Simple Additive Weighting (SAW) & Weighted Product Method (WPM)

A Comprehensive Reference Manual for Multi-Criteria Decision Making (MCDM)

Introduction to MCDM
Decision-Making Terminology & Notation
The Simple Additive Weighting (SAW) Method

Core Mechanics & Normalization
Mathematical Model & Algorithmic Pipeline
Comprehensive Worked Numerical Examples

The Weighted Product Method (WPM)

Core Mechanics & Multiplicative Aggregation
Mathematical Model & Algorithmic Pipeline
Worked Numerical Example

Comparative Analysis Matrix: SAW vs. WPM
Sensitivity Analysis Foundations
Empirical Applications Across Domains
Methodological Best Practices
Rigorous Examination & Review Questions
Strategic Summary & Framework Synthesis

1. Introduction to Multi-Criteria Decision Making (MCDM)

In complex engineering systems, operational management, and strategic procurement, decision-makers rarely evaluate options based on a single metric. Instead, they must navigate a landscape of conflicting objectives—such as maximizing performance while minimizing cost.
Multi-Criteria Decision Making (MCDM) provides the mathematical scaffolding required to resolve these trade-offs systematically. Rather than relying on subjective intuition, MCDM formalizes the evaluation process, ensuring transparency, reproducibility, and auditability.
While advanced frameworks exist, Simple Additive Weighting (SAW) and the Weighted Product Method (WPM) form the analytical bedrock of decision theory. They serve as the introductory entry point for M.Tech coursework, operational research, and engineering optimization studies.

2. Decision-Making Terminology & Notation

To maintain mathematical consistency throughout this manual, the following standard matrix notations and variables are defined:

m: The total number of alternatives under evaluation (i = 1, 2, \dots, m).
n: The total number of evaluation criteria (j = 1, 2, \dots, n).
X: The raw Decision Matrix of size m \times n, where x_{ij} represents the performance of alternative i against criterion j.
R: The Normalized Decision Matrix of size m \times n, where r_{ij} represents the dimensionless, scaled performance value (0 \le r_{ij} \le 1).
w_j: The relative importance weight assigned to criterion j, where the weight vector satisfies the probability constraint:
V_i: The final preference score of alternative i derived via the SAW method.
S_i: The final preference score of alternative i derived via the WPM method.

3. The Simple Additive Weighting (SAW) Method

Core Mechanics & Normalization

Simple Additive Weighting (SAW)—traditionally referred to as the Weighted Sum Method (WSM)—operates on the principle of linear compensation. In this framework, a low score in one criterion can be fully counterbalanced by an exceptionally high score in another.
Because raw criteria data contain distinct engineering units (e.g., Gigabytes, Hours, Indian Rupees, Grams), they cannot be directly added. SAW resolves this by employing Linear Max/Min Scale Normalization to transform all parameters into dimensionless values between 0 and 1.
Criteria are mathematically segregated into two operational types:

1. Benefit Criteria (Maximization Objective)

Criteria where higher performance values are optimal (e.g., throughput, efficiency, capacity). The raw value is divided by the maximum value in the column:

2. Cost Criteria (Minimization Objective)

Criteria where lower performance values are optimal (e.g., price, mass, defect rate). The minimum value in the column is divided by the raw value to invert the scaling:

Mathematical Model

Once normalized, the total utility or preference score V_i for each alternative is computed using an additive linear aggregation function:
The alternatives are ordered based on descending values of V_i. The optimal alternative A^* is selected such that:

Algorithmic Pipeline

┌────────────────────────────────────────────────────────┐  
│ 1. Construct the Raw Decision Matrix (X_mxn)          │  
└───────────────────────────┬────────────────────────────┘  
                            ▼  
┌────────────────────────────────────────────────────────┐  
│ 2. Define Weights (w_j) where sum(w_j) = 1.0           │  
└───────────────────────────┬────────────────────────────┘  
                            ▼  
┌────────────────────────────────────────────────────────┐  
│ 3. segregate Criteria into Benefit vs. Cost Types      │  
└───────────────────────────┬────────────────────────────┘  
                            ▼  
┌────────────────────────────────────────────────────────┐  
│ 4. Execute Linear Normalization to construct R_mxn     │  
└───────────────────────────┬────────────────────────────┘  
                            ▼  
┌────────────────────────────────────────────────────────┐  
│ 5. Compute Weighted Sums: V_i = sum(w_j * r_ij)        │  
└───────────────────────────┬────────────────────────────┘  
                            ▼  
┌────────────────────────────────────────────────────────┐  
│ 6. Rank Alternatives in Descending Order of V_i        │  
└────────────────────────────────────────────────────────┘

