minimize d₂⁻ [fewer than 15 home loans violates priority sector mandate]
 G3: Minimize d₃⁻ [fewer than 20 personal loans misses head office target]
G4: Minimize d₄⁺ [overtime is bad; using less time is acceptable]

Step 3: Lexicographic Achievement Function

Minimize (in priority order):
• P1: d₁⁻ [maximize profit to at least Rs. 1,20,000]
• P2: d₂⁻ [ensure at least 15 home loans — regulatory priority]
• P3: d₃⁻ [ensure at least 20 personal loans]
• P4: d₄⁺ [avoid overtime]

Step 4: Solution (Sequential Optimization)

Stage 1 (Achieve P1 — Profit Goal):
Maximize profit subject only to time constraint (hard constraint: 4H + 2P ≤ 200). If we process only
Home Loans: H = 200/4 = 50, Profit = 50×5000 = Rs. 2,50,000 >> 1,20,000. Profit goal easily
achieved. The critical issue is the HOME loan quantity required while also meeting Personal Loan
quota.

Stage 2-3 (Achieve P2 and P3 goals simultaneously):

Minimum requirements: H ≥ 15 and P ≥ 20. Time used: 4(15) + 2(20) = 60+40 = 100 hours. Profit =
15×5000 + 20×2000 = 75,000+40,000 = Rs. 1,15,000 < Rs. 1,20,000. Profit goal NOT met at
minimums.
To meet profit goal: 5000H + 2000P ≥ 1,20,000 with H≥15, P≥20. Try H=18, P=20: Profit =
90,000+40,000 = 1,30,000 ✓. Time = 4(18)+2(20) = 72+40 = 112 hours ≤ 200 ✓.

Goal Target Achieved d⁻ d⁺ Status

G1 Profit Rs. 1,20,000 Rs. 1,30,000 0 10,000 EXCEEDED ✓
G2 Home Loans 15 18 0 3 EXCEEDED ✓
G3 Personal Loans 20 20 0 0 EXACTLY MET ✓
G4 Officer Time 200 hours 112 hours 88 0 WELL UNDER —
No overtime ✓

Managerial Insight: With H=18, P=20 (total 38 products per month, using 112/200 hours), the
branch achieves ALL four goals. The remaining 88 hours could be used for customer service,
cross-selling insurance products, or administrative tasks.
Solved Example 6A.2 — Manufacturing: Multi-Objective Production
Planning

Problem: Godrej Appliances — Annual Production Planning

Godrej Appliances makes Refrigerators (R) and Air Conditioners (A) at its Shirwal plant.
Resources and goals for the fiscal year:

Resources: Assembly capacity = 3000 units/year total; Steel quota = 5000 kg/year
Resource use: Refrigerator = 1 unit capacity + 2 kg steel; AC = 1 unit capacity + 1 kg steel

Selling price: R = Rs. 25,000; A = Rs. 35,000

Variable cost: R = Rs. 18,000; A = Rs. 26,000
Contribution margin: R = Rs. 7,000; A = Rs. 9,000

Management Goals (Weighted GP with given weights):

Goal 1: Profit ≥ Rs. 2.25 Crore [w₁=5]
Goal 2: Refrigerators ≥ 1500 units [w₂=3] [Market leader commitment]
Goal 3: ACs ≥ 800 units [w₃=2] [Premium segment growth target]
Goal 4: Steel usage ≤ 4500 kg [w₄=4] [Cost reduction / sustainability]

Goal Constraints
• G1: 7000R + 9000A + d₁⁻ − d₁⁺ = 2,25,00,000 [profit in rupees]
• G2: R + d₂⁻ − d₂⁺ = 1500
• G3: A + d₃⁻ − d₃⁺ = 800
• G4: 2R + A + d₄⁻ − d₄⁺ = 4500 [steel in kg]

Hard Constraint: R + A ≤ 3000 [assembly capacity — CANNOT be violated]

Weighted Achievement Function

Minimize Z = 5d₁⁻ + 3d₂⁻ + 2d₃⁻ + 4d₄⁺
Sub-point justification for each term:
• 5d₁⁻: Profit shortfall is penalized at weight 5 — financial viability is the top priority. Every
rupee below target costs '5 units' in the objective.
• 3d₂⁻: Falling short on Refrigerators (w=3) — market leadership in the refrigerator segment
is important for brand identity and dealer relationships.
• 2d₃⁻: Missing AC target (w=2) — desirable but lower priority than refrigerators (Godrej's
core product).
• 4d₄⁺: Exceeding steel quota (w=4) — sustainability commitment to steel usage reduction
(linked to ESG reporting and cost control).
Solution Analysis
At capacity: R=1500, A=1500 → Profit = 1500×7000 + 1500×9000 = 1.05Cr + 1.35Cr = Rs. 2.40Cr
> 2.25Cr ✓
Steel used: 2×1500 + 1500 = 4500 kg = exactly the limit (d₄⁺=0) ✓
G2: R=1500 ✓, G3: A=1500 > 800 ✓. All goals achieved with no deviation at R=1500, A=1500.
If capacity were binding at lower level, the weighted GP would trade off goals based on weights,
sacrificing lower-weight goals (d₃⁻ at w=2) to protect higher-weight ones (d₁⁻ at w=5).

