Business Decision-Making: Models, Types & Tools

TAKEAWAYS:

• Decisions vary: Structured (repetitive, quantifiable) vs. unstructured (novel, ambiguous) problems require different tools.
• Models matter: Rational, bounded rationality, and satisficing models explain how humans (and AI) make choices under uncertainty.
• Decision trees visualize trade-offs: Branches show outcomes, probabilities, and expected values for optimal choices.
• Payoff matrices reveal risks: They compare strategies against uncertain events (e.g., Daraz’s inventory vs. demand).
• Cost-benefit analysis justifies investments: Used by NTC for infrastructure projects or NEPSE for stock listings.
• Heuristics simplify complex choices: Shortcuts like "satisficing" (choosing "good enough") are common in real-world DSS.

1. Types of Business Decisions

Decisions in organizations fall into three categories, each requiring different tools and approaches:

1.1 Structured vs. Unstructured vs. Semi-Structured Decisions

A labeled flowchart showing structured (left), semi-structured (middle), and unstructured (right) paths. (Image: DavidLevinson, CC BY-SA 3.0, via Wikimedia Commons)

Why it matters:

• Structured decisions (e.g., bank loan approvals) can be automated with rules.
• Unstructured decisions (e.g., launching a new Pathao service route) require human expertise or AI.
• Semi-structured (e.g., NTC’s tariff adjustments) often use DSS to combine data with expert input.

2. Decision-Making Models

How do humans (and systems) actually make choices? Three key models explain this:

2.1 Rational Model (Optimizing)

Assumptions:

• Full information available.
• Clear, measurable goals.
• Logical, unbiased analysis.

Process:

1. Define objectives (e.g., maximize profit).
2. List all alternatives (e.g., Daraz: expand to Chitwan or Pokhara).
3. Evaluate each alternative’s outcomes.
4. Choose the optimal (best) solution.

Limitation:

• Rare in real life (e.g., no business has all data on future demand).

Worked Example: NEPSE Stock Listing Suppose NEPSE must decide whether to list a new company. Using the rational model:

• Objective: Maximize long-term investor trust.
• Alternatives:
  • List now (high short-term revenue but risk of scams).
  • Delay listing (lower revenue but safer).
• Outcomes:
  • List now: 80% chance of +₹50M revenue, 20% chance of -₹20M (scandal).
  • Delay: 100% chance of +₹30M revenue.
• Expected Value (EV):
  • List now: $0.8\times 50+0.2\times (-20)=36$ (in ₹M).
  • Delay: $30$.
  • Decision: Delay (higher EV).

Visual: Payoff Matrix

graph TD
    A["List Now"] -->|"80%"| B["+₹50M"]
    A -->|"20%"| C["-₹20M"]
    D["Delay"] --> E["+₹30M"]

2.2 Bounded Rationality (Simon’s Model)

Key Idea: Humans are cognitive misers—we simplify problems due to limited time, knowledge, or mental capacity.

Steps:

1. Satisfice: Choose the "good enough" option (not necessarily optimal).
2. Use heuristics: Mental shortcuts (e.g., "if it worked last time, do it again").
3. Incomplete information: Accept uncertainty (e.g., NTC predicting fiber demand).

Example: Pathao’s Route Planning

• Rational model: Calculate every possible route to minimize time (impossible).
• Bounded rationality: Use past data + driver feedback to pick a "good enough" route.

Advantage: Faster, practical for real-world DSS. Disadvantage: May miss better solutions.

2.3 Satisficing Model

• Definition: Selecting the first acceptable option (not the best).
• Example: A bank approves a loan when the applicant meets minimum criteria (e.g., 60% credit score), even if a slightly riskier candidate with 65% exists.

Comparison Table

3. Decision Support Systems (DSS) Tools

DSS help managers make semi-structured decisions by combining data, models, and user input.

3.1 Decision Trees

What it shows: All possible outcomes of a decision, with probabilities and payoffs.

Structure:

• Root: Decision point (e.g., "Should Daraz expand to Chitwan?").
• Branches: Alternatives (Yes/No).
• Leaves: Outcomes (profit/loss, success/failure).

Worked Example: Daraz’s Expansion Decision Daraz wants to expand to Chitwan. Data:

• Market size: 50,000 potential customers.
• Cost: ₹10M setup.
• Probabilities:
  • Success (high demand): 60% → ₹20M profit.
  • Failure (low demand): 40% → ₹5M loss.

Decision Tree:

graph TD
    A["Expand to Chitwan?"] --> B["Yes"]
    A --> C["No"]
    B --> D["Success: +₹20M (60%)"]
    B --> E["Failure: -₹5M (40%)"]
    C --> F["No expansion: ₹0"]

Calculations:

• EV(Yes): $0.6\times 20+0.4\times (-5)=12-2=₹10M$.
• EV(No): ₹0.
• Decision: Expand (EV = ₹10M > ₹0).