Comprehensive Worked Numerical Examples

Example 3.1: Academic Laptop Selection Problem

A computer science student requires an optimal workstation based on three benefit criteria: RAM (GB), SSD (GB), and Battery Life (Hours).
1. Raw Decision Matrix & Weights

Weights: w_{\text{RAM}} = 0.4, w_{\text{SSD}} = 0.3, w_{\text{Battery}} = 0.3
| Alternative (i) | RAM (GB) | SSD (GB) | Battery (HR) |
|---|---|---|---|
| Laptop A | 8 | 512 | 6 |
| Laptop B | 16 | 256 | 8 |
| Laptop C | 12 | 512 | 7 |
| Max Vector | 16 | 512 | 8 |
2. Normalization Process (Benefit Criteria)
r_{A,1} = 8 / 16 = 0.50; r_{A,2} = 512 / 512 = 1.00; r_{A,3} = 6 / 8 = 0.75
r_{B,1} = 16 / 16 = 1.00; r_{B,2} = 256 / 512 = 0.50; r_{B,3} = 8 / 8 = 1.00
r_{C,1} = 12 / 16 = 0.75; r_{C,2} = 512 / 512 = 1.00; r_{C,3} = 7 / 8 = 0.875
3. Weighted Score Aggregation (V_i)
- - 4. Ranking Resolution
    | Rank | Alternative | SAW Score (V_i) | Status |
    |---|---|---|---|
    | 1 | Laptop C | 0.8625 | Optimal Selection |
    | 2 | Laptop B | 0.8500 | Acceptable Alternative |
    | 3 | Laptop A | 0.7250 | Suboptimal |

Example 3.2: Advanced Manufacturing System Selection Problem

An industrial engineering firm needs to evaluate four automated flexible manufacturing configurations (P_1 to P_4) against five diverse criteria.

Benefit Criteria: X_1, X_2, X_3, X_5
Cost Criterion: X_4 (Lower processing cost per unit is preferred)
Weight Vector: W = [0.3, 0.2, 0.1, 0.2, 0.3]
1. Raw Decision Matrix (X)
| Alternative | X_1 (Benefit) | X_2 (Benefit) | X_3 (Benefit) | X_4 (Cost) | X_5 (Benefit) |
|---|---|---|---|---|---|
| P_1 | 9 | 5 | 3 | 7 | 3 |
| P_2 | 5 | 7 | 5 | 5 | 5 |
| P_3 | 3 | 9 | 7 | 3 | 7 |
| P_4 | 1 | 1 | 1 | 1 | 9 |
| Ideal Reference | Max = 9 | Max = 9 | Max = 7 | Min = 1 | Max = 9 |
2. Normalized Decision Matrix (R)
For X_1, X_2, X_3, X_5: r_{ij} = x_{ij} / \max(x_j)
For X_4: r_{i4} = \min(x_4) / x_{i4} (e.g., for P_1, r_{14} = 1 / 7 \approx 0.143)
| Alternative | X_1 | X_2 | X_3 | X_4 | X_5 |
|---|---|---|---|---|---|
| P_1 | 1.000 | 0.556 | 0.429 | 0.143 | 0.333 |
| P_2 | 0.556 | 0.778 | 0.714 | 0.200 | 0.556 |
| P_3 | 0.333 | 1.000 | 1.000 | 0.333 | 0.778 |
| P_4 | 0.111 | 0.111 | 0.143 | 1.000 | 1.000 |
3. Weighted Score Aggregation (V_i)
- - - 4. Final Operational Ranking
      | Rank | System Configuration | SAW Score (V_i) |
      |---|---|---|
      | 1 | P_3 | 0.6999 |
      | 2 | P_2 | 0.6006 |
      | 3 | P_1 | 0.5826 |
      | 4 | P_4 | 0.5698 |