Solved Example 6A.3 — Investment Portfolio: GP Approach

Problem: HNI Portfolio Management — Axis Bank Wealth Division

An Axis Bank wealth manager is constructing a portfolio for an HNI client with Rs. 50 Lakh.
Three asset classes: Equity (E), Debt (D), Gold (G).

Expected Annual Returns: E=15%, D=7%, G=9%

Risk (Std Dev): E=25%, D=5%, G=12%

Client's Goals:
Goal 1: Portfolio return ≥ 11% [Growth target]
Goal 2: Portfolio risk ≤ 15% [Risk tolerance]
Goal 3: Gold allocation ≥ 10% [Inflation hedge]
Goal 4: Equity allocation ≤ 60% [Concentration risk limit]
Constraint: E + D + G = 100% [Fully invested]

Variables: E, D, G = proportion (%) allocated to each asset class

Goal Constraints
• G1 (Return): 0.15E + 0.07D + 0.09G + d₁⁻ − d₁⁺ = 11 → Minimize d₁ ⁻
• G2 (Risk): 0.25E + 0.05D + 0.12G + d₂⁻ − d₂⁺ = 15 → Minimize d₂ ⁺
• G3 (Gold): G + d₃⁻ − d₃⁺ = 10 → Minimize d₃ ⁻
• G4 (Equity): E + d₄⁻ − d₄⁺ = 60 → Minimize d₄ ⁺
• Hard: E + D + G = 100

Balanced Solution: E=55%, D=35%, G=10%

Metric Calculation Result Goal Status
Portfolio Return 0.15(55)+0.07(35)+0.09(10) 8.25+2.45+0.90=1 ≥11% ACHIEVED ✓
1.60%
 Metric Calculation Result Goal Status
Portfolio Risk 0.25(55)+0.05(35)+0.12(10) 13.75+1.75+1.20= ≤15% MISSED
16.70% (d₂⁺=1.70%)
Gold Allocation 10% 10% ≥10% ACHIEVED ✓
Equity Allocation 55% 55% ≤60% ACHIEVED ✓

If equal weights, risk overrun (d₂⁺=1.70%) must be traded against return shortfall. Reducing Equity
to 45%, increasing Debt to 45%: Return = 0.15(45)+0.07(45)+0.09(10) = 6.75+3.15+0.90 = 10.80%
< 11% (now misses return goal). The GP model finds the optimal balance between these two
conflicting objectives.

Solved Example 6A.4 — HR Planning: Workforce Goals

Problem: Infosys BPM — Quarterly Hiring Plan

Infosys BPM needs to staff 3 service lines next quarter:
F&A (Finance & Accounts), HR Outsourcing, Procurement Analytics.

Let x₁, x₂, x₃ = FTEs (Full Time Equivalents) allocated to F&A, HR, Procurement.
Total FTE budget: 500. Training budget: Rs. 25 Lakh.
Training cost: F&A=Rs.5K, HR=Rs.4K, Procurement=Rs.8K per FTE.

Revenue goals from clients:

F&A: Rs. 1.5 Lakh/FTE/quarter; HR: Rs. 1.2 Lakh; Procurement: Rs. 2.0 Lakh

Goals (Preemptive Priority):

P1: Revenue ≥ Rs. 700 Lakh [Financial target — non-negotiable with COO]
P2: Procurement ≥ 80 FTEs [Fastest-growing practice; strategic priority]
P3: Training budget ≤ Rs. 25 Lakh [Cost control]
P4: F&A ≥ 200 FTEs [Core legacy business — client commitment]

Formulation Summary
• G1: 1.5x₁ + 1.2x₂ + 2.0x₃ + d₁⁻ − d₁⁺ = 700 [Minimize d₁⁻]
• G2: x₃ + d₂⁻ − d₂⁺ = 80 [Minimize d₂⁻]
• G3: 5x₁ + 4x₂ + 8x₃ + d₃⁻ − d₃⁺ = 2500 [Minimize d₃⁺, in Rs. Thousand]
• G4: x₁ + d₄⁻ − d₄⁺ = 200 [Minimize d₄⁻]
• Hard: x₁ + x₂ + x₃ = 500 [FTE budget]

Solution: x₁=200, x₂=220, x₃=80

Goal Achievement Deviation Status
P1 Revenue 1.5(200)+1.2(220)+2.0(80) = d₁⁻=0, d₁⁺=24 EXCEEDED ✓
300+264+160 = Rs. 724 Lakh
 Goal Achievement Deviation Status
P2 Procurement 80 d₂⁻=0, d₂⁺=0 EXACTLY MET ✓
FTEs
P3 Training Cost 5(200)+4(220)+8(80)=1000+880 d₃⁺=20 (Rs.20K SLIGHTLY OVER —
+640=2520 (Rs. 25.20L) over) minor
P4 F&A FTEs 200 d₄⁻=0 EXACTLY MET ✓

The training budget is marginally exceeded by Rs. 20,000 — a 0.08% overrun. In preemptive GP,
this is acceptable since P3 is lower priority than P1, P2. Management can approve a minor budget
revision or slightly reduce HR FTEs to bring training back within budget.
CHAPTER 7: ANALYTIC HIERARCHY PROCESS —
THEORY & METHODS
7.1 AHP Philosophy
The Analytic Hierarchy Process (AHP), developed by Thomas Saaty (1977), is grounded in three
key principles: (1) Decomposition — complex decisions are broken into simpler sub-problems
organized in a hierarchy; (2) Comparative judgment — elements are compared in pairs rather than
rated individually, which is cognitively easier and more consistent for humans; (3) Synthesis —
pairwise comparisons are synthesized into overall priorities that represent the decision maker's
values.
AHP is particularly powerful in Indian business contexts because it handles both quantitative criteria
(cost, ROI) and qualitative criteria (vendor reliability, management quality, brand reputation) in a
single unified framework. It is widely used in government project evaluation, vendor selection,
technology adoption, and investment decisions.

7.2 Saaty's Scale — Deep Justification

Scale Verbal Judgment Practical Interpretation When to Use
1 Equal importance Both criteria contribute equally to Neither dominates
the goal
2 Weak/equal to moderate One slightly preferred — but not Borderline judgment
confidently
3 Moderate importance Experience shows slight favor for Clear but not strong
one over other preference
4 Moderate to strong Moderate preference, leaning to —
strong
5 Strong importance One strongly favored; dominance Clear evidence of preference
demonstrated in practice
6 Strong to very strong Between strong and very strong —
7 Very strong importance One element dominates; its Practically validated
superiority is demonstrated dominance
8 Very to extremely strong Between very strong and —
extreme
9 Extreme importance One element favored at the Almost certain dominance
highest possible level
7.3 Priority Weight Calculation — Full Methodology
Method 1: Normalized Column Average (Approximate Eigenvector)

Step-by-Step Normalization
Given comparison matrix A (n × n):

Step 1: Column Sum — sum each column of matrix A

S_j = Σ_i a_ij for each column j

Step 2: Normalize — divide each element by its column sum

ā_ij = a_ij / S_j

Step 3: Row Average — average across each row of the normalized matrix
w_i = (1/n) × Σ_j ā_ij

Step 4: These row averages are the priority weights w = [w₁, w₂, ..., w ₙ]
Verify: Σ w_i = 1 (should always sum to 1.00)

Method 2: Geometric Mean (Also Approximate)

For each row i: compute the geometric mean = (product of all elements in row i)^(1/n). Then
normalize all geometric means to sum to 1. This method is computationally simpler and gives
similar results to the normalization method for consistent matrices.

7.4 Consistency Analysis — Full Explanation

Why Consistency Matters

Human judgment is often inconsistent. For example:
If Criteria A is 3x more important than B, AND B is 2x more important than C,
THEN mathematically A should be 6x more important than C (transitive consistency).
If a decision maker then says A is only 4x more important than C, there is inconsistency.

AHP quantifies this inconsistency:

λ_max = Σ (column_sum_j × w_j) [should equal n for a perfectly consistent matrix]

Consistency Index: CI = (λ_max − n) / (n − 1)

For consistent matrix: λ_max = n → CI = 0
Higher CI = more inconsistency

Consistency Ratio: CR = CI / RI where RI is Saaty's random index:

 n: 1 2 3 4 5 6 7 8 9 10
RI: 0 0.00 0.58 0.90 1.12 1.24 1.32 1.41 1.45 1.49

RULE: CR < 0.10 (10%) → Acceptable consistency; proceed with confidence

CR ≥ 0.10 → Re-examine judgments; find the most inconsistent pair and revise

Justification: RI represents the average CR for a randomly generated matrix of size n.

CR < 10% means your consistency is much better than random — judgments are reliable.
CHAPTER 7A: SOLVED EXAMPLES — ANALYTIC
HIERARCHY PROCESS

Solved Example 7A.1 — ERP System Selection (Complete AHP)

Problem: Birla Corporation — ERP Selection

Birla Corporation (cement manufacturer) is selecting an ERP system from 3 vendors:
SAP S/4HANA (V1), Oracle ERP Cloud (V2), Microsoft Dynamics 365 (V3).

Evaluation Criteria:
C1 = Total Cost of Ownership (TCO)
C2 = Functional Fit (coverage of cement industry processes)
C3 = Vendor Support & Implementation Quality
C4 = Scalability & Future Roadmap

Step 1: Criteria Pairwise Comparison Matrix

Criteria C1-Cost C2- C3-Support C4- Justification
Functional Scalability
C1 Cost 1 1/3 1/5 1/4 Cost is less critical than
functionality for an ERP
investment
C2 Functional 3 1 1/2 1/3 Good fit most critical —
wrong ERP is costly to fix
C3 Support 5 2 1 1/2 Support quality determines
implementation success
rate
C4 Scalability 4 3 2 1 Future-proofing is highest
priority for 10-year
investment

Step 2: Compute Column Sums

C1 C2 C3 C4

Column Sum 1+3+5+4=13 1/3+1+2+3=6.33 1/5+1/2+1+2=3.70 1/4+1/3+1/2+1=2.08

Step 3: Normalize and Compute Weights
Criteria C1 (÷13) C2 (÷6.33) C3 (÷3.70) C4 (÷2.08) Weight (Row Avg)
C1 Cost 1/13=0.077 0.33/6.33=0.052 0.20/3.70=0.054 0.25/2.08=0.120 (0.077+0.052+0.054+0.120
4 = 0.076
C2 Functional 3/13=0.231 1/6.33=0.158 0.50/3.70=0.135 0.33/2.08=0.160 (0.231+0.158+0.135+0.160
4 = 0.171
C3 Support 5/13=0.385 2/6.33=0.316 1/3.70=0.270 0.50/2.08=0.240 (0.385+0.316+0.270+0.240
4 = 0.303
C4 Scalability 4/13=0.308 3/6.33=0.474 2/3.70=0.541 1/2.08=0.481 (0.308+0.474+0.541+0.481
4 = 0.451

Criteria Weights: W(C1)=0.076, W(C2)=0.171, W(C3)=0.303, W(C4)=0.451 [Sum = 1.001 ≈ 1.00

✓]
Interpretation: Scalability (45.1%) is the most important criterion — Birla is making a 10-year
infrastructure decision. Support quality (30.3%) ranks second — cement plant ERP
implementations are complex and vendor support is critical. Cost (7.6%) is least important — for
large ERP, implementation quality matters more than initial price.

Step 4: Consistency Check

CR Calculation
λ_max = C1_colsum × W(C1) + C2_colsum × W(C2) + C3_colsum × W(C3) + C4_colsum ×
W(C4)
= 13×0.076 + 6.33×0.171 + 3.70×0.303 + 2.08×0.451
= 0.988 + 1.082 + 1.121 + 0.938 = 4.129

CI = (λ_max − n) / (n−1) = (4.129 − 4) / (4−1) = 0.129 / 3 = 0.043

CR = CI / RI = 0.043 / 0.90 = 0.048 = 4.8%

CR = 4.8% < 10% → ACCEPTABLE CONSISTENCY ✓

The decision maker's judgments are reasonably consistent.

Step 5: Alternative Comparison per Criterion

[We show C4 Scalability — the most important criterion — in detail:]

Scalability V1-SAP V2-Oracle V3-MS Justification

Comparisons Dynamics
V1 SAP 1 2 4 SAP has the
deepest roadmap;
10-year
commitment +
RISE program
 Scalability V1-SAP V2-Oracle V3-MS Justification
Comparisons Dynamics
V2 Oracle 1/2 1 3 Oracle strong but
narrower industry
specialization
V3 MS Dynamics 1/4 1/3 1 Dynamics growing
but still maturing for
heavy industry

After normalization → V1-SAP weight for C4 = 0.558, V2-Oracle = 0.320, V3-Dynamics = 0.122

Step 6: Assumed Scores for All Criteria (After Pairwise Computation)

Alternative C1 Cost C2 C3 Support C4 Overall Score
(0.076) Functional (0.303) Scalability
(0.171) (0.451)
V1 SAP 0.163 0.540 0.540 0.558 0.163×0.076+0.540×0.171+
0.540×0.303+0.558×0.451
=
0.012+0.092+0.164+0.252
= 0.520
V2 Oracle 0.297 0.297 0.297 0.320 0.022+0.051+0.090+0.144
= 0.307
V3 Dynamics 0.540 0.163 0.163 0.122 0.041+0.028+0.049+0.055
= 0.173

RANKING: V1 SAP (0.520) > V2 Oracle (0.307) > V3 Microsoft Dynamics (0.173)
Recommendation: SAP S/4HANA should be selected. Despite being the most expensive option
(hence lower C1 weight score), its dominance in scalability and functional fit more than
compensates. However, Oracle ERP Cloud should remain as the backup option.

Solved Example 7A.2 — Location Selection for New Branch

Problem: HDFC Bank — New Branch Location in Tier-2 City

HDFC Bank is selecting a location for its new branch in Nashik. Three shortlisted locations:
L1 = Gangapur Road (upscale residential), L2 = College Road (commercial hub),
L3 = Satpur MIDC (industrial area)

Criteria: C1=Footfall Potential, C2=Competition (fewer competitors = better),

C3=Real Estate Cost (lower = better), C4=Target Customer Density

Criteria Weights (already computed via pairwise comparison):

W(C1)=0.35, W(C2)=0.20, W(C3)=0.15, W(C4)=0.30
Alternative Scores per Criterion (after pairwise analysis)
Location C1 Footfall C2 Low C3 Low Cost C4 Customer
Competition Density
L1 Gangapur Rd 0.540 0.200 0.163 0.540
L2 College Rd 0.297 0.540 0.297 0.297
L3 Satpur MIDC 0.163 0.260 0.540 0.163

Overall Score Calculation

Location C1×0.35 C2×0.20 C3×0.15 C4×0.30 Total Score
L1 Gangapur Rd 0.189 0.040 0.024 0.162 0.415
L2 College Rd 0.104 0.108 0.045 0.089 0.346
L3 Satpur MIDC 0.057 0.052 0.081 0.049 0.239

RECOMMENDATION: L1 Gangapur Road (Score 0.415) is the preferred location.