A real screenshot of a decision tree from a business textbook (e.g., Harvard case study). (Image: Dhrm77, CC0, via Wikimedia Commons)

3.2 Payoff Matrices

What it shows: Outcomes of strategies vs. uncertain events (e.g., Ncell’s network expansion).

Example: Ncell’s 5G Rollout

Optimal Strategy:

• If high demand is likely, rollout 5G.
• If low demand is likely, delay.

Real-World Use:

• NTC uses payoff matrices to decide on infrastructure projects (e.g., fiber vs. copper).
• Banks use them for loan approvals (e.g., "approve if credit score > 65%").

3.3 Cost-Benefit Analysis (CBA)

Formula: $$\text{Net Benefit}=\text{Total Benefits}-\text{Total Costs}$$ Example: NTC’s Fiber Expansion

• Cost: ₹500M for new cables.
• Benefits:
  • 5 years of ₹150M/year revenue.
  • Social benefit: better connectivity in rural areas (₹50M/year).
• Net Present Value (NPV): $$\text{NPV}=-500+\frac{150}{1.1^1}+\frac{150}{1.1^2}+\frac{200}{1.1^3}+\frac{200}{1.1^4}+\frac{200}{1.1^5}\approx ₹120M$$
• Decision: Expand (NPV > 0).

4. Heuristics in Decision Making

Definition: Mental shortcuts to simplify complex decisions.

Example: Bank Loan Approval

• Heuristic: "If the applicant is from Kathmandu University, approve."
• Risk: May reject a rural entrepreneur with better potential.

## In the Real World

1. Ncell’s Network Expansion

   • Tool: Payoff matrices and decision trees.
   • How: Ncell uses historical data to predict demand in new areas (e.g., Pokhara vs. Dhangadi). A decision tree helps weigh the cost of laying cables against potential subscriber growth.

2. Daraz’s Inventory Management

   • Tool: Satisficing + bounded rationality.
   • How: Daraz doesn’t optimize stock levels perfectly (impossible). Instead, it uses past sales data to order "good enough" quantities, avoiding both stockouts and overstocking.

3. Khalti’s Fraud Detection

   • Tool: Heuristics + cost-benefit analysis.
   • How: Khalti flags transactions that match fraud patterns (heuristic). The CBA decides whether to block a transaction (cost: customer frustration; benefit: preventing loss).

4. NTC’s Tariff Adjustments

   • Tool: Cost-benefit analysis.
   • How: Before raising internet prices, NTC models the impact on revenue vs. customer churn. If NPV is negative, they delay or adjust the hike.

5. NEPSE’s IPO Approvals

   • Tool: Rational model (partially).
   • How: NEPSE evaluates companies based on financials, but also uses bounded rationality—approving an IPO if it meets minimum criteria (e.g., 3 years of profits), even if a slightly riskier company exists.

## Exam Tip

How this unit is tested in TU/PU exams:

1. Definitions: Expect 2–3 marks on distinguishing structured/unstructured decisions or rational vs. bounded rationality.

   • Example question: "Differentiate between satisficing and optimizing with an example from a Nepalese company."

2. Calculations: Decision trees and payoff matrices are high-yield for numerical questions (5–10 marks).

   • Example: Given a payoff matrix for a bank loan, calculate the expected value of approving/rejecting a loan applicant.

3. Applications: Link concepts to real-world scenarios (e.g., "How would Ncell use a decision tree for 5G expansion?").

   • Tip: Always use Nepalese examples (Ncell, Daraz, Khalti) in answers—they show contextual understanding.

4. Heuristics and Biases: Short-answer questions on common biases (e.g., "Explain anchoring with a Pathao route-planning example").

5. Cost-Benefit Analysis: One question will ask you to compute NPV or net benefit for a project (e.g., NTC’s fiber expansion).

   • Formula to remember: $$\text{NPV}=\sum \frac{{\text{Benefits}}_t}{(1+r{)}^t}-\text{Initial Cost}$$ where $r$ = discount rate (e.g., 10% = 0.1).

Common Mistakes to Avoid:

• Ignoring probabilities in decision trees (always calculate EV).
• Confusing structured (automated) vs. unstructured (human judgment) decisions.
• Forgetting to label axes in graphs (e.g., "Profit (₹)" vs. "Probability").

Quick Revision Checklist:

• Can you draw a decision tree for a given scenario?
• Do you know the 3 decision-making models and their differences?
• Can you compute NPV for a project?
• Are you familiar with 2 heuristics and their biases?