Example 3.3: Enterprise Cell Phone Deployment Problem

A procurement manager selects a standard mobile platform for field technicians based on six criteria:

Cost Criteria: Price (INR), Weight (grams)
Benefit Criteria: Storage (GB), RAM (GB), Camera (MP), Battery (mAh)
Weight Vector (W): Price=0.30, Weight=0.10, Storage=0.15, RAM=0.15, Camera=0.20, Battery=0.10
1. Raw Matrix Data & Reference Optima
\min(\text{Price}) = 12,500 | \min(\text{Weight}) = 130
\max(\text{Storage}) = 64 | \max(\text{RAM}) = 4 | \max(\text{Camera}) = 13 | \max(\text{Battery}) = 5,000
| ID | Price (INR) | Weight (g) | Storage (GB) | RAM (GB) | Camera (MP) | Battery (mAh) |
|---|---|---|---|---|---|---|
| 1 | 14,000 | 180 | 64 | 3 | 13 | 3,300 |
| 2 | 12,500 | 170 | 64 | 4 | 12 | 4,100 |
| 3 | 15,500 | 130 | 32 | 3 | 8 | 5,000 |
| 4 | 13,000 | 150 | 64 | 4 | 12 | 4,000 |
| 5 | 19,500 | 130 | 32 | 4 | 8 | 3,300 |
| 6 | 17,000 | 190 | 64 | 3 | 13 | 5,000 |
2. Normalized Matrix Mapping (R)
| ID | Price | Weight | Storage | RAM | Camera | Battery |
|---|---|---|---|---|---|---|
| 1 | 0.8929 | 0.7222 | 1.0000 | 0.7500 | 1.0000 | 0.6600 |
| 2 | 1.0000 | 0.7647 | 1.0000 | 1.0000 | 0.9231 | 0.8200 |
| 3 | 0.8065 | 1.0000 | 0.5000 | 0.7500 | 0.6154 | 1.0000 |
| 4 | 0.9615 | 0.8667 | 1.0000 | 1.0000 | 0.9231 | 0.8000 |
| 5 | 0.6410 | 1.0000 | 0.5000 | 1.0000 | 0.6154 | 0.6600 |
| 6 | 0.7353 | 0.6842 | 1.0000 | 0.7500 | 1.0000 | 1.0000 |
3. Final Multi-Attribute Aggregation & Sorting
- - - - | Final Rank | Platform ID | Score (V_i) |
        |---|---|---|
        | 1 | Phone 4 | 0.9498 |
        | 2 | Phone 2 | 0.9431 |
        | 3 | Phone 1 | 0.8694 |
        | 4 | Phone 6 | 0.8524 |
        | 5 | Phone 3 | 0.7525 |
        | 6 | Phone 5 | 0.7054 |

Strategic Note: While Phone 2 has a lower raw purchase price, Phone 4's superior weight profile yields a higher cumulative linear score, making it the mathematically superior alternative.

Advantages & Limitations of SAW

Advantages

Computational Simplicity: Easily managed within basic spreadsheet environments without requiring high-level programming scripts.
Intuitive Transparency: Directly maps raw performance variances to a final unified baseline.
Linear Preservations: Standardized scores maintain relative linear proportional spacing between alternatives.

Limitations

Compensatory Distortion: Extreme underperformance in high-risk categories (e.g., critical safety violations) can be artificially masked by superior marks in less critical benefit fields.
Strict Unit Scaling Dependencies: Relies entirely on linear normalization vectors. If the scale limits shift, the ultimate option sequence can alter unexpectedly.

4. The Weighted Product Method (WPM)

Core Mechanics & Multiplicative Aggregation

The Weighted Product Method (WPM) replaces linear addition with multiplicative aggregation. Instead of multiplying performance metrics by weights, criteria values are raised to the power of their corresponding weights.
A core advantage of WPM is its dimensionless nature. Because it evaluates performance as relative ratio comparisons, the raw units cancel out mathematically, frequently eliminating the requirement for preliminary data normalization.
To handle cost attributes without altering the underlying multiplication mechanics, WPM employs exponent signed configuration:

Benefit Criteria: Exponents are assigned positive weights (+w_j).
Cost Criteria: Exponents are assigned negative weights (-w_j).