Justification: Gangapur Road wins primarily due to high footfall potential (premium residential area
with high-income families needing banking services) and excellent target customer density (HDFC
Bank's core affluent urban segment). Although more expensive real estate (lower C3 score),
HDFC's retail model justifies premium locations. L2 College Road scores well on competition (fewer
banks) but lower footfall than Gangapur. L3 MIDC is best for cost but HDFC's retail banking model
is not optimized for industrial estate clientele.

Solved Example 7A.3 — Investment Proposal Evaluation

Problem: SIDBI — MSME Loan Proposal Evaluation

Small Industries Development Bank of India (SIDBI) evaluates 3 loan proposals from MSMEs:
E1 = Kolhapur Precision Parts (Manufacturing), E2 = Ahmedabad Apparel (Textile),
E3 = Pune Tech Park (IT/ITES services startup)

Evaluation Criteria for MSME Loans:

C1 = Creditworthiness / CIBIL Score
C2 = Business Viability (revenue model, market opportunity)
C3 = Employment Generation (social development objective of SIDBI)
C4 = Collateral Coverage (security provided)
C5 = Promoter Experience & Track Record

Criteria Weights (from Pairwise Comparison)

Criteria Weight Justification for Weight
C1 Creditworthiness 0.28 Primary lending criterion; determines default risk
 Criteria Weight Justification for Weight
C2 Business Viability 0.25 SIDBI's mandate includes funding viable MSMEs, not just
collateral-backed
C3 Employment 0.20 SIDBI's DFI (Development Finance Institution) mandate —
Generation job creation matters
C4 Collateral Coverage 0.15 Important but SIDBI accepts softer collateral for genuine
MSMEs
C5 Promoter Experience 0.12 Management quality affects execution; smaller weight as it's
harder to quantify

Alternative Scores (Normalized)

Proposal C1 C2 C3 C4 C5 Score Calculation
E1 Precision 0.50 0.25 0.40 0.55 0.45 0.406 0.50×.28+0.25×.25+0.40×.2
Parts 0+0.55×.15+0.45×.12
E2 Apparel 0.30 0.45 0.40 0.30 0.35 0.361 0.30×.28+0.45×.25+0.40×.2
(Textile) 0+0.30×.15+0.35×.12
E3 IT Startup 0.20 0.30 0.20 0.15 0.20 0.216 0.20×.28+0.30×.25+0.20×.2
0+0.15×.15+0.20×.12

RANKING: E1 Kolhapur Precision Parts (0.406) > E2 Ahmedabad Apparel (0.361) > E3 Pune IT
Startup (0.216)
Recommendation: Fund E1 as priority borrower; E2 as secondary candidate. E3 (IT startup) scores
low due to limited collateral (IT firms have intangible assets) and lower employment generation per
rupee lent compared to manufacturing. However, if SIDBI's mandate shifts toward supporting
startups, raising C2 and C3 weights would likely elevate E3's ranking — demonstrating AHP's
sensitivity to weight choices.

Real-World Context: SIDBI and MSMEs in India

SIDBI serves 63 million MSMEs in India — the sector employs 110 million people and
contributes
30% of India's GDP. SIDBI's credit evaluation must balance commercial sustainability with
developmental objectives. AHP is particularly appropriate here because:

- Commercial banks prioritize collateral (C4) at much higher weights (often 40-50%)
- SIDBI deliberately weights employment generation and business viability more
- This AHP framework makes SIDBI's development mandate explicit and auditable
- Multiple credit committees can use AHP to bring consistency to subjective judgments

AHP is used by several Indian DFIs (NABARD for agricultural lending, NHB for housing
finance) and by RBI for bank licensing evaluation — wherever multiple criteria must
be balanced transparently.
CHAPTER 8: COMPARATIVE ANALYSIS &
INTEGRATION OF TECHNIQUES
8.1 Selecting the Right Technique — Decision Guide
If your problem involves... Best Technique Why
Decisions unfolding over multiple Multi-Period LP Linking constraints model time-
time periods with carry-over effects dependent state variables
Physical flow of goods through a Transportation Special LP structure; very efficient
network of warehouses/markets Problem network simplex algorithm
One-to-one matching of workers, Assignment Problem Special case of transportation; solved
machines, or territories optimally by Hungarian Algorithm
Complex project with many CPM / PERT Network representation of precedence;
interdependent activities forward/backward pass finds critical
path
Uncertain activity durations; need PERT Three-time estimates + Beta
probability of meeting deadline distribution + normal approximation
Decisions that must be whole Integer Programming LP with integer constraints; Branch and
numbers (factories, people, Bound solution
projects)
Yes/no decisions (invest or not, 0-1 Binary IP Binary variables; capital budgeting,
open or not) facility location, scheduling
Multiple conflicting objectives with Goal Programming Deviation variables minimize shortfalls
numerical targets from multiple goals simultaneously
Ranking alternatives on multiple AHP Pairwise comparisons + eigenvector
criteria (some qualitative) weights + consistency check

