Research Design: Types, Models & Applications in Business

What is Research Design?

Research design is the blueprint of a study. It outlines:

• How data will be collected (surveys, experiments, observations).
• When and where (field vs. lab, cross-sectional vs. longitudinal).
• Why a specific approach is chosen (to answer the research question).

A poorly designed study leads to invalid conclusions, while a well-structured design ensures reliable, actionable insights.

Why Does It Matter?

• Avoids bias: Ensures results reflect reality, not researcher assumptions.
• Saves resources: Prevents wasted time/money on flawed methods.
• Guides decisions: Helps businesses (e.g., NTC’s network expansion) make data-driven choices.

Types of Research Design

Research designs are categorized based on purpose, time dimension, and control over variables. Below is a comparison table of the three primary purpose-based designs:

1. Exploratory Design

Goal: Generate insights, define problems, or develop hypotheses. Methods:

• Secondary research (literature reviews, existing data).
• Qualitative techniques: Interviews, focus groups, case studies.

How It Works (Trace Example)

Problem: Why are Nepali students switching from traditional banks to digital wallets like eSewa?

1. Review data: Analyze NPR’s reports on fintech growth.
2. Conduct interviews: Talk to 10 students in Pokhara about their habits.
3. Identify themes: Lack of branch access, faster transactions, trust in digital security.
4. Formulate hypothesis: "Convenience and speed are primary drivers of wallet adoption."

Visual: Exploratory Research Process

flowchart TD
    A["Problem Identification"] --> B["Secondary Research"]
    B --> C["Qualitative Data Collection\n(Interviews/Focus Groups)"]
    C --> D["Theme Analysis"]
    D --> E["Hypothesis Development"]

Real-World Tie:

• Daraz used exploratory research to understand why customers abandoned carts. They found shipping costs were the top reason, leading to a "free delivery above Rs. 1000" policy.

2. Descriptive Design

Goal: Paint a clear picture of a population or situation. Methods:

• Surveys (structured questionnaires).
• Observations (e.g., traffic patterns in Kathmandu).
• Case studies (e.g., analyzing a single bank’s loan defaults).

Worked Example: NTC’s Internet Usage Study

Research Question: What are the internet usage patterns in Nepal’s rural vs. urban areas? Design: