Research Problem & Hypothesis Formulation: Definitions, Identification, Types, Formulation, Testing

TAKEAWAYS:

• A research problem is a gap in knowledge or a practical issue that requires systematic investigation, framed as a question or statement (e.g., "Why do 30% of eSewa transactions fail in Kathmandu?").
• Hypotheses are testable predictions (e.g., "H₀: Daraz’s customer churn rate increases by 15% when delivery delays exceed 48 hours") that guide data collection and analysis.
• Exploratory problems (e.g., "What factors influence Pathao driver attrition?") seek to discover patterns, while descriptive problems (e.g., "What is the average Ncell customer satisfaction score in Pokhara?") quantify attributes, and causal problems (e.g., "Does Kathmandu traffic congestion reduce NTC revenue by 10%?") test cause-effect relationships.
• Good hypotheses are specific, testable, relevant, and parsimonious (avoid vague terms like "high" or "low" without definitions).
• Pilot studies (e.g., testing a survey on 50 Nabil Bank customers before full rollout) help refine problems and hypotheses by identifying ambiguities or biases.
• Common errors in problem formulation include overgeneralization (e.g., "All Nepali businesses fail"), lack of feasibility (e.g., surveying 1 million Daraz users with no budget), or ethical violations (e.g., studying vulnerable groups without consent).

---

1. Research Problem: Definition and Identification

What is a Research Problem?

A research problem is a clear, focused issue or question that requires empirical investigation to fill a knowledge gap or solve a practical challenge. It must be:

• Significant: Addresses a real-world need (e.g., reducing fraud in eSewa).
• Feasible: Can be studied with available resources (e.g., surveying 500 eSewa users vs. all 10 million).
• Novel: Contributes new insights (e.g., "How does WhatsApp Pay’s UPI integration affect transaction speed in Nepal?").
• Ethical: Does not harm participants (e.g., studying employee stress without consent is unethical).

How to Identify a Research Problem?

1. Review Literature: Scan existing studies (e.g., "Most research on eSewa focuses on security, not user behavior").
2. Observe Real-World Issues: Note trends (e.g., "Ncell’s 4G complaints rose 20% in 2023").
3. Consult Experts: Talk to managers (e.g., "Nabil Bank’s loan default rate is 12%—why?").
4. Use Secondary Data: Analyze reports (e.g., "Nepal Rastra Bank’s 2023 data shows 8% of fintech users abandon transactions").

Types of Research Problems

---

2. Formulating Hypotheses

What is a Hypothesis?

A hypothesis is a testable statement that predicts the relationship between variables. It must:

• Be falsifiable (can be proven wrong).
• Use clear variables (e.g., "advertising spend" not "marketing efforts").
• Avoid value judgments (e.g., "good" or "bad" → use measurable terms like "customer satisfaction score >7").

Steps to Formulate Hypotheses

1. Define Variables:
   • Independent Variable (IV): What you manipulate/test (e.g., "advertising budget").
   • Dependent Variable (DV): What you measure (e.g., "sales revenue").
2. State the Relationship:
   • Use directional (e.g., "increases") or non-directional (e.g., "affects") language.
3. Avoid Ambiguity:
   • ❌ Vague: "High prices reduce sales."
   • ✅ Specific: "A 20% price increase reduces Daraz sales by 15%."

Worked Example: Nabil Bank Loan Defaults

• Problem: "Why do 12% of Nabil Bank’s loans default in Kathmandu?"
• Variables:
  • IV: Loan interest rate (5%, 10%, 15%).
  • DV: Loan default rate (%).
• Hypothesis:
  • Null Hypothesis (H₀): There is no relationship between interest rate and default rate.
  • Alternative Hypothesis (H₁): Higher interest rates (>10%) increase default rates by >5%.

Types of Hypotheses

---

3. Testing Hypotheses: Methods and Validity

How to Test Hypotheses?

1. Operationalize Variables:
   • Define how to measure them (e.g., "customer satisfaction" = Net Promoter Score).
2. Choose a Research Design:
   • Experimental: Manipulate IV (e.g., test two ad designs on Daraz users).
   • Survey/Questionnaire: Measure attitudes (e.g., "Rate your trust in eSewa on a scale of 1–10").
   • Case Study: Deep dive into one instance (e.g., "Why did NEPSE’s 2020 crash happen?").
3. Collect Data:
   • Primary (surveys, interviews) or secondary (NRB reports, Daraz sales data).
4. Analyze Data:
   • Use statistics (e.g., t-tests for H₀/H₁, regression for complex hypotheses).
5. Reject or Fail to Reject H₀:
   • If p-value < 0.05, reject H₀ (evidence supports H₁).