Mathematical Model

The total structural score S_i for an alternative is computed via the product of its exponentiated features:
Alternatively, when comparing two options (A_k and A_l) directly to determine dominance, a relative ratio layout is used:

If S_k / S_l > 1.0, alternative A_k dominates alternative A_l.

Algorithmic Pipeline

┌────────────────────────────────────────────────────────┐  
│ 1. Populate Raw Decision Performance Matrix           │  
└───────────────────────────┬────────────────────────────┘  
                            ▼  
┌────────────────────────────────────────────────────────┐  
│ 2. Formalize Exponent Weight Vector (w_j)              │  
└───────────────────────────┬────────────────────────────┘  
                            ▼  
┌────────────────────────────────────────────────────────┐  
│ 3. Apply Sign Shifts: (+w_j) Benefit / (-w_j) Cost     │  
└───────────────────────────┬────────────────────────────┘  
                            ▼  
┌────────────────────────────────────────────────────────┐  
│ 4. Execute Multiplicative Exponentiation per Row       │  
└───────────────────────────┬────────────────────────────┘  
                            ▼  
┌────────────────────────────────────────────────────────┐  
│ 5. Compile and Sort Final Product Signatures (S_i)     │  
└────────────────────────────────────────────────────────┘

Worked Numerical WPM Example

Example 4.1: Industrial Lab Instrument Procurement

A chemical research center needs to choose a precision diagnostic spectrometer from three alternatives (A_1, A_2, A_3) based on Accuracy (Benefit), Speed (Benefit), and Cost (Cost).

Weight Profiling: w_{\text{Accuracy}} = 0.5, w_{\text{Speed}} = 0.3, w_{\text{Cost}} = 0.2
Operational Exponents: \text{Accuracy} = +0.5, \text{Speed} = +0.3, \text{Cost} = -0.2
1. Raw Matrix Data
| Alternative | Accuracy (Benefit) | Speed (Benefit) | Cost (Cost) |
|---|---|---|---|
| Instrument A_1 | 9 | 7 | 5 |
| Instrument A_2 | 8 | 9 | 6 |
| Instrument A_3 | 7 | 8 | 4 |
2. Multiplicative Compounding Calculations
- - 3. Ranking Synthesis
    | Rank Position | Instrument Alternative | WPM Multiplicative Score (S_i) |
    |---|---|---|
    | 1 | Instrument A_1 | 3.8983 |
    | 2 | Instrument A_2 | 3.8212 |
    | 3 | Instrument A_3 | 3.7420 |

Analytical Verification: Instrument A_1 scores highest primarily because its lower operational unit cost vector (5 vs 6) penalizes its final score less than A_2's, keeping A_1 at the top of the ranking.

Advantages & Limitations of WPM

Advantages

Elimination of Scale Distortion: By avoiding linear transformations, it prevents mathematical artifacts introduced by normalization scaling.
Non-Compensatory Sensitivity: A near-zero performance value in a critical criterion severely drops the product metric, preventing fatal single-point deficiencies from being masked.

Limitations

Zero-Value Vulnerability: If any raw criterion value equals zero (x_{ij}=0), the cumulative product drops to zero instantly, requiring data shifts or scale adjustments.
Mathematical Complexity: Exponentiation fractions require scientific computation platforms, making rapid field assessments harder to execute manually.

5. Comparative Analysis Matrix: SAW vs. WPM

The table below outlines the core differences between the two methods:

Architectural Metric	Simple Additive Weighting (SAW)	Weighted Product Method (WPM)
Core Operation	Linear Additive Synthesis	Geometric Multiplicative Synthesis
Mathematical Equation	V_i = \sum w_j r_{ij}	S_i = \prod x_{ij}^{w_j}
Normalization Necessity	Mandatory prior to addition	Not required for dimensionless inputs
Cost Attributes Handling	Scaled division: \min(x)/x	Signed Exponents Configuration: -w_j
Compensatory Degree	High (Full parameter trade-off)	Low (Strict penalty for weak scores)
Zero Values Handling	No systemic distortion	Destroys product outputs (S_i \to 0)
Optimal Use Cases	Unified, standard linear scale problems	Non-linear engineering design contexts

6. Sensitivity Analysis Foundations

A critical step in any MCDM process is verifying the stability of the final decision through Sensitivity Analysis. Because weights are often determined by subjective expert opinions or group consensus, we must evaluate how variations in these weights impact the final rankings.