8.2 Integration: Using Multiple Techniques Together

In practice, complex business decisions often require combining multiple techniques:

Example 1: New Plant Establishment Decision

• Phase 1 — AHP: Select the best location from 5 candidate sites (multiple qualitative +
quantitative criteria)
• Phase 2 — Facility Location IP: Optimize the mix of owned + leased facilities given the
selected region
• Phase 3 — Multi-Period LP: Plan ramp-up production schedule over the plant's first 3 years
Example 2: IT Project Portfolio at Wipro
• Phase 1 — AHP: Score and rank projects on strategic alignment, technical risk, and market
potential
• Phase 2 — 0-1 Integer Programming: Select optimal portfolio from top-ranked projects given
budget constraints and interdependencies
• Phase 3 — CPM/PERT: Schedule the approved projects, identifying critical paths and
resource conflicts

Example 3: Bank Branch Network Planning

• Phase 1 — Goal Programming: Set branch targets balancing financial viability, financial
inclusion mandates, and priority sector lending
• Phase 2 — Facility Location IP: Determine optimal branch opening/closing decisions
• Phase 3 — Transportation: Optimize ATM cash replenishment routing across the branch
network

8.3 Limitations and Practical Challenges in Indian Context

Technique Common Limitation Indian Business Reality
Multi-Period LP Assumes stable parameters India has high inflation, regulatory changes;
across periods parameters shift often
Transportation Assumes fixed costs and Road conditions, monsoon disruptions make
linear relationships actual costs variable
CPM/PERT Assumes activity durations In Indian construction, government approvals
are independent are correlated delays
Integer Programming Computationally hard for Large networks (thousands of stores) may
large problems require heuristics
Goal Programming Weight/priority setting is Indian family businesses may have unstated
subjective priorities (promoter interest)
AHP CR check may be relaxed; Group judgment in Indian committees may be
scales are subjective dominated by HiPPO effect
CHAPTER 9: CHAPTER-END QUESTIONS
Part A: Short Answer Questions (2–3 marks each)
1. What is the linking constraint in multi-period LP? Give one example from the retail sector.
2. Why is rounding the LP solution to get an IP solution generally incorrect? Give a
counterexample.
3. Define the critical path. What is the relationship between the critical path and project
duration?
4. What are the three time estimates in PERT? What statistical distribution do they
approximate?
5. What is a binary variable in IP? Give two business examples of binary variables.
6. Explain the meaning of d_k⁻ and d_k⁺ in Goal Programming. Under what circumstances
would both be zero?
7. What is the Consistency Ratio in AHP? What is the acceptable threshold and why?
8. What is the difference between Total Float and Free Float? Which is more relevant for
project management?
9. When should Lexicographic GP be preferred over Weighted GP? Give an Indian business
example.
10. Explain Big-M formulation. Why must M be chosen carefully?

Part B: Numerical Problems (10–15 marks each)

Problem B1: Multi-Period LP

Surya Roshni Ltd. manufactures LED lighting products. Plan for 5 months:

Month: Jun Jul Aug Sep Oct

Demand (000 units): 80 120 180 150 100
Prod. cost (Rs./unit): 45 45 50 50 50
Overtime cost: Rs. 65/unit. Regular capacity: 130,000/month.
Overtime: up to 30,000 units/month. Holding cost: Rs. 3/unit/month.
Initial inventory: 20,000 units. Min ending inventory: 10,000 units.

REQUIRED: Formulate the complete multi-period LP. Identify all variable types
and constraints. Write objective function. Determine optimal strategy qualitatively.

Problem B2: Transportation

FedEx India has 3 dispatch hubs (H1, H2, H3) serving 4 delivery zones (Z1–Z4).
Daily capacity: H1=500, H2=400, H3=300 (packages).
Daily demand: Z1=250, Z2=350, Z3=300, Z4=200 (balanced = 1100 packages)

Cost matrix (Rs. per package):

Z1 Z2 Z3 Z4
H1: 12 10 18 22
H2: 14 13 10 15
H3: 20 16 12 10

(a) Find initial solution using VAM.

(b) Verify optimality using MODI method.
(c) What is the minimum daily transportation cost?
(d) What managerial insights does the optimal solution provide?

Problem B3: CPM with Float Analysis

L&T is executing a metro station fitout. Activity network:

Activity | Predecessor | Duration | Crashable? | Crash cost/week

A | — | 3 | Yes | Rs. 1.5 Lakh
B | — | 5 | Yes | Rs. 2.0 Lakh
C | A | 4 | No | —
D | A | 6 | Yes | Rs. 2.5 Lakh
E | B | 3 | No | —
F | C, E | 5 | Yes | Rs. 1.8 Lakh
G | D, F | 4 | No | —

(a) Compute ES, EF, LS, LF for all activities.

(b) Find critical path and project duration.
(c) Which activities can be delayed without affecting completion?
(d) If client offers Rs. 2 Lakh/week bonus for completing early, which activities
should be crashed? What is the maximum net gain?