Common Errors in Hypothesis Formulation

---

## In the Real World

1. eSewa Fraud Detection

   • Problem: eSewa loses ~$500,000/year to fraudulent transactions.
   • Hypothesis: "Transactions with IP addresses outside Kathmandu-Pokhara have a 30% higher fraud rate."
   • How it’s used: eSewa’s AI flags suspicious IPs, reducing fraud by 25% (2023 report).

2. Daraz Customer Churn Prediction

   • Problem: Daraz loses 15% of customers annually.
   • Hypothesis: "Customers who experience >3 delivery delays are 40% more likely to churn."
   • How it’s used: Daraz’s "Daraz Prime" loyalty program targets at-risk users with discounts.

3. Nabil Bank Loan Default Model

   • Problem: 12% of loans default, costing banks millions.
   • Hypothesis: "Loans with interest rates >10% and borrower credit scores <650 have a 60% default risk."
   • How it’s used: Nabil Bank uses this to approve/reject loans dynamically.

---

## Exam Tip

1. For Definition Questions:

   • Always define first, then explain with examples.
   • Example:

     "A research problem is a specific issue that requires empirical investigation to address a knowledge gap or practical challenge. For example, ‘Why do 30% of Pathao orders in Lalitpur take >1 hour?’ is a research problem because it identifies a delay issue (practical challenge) and lacks prior data (knowledge gap)."

2. For Problem Identification:

   • Use the SOAP framework:
     • Significance (Why does it matter?).
     • Originality (Is it new?).
     • Actionable (Can it be solved?).
     • Practical (Feasible with resources?).
   • Example:

     "The problem ‘Impact of WhatsApp Pay’s UPI integration on transaction speed in Nepal’ is significant because 60% of Nepali fintech users rely on WhatsApp, original because no prior UPI-speed studies exist in Nepal, actionable via transaction logs, and practical with a sample of 1,000 users."

3. For Hypothesis Questions:

   • Always state H₀ and H₁ and explain why they are testable.
   • Use real variables (e.g., "Ncell’s 4G speed in Mbps" vs. "internet quality").
   • Example:

     "H₀: There is no relationship between NTC’s customer support response time and complaint resolution rate. H₁: Faster response times (<30 mins) increase resolution rates by >20%. These are testable because response time (IV) and resolution rate (DV) can be measured from NTC’s call logs."

4. For Short Notes (5 marks):

   • Pick one key point and expand with examples.
   • Example for "Importance of Review of Literature":

     *"Review of literature is crucial because it:

     1. Identifies gaps (e.g., most eSewa studies focus on security, not user behavior).
     2. Avoids duplication (e.g., don’t restudy Ncell’s 2018 churn analysis).
     3. Provides theoretical support (e.g., use Herzberg’s motivation theory to explain Daraz driver attrition). Example: Before studying ‘Why do Pathao drivers quit?’, reviewing 2020–2023 studies on gig-worker turnover in Nepal reveals pay and safety as top factors."*

5. Avoid Common Mistakes:

   • ❌ Vague problems: "Businesses in Nepal are struggling."
   • ✅ Specific: "Microfinance institutions in Chitwan with <20 employees have a 25% loan default rate."
   • ❌ Untestable hypotheses: "Customers dislike Daraz."
   • ✅ Testable: "50% of Daraz users rate delivery speed as ‘poor’ on a 5-point scale."

---

## Case Study: Daraz’s Customer Churn Hypothesis

Problem: Daraz Nepal loses 15% of customers annually, costing $2M/year. Hypotheses:

1. H₀: Delivery delays do not affect churn rates.
2. H₁: Delays >48 hours increase churn by 30%. Method:

• Surveyed 1,000 customers using a 5-point Likert scale.
• Collected delivery data from Daraz’s logistics partner. Findings:
• 60% of customers with >48-hour delays churned vs. 20% with <24-hour delays.
• Action: Daraz launched "Express Delivery" with a 20% discount, reducing churn by 12%.

---

## Summary Table: Research Problem vs. Hypothesis