If a small ±2% shift in the weight of a single criterion flips the rank order of the top alternatives, the decision model is sensitive, high-risk, and requires further optimization.

Methodological Execution Pipeline

Isolate Target Criterion: Incrementally adjust a target weight w_k by an offset \Delta.
Re-proportionalize Remaining Vector: Dynamically adjust all remaining weights to satisfy the total sum constraint:
Recalculate Preferences: Re-run the SAW or WPM aggregation step.
Map Rank Stability Boundaries: Identify the exact threshold values where rank switching occurs to determine the robustness of the selection.

7. Empirical Applications Across Domains

MCDM methods are widely applied across multiple industrial and academic fields:

┌───────────────────────────────────────────────────────────────────────────┐  
│                       CROSS-DOMAIN APPLICATIONS                           │  
├─────────────────────┬───────────────────────────────┬─────────────────────┤  
│     ENGINEERING     │       PROJECT MANAGEMENT      │    MANUFACTURING    │  
│  Material Selection │    Contractor Selection       │ Supplier Evaluation │  
│  Design Validation  │    Risk Prioritization        │ Facility Location   │  
├─────────────────────┼───────────────────────────────┼─────────────────────┤  
│     HEALTHCARE      │           EDUCATION           │      BUSINESS       │  
│  Equipment Purchase │    University Rankings        │ Vendor Procurement  │  
│  Treatment Audits   │    Faculty Appraisals         │ Portfolio Analysis  │  
└─────────────────────┴───────────────────────────────┴─────────────────────┘

Civil & Structural Engineering: Evaluating bids from contractors by balancing bid cost against historical project quality and safety records.
Supply Chain Operations: Selecting suppliers based on conflicting metrics like lead time, component cost, carbon footprint, and defect rates.

8. Methodological Best Practices

To minimize structural errors during analysis, ensure these rules are followed:

Enforce Sum Constraints: Always verify that the final criteria weight vectors sum to exactly 1.000.
Isolate Cost Scales: Clearly identify cost criteria before normalizing to prevent inverting the intended optimization goal.
Apply Data Offsets: Add a small baseline offset (\epsilon = 0.001) to any zero values in raw data before running WPM calculations to prevent mathematical cancellation.

9. Rigorous Examination & Review Questions

Analytical & Theoretical Inquiries

Explain the concept of compensatory models in MCDM. Prove mathematically how SAW exhibits a high compensatory nature compared to WPM.
Detail the mathematical steps required to handle a minimization (cost) criterion in both SAW and WPM frameworks.
Why does WPM eliminate the need for standard vector or linear normalization when evaluating datasets with mixed physical units? Explain the underlying dimensional analysis.

Applied Computational Challenge

Problem: A logistics provider must select a new fleet vehicle based on four criteria: Fuel Efficiency (Benefit, w=0.35), Cargo Volume (Benefit, w=0.25), Initial Capital Cost (Cost, w=0.30), and Carbon Emissions (Cost, w=0.10).
Given the raw matrix below, compute the final operational rankings using both the SAW and WPM frameworks, and explain any variations in the resulting sequence.

Fleet Alternative	Efficiency (km/L)	Volume (\text{m}^3)	Cost (INR)	Emissions (g/km)
Vehicle V1	15	12	1,800,000	120
Vehicle V2	18	10	2,100,000	95
Vehicle V3	14	15	1,950,000	110

10. Strategic Summary & Framework Synthesis

Choosing between SAW and WPM depends on the operational structure of the problem:

SAW is best suited for general decision-making problems with independent criteria that can be easily normalized, and where simplicity and stakeholder transparency are the primary requirements.
WPM is ideal for highly sensitive engineering design problems involving multi-scale units and cost/benefit trade-offs, where poor performance in a critical area cannot be overlooked.
By combining these foundational frameworks with sensitivity analysis, engineers and managers can make rational, data-driven decisions that stand up to rigorous auditing.

Monday, 20 July 2026