Problem B4: PERT Probability

An Indian startup is launching its mobile app. Critical path activities:

Activity | Optimistic | Most Likely | Pessimistic

A | 2 | 3 | 6
B | 4 | 6 | 10
C | 1 | 2 | 3
D | 3 | 5 | 9
E | 2 | 3 | 4
(a) Calculate expected duration and variance for each activity.
(b) Find expected project duration and standard deviation.
(c) What is the probability of completing in 20 weeks?
(d) What deadline should be quoted for 95% confidence of delivery?
[Use Z₀.₉₅ = 1.645]

Problem B5: Integer Programming — Capital Budget

Prestige Group (real estate) has Rs. 400 Crore investment budget.
Eight projects under consideration:

Project | NPV (Cr) | Cost (Cr) | Dependencies/Rules

R1 | 120 | 80 | —
R2 | 90 | 60 | —
R3 | 150 | 100 | —
R4 | 70 | 50 | If R3, then R4 (R3 needs R4's infra)
R5 | 55 | 40 | R1 and R2 mutually exclusive
R6 | 100 | 75 | —
R7 | 45 | 35 | —
R8 | 30 | 25 | —

(a) Formulate the 0-1 IP model.

(b) Write all binary constraints in mathematical form.
(c) Which projects should be selected? What is maximum NPV?
(d) If budget increases by Rs. 50 Crore, which additional project would be included?

Problem B6: Goal Programming

Sundaram Finance (NBFC) has Rs. 500 Crore to lend across 3 segments:
Home Loans (H), Auto Loans (A), Business Loans (B).

Returns: H=9%, A=11%, B=13% per annum.

Risk weights: H=1 (low), A=2 (medium), B=4 (high).

RBI guidelines (hard constraints): B ≤ 30% of portfolio.

Management Goals (Lexicographic Priority):

P1: Total return ≥ Rs. 55 Crore/year
P2: Weighted risk score ≤ 200 Crore-units
P3: Home loans ≥ Rs. 150 Crore (priority sector)
P4: Auto loans ≥ Rs. 100 Crore (core business segment)
 (a) Formulate the complete Goal Programming model.
(b) Identify deviation variables for each goal.
(c) Write the lexicographic achievement function.
(d) Which goals conflict with each other? Explain economically.

Problem B7: Complete AHP Problem

Tata Capital is selecting a fintech partner for its digital lending platform.
Three fintech firms: FinX (Mumbai), PayStar (Bangalore), CreditAI (Hyderabad).

Criteria and pairwise comparison matrix:

C1-Tech C2-Scale C3-Compliance C4-Cost
C1 Tech: 1 1/2 3 2
C2 Scale: 2 1 4 3
C3 Comp.: 1/3 1/4 1 1/2
C4 Cost: 1/2 1/3 2 1

Alternative scores (from their individual pairwise matrices — given):

C1 C2 C3 C4
FinX: 0.54 0.30 0.16 0.54
PayStar: 0.30 0.54 0.54 0.30
CreditAI: 0.16 0.16 0.30 0.16

(a) Compute criteria weights using normalization method.

(b) Verify consistency (CR must be < 10%).
(c) Compute overall scores for each fintech firm.
(d) Recommend the best partner with business justification.

Part C: Case-Based and Essay Questions (15–20 marks each)

11. Reliance Jio expanded to 400 million subscribers in record time, requiring massive tower
infrastructure rollout across India. Design a multi-period LP model for Jio's tower installation
planning over 8 quarters. What are the key decision variables, linking constraints, capacity
constraints, and objective function? What data would Jio need to operationalize this model?
12. The Airports Authority of India (AAI) is evaluating whether to expand Tier-2 city airports (10
candidates) under UDAN scheme. Formulate a 0-1 Integer Programming model for airport
expansion selection. Include budget constraints, regional balance requirements, traffic
threshold conditions, and interdependency constraints (feeder airports require hub
expansion first). Discuss what non-modeled factors might override the mathematical
solution.
13. Nestle India manufactures products across multiple categories (Maggi, KitKat, Nescafe) at
plants in Pune, Pantnagar, and Moga. Design a multi-period, multi-product transportation-
production model covering 4 quarters. How would you handle seasonal demand (Maggi
 demand peaks in winter; KitKat peaks during summers and festivals)? How does the model
differ from a standard single-period transportation problem?
14. 'AHP and Goal Programming can be combined to make better complex decisions than either
technique alone.' Using the example of SEBI's evaluation of new stock exchange license
applications, design a two-stage MCDM framework that uses AHP for criterion weighting
and GP for resource allocation. Critically evaluate the advantages and limitations of this
combined approach.
15. Critically analyze the practical barriers to the adoption of optimization analytics in Indian
small and medium enterprises (SMEs). How do factors such as data quality, management
education, organizational culture, and software costs affect adoption? What steps could the
government, business schools, and industry bodies take to democratize optimization
analytics for Indian MSMEs?
CHAPTER 10: REFERENCES AND FURTHER
READING
10.1 Primary Textbooks
Author(s) Title Publisher Relevance to
Module 4
Hillier, F.S. & Introduction to Operations McGraw-Hill Comprehensive
Lieberman, G.J. Research, 10th Ed. Education coverage;
Chapters
3,4,7,9,14 directly
relevant
Taha, H.A. Operations Research: An Pearson Excellent on IP
Introduction, 10th Ed. and Goal
Programming;
Indian edition
available
Sharma, J.K. Operations Research: Theory and Macmillan India India-specific
Applications, 6th Ed. examples; widely
used in Indian
universities
Kapoor, V.K. Operations Research: Problems Sultan Chand & Extensive solved
and Solutions Sons problems; useful
for exam
preparation
Render, B., Stair R.M., Quantitative Analysis for Pearson Commerce-
Hanna M.E. Management, 12th Ed. oriented; strong on
project
management and
AHP

10.2 Multi-Period LP and Production Planning

Reference Details
Nahmias, S. & Olsen, T. Production and Operations Analysis, 7th Ed. — Waveland Press.
Chapters on inventory and production planning.
Silver, E.A., Pyke, D.F., Inventory Management and Production Planning and Scheduling —
Peterson, R. Wiley. The standard reference for multi-period inventory models.
Chopra, S. & Meindl, P. Supply Chain Management: Strategy, Planning, and Operation —
Pearson. Excellent on India supply chain applications.
10.3 Project Management (CPM/PERT)
Reference Details
Kerzner, H. Project Management: A Systems Approach to Planning, Scheduling,
and Controlling, 12th Ed. — Wiley.
PMBOK Guide A Guide to the Project Management Body of Knowledge, 7th Ed. —
PMI. Industry standard reference.
Moder, J.J., Phillips, C.R., Project Management with CPM, PERT and Precedence
Davis, E.W. Diagramming — Van Nostrand Reinhold.

10.4 Integer Programming

Reference Details
Wolsey, L.A. Integer Programming — Wiley. Advanced but authoritative on IP
theory including Branch and Bound and cutting planes.
Nemhauser, G.L. & Wolsey, Integer and Combinatorial Optimization — Wiley. Graduate-level
L.A. comprehensive reference.
Williams, H.P. Model Building in Mathematical Programming, 5th Ed. — Wiley.
Excellent on formulating real-world IP problems.

10.5 Goal Programming

Reference Details
Charnes, A. & Cooper, W.W. Management Models and Industrial Applications of Linear
Programming — Wiley. Original GP reference; foundation paper.
Ignizio, J.P. Goal Programming and Extensions — Lexington Books.
Comprehensive treatment of all GP variants.
Jones, D.F. & Tamiz, M. Practical Goal Programming — Springer. Modern applied
perspective including software implementation.

10.6 Analytic Hierarchy Process

Reference Details
Saaty, T.L. The Analytic Hierarchy Process — McGraw-Hill. The original and
definitive AHP reference by the inventor.
Saaty, T.L. & Vargas, L.G. Models, Methods, Concepts & Applications of the Analytic Hierarchy
Process — Springer. Applied AHP case studies.
Mu, E. & Pereyra-Rojas, M. Practical Decision Making using Super Decisions v3 — Springer.
Practical implementation guide.
10.7 Indian Business & Operations Research
Reference Details
Krishnamoorthy, C.S. & Artificial Intelligence and Expert Systems for Engineers — CRC
Rajeev, S. Press. Indian engineering management applications.
IIMB Working Papers IIM Bangalore Case Research Centre — India-specific optimization
and operations research case studies ([Link]/node/4).
Reddy, J.N. An Introduction to the Finite Element Method and applications in
Indian Engineering practice.
Economic Times & Business Indian business newspapers — ongoing coverage of analytics
Standard adoption by Indian corporations.

10.8 Software Resources

Software Purpose Access
Microsoft Excel Solver Basic LP, IP, nonlinear Built into Excel (Data → Solver Add-
optimization in); free
LINGO 19 Algebraic LP/IP/NLP modeling; [Link]; student version available
educational license
IBM CPLEX Industrial-grade LP/MIP solver Free academic license available; IBM
Academic Initiative
Gurobi Optimizer State-of-the-art MIP solver Free academic license for students
and researchers
Python PuLP Open-source LP/IP in Python pip install pulp; excellent tutorials on
GitHub
R: lpSolve package LP/IP in R statistical [Link]('lpSolve')
environment
Expert Choice AHP software tool [Link]; trial version
(commercial) available
Super Decisions Free AHP/ANP software by [Link]; free download
Saaty's team

Online Learning Resources

NPTEL ([Link]): 'Operations Research' courses by IIT faculty — free video lectures with
transcripts.
Coursera: 'Discrete Optimization' (Univ. of Melbourne) — excellent for IP and Branch & Bound.
MIT OpenCourseWare: 15.053 'Optimization Methods in Management Science' — full course
materials free.
INFORMS ([Link]): OR/MS Today journal — industry case studies including Indian
companies.
ORSI (Operational Research Society of India, [Link]): Indian OR community; conferences and
journals.

— END OF MODULE 4: OPTIMIZATION ANALYTICS —

Comprehensive Study Material with 40+ Solved Examples | Commerce & Management Students
Indian Company Cases: ITC, Amul, Tata Motors, NHAI, Infosys, SBI, BHEL, Wipro, HDFC, NTPC, SIDBI &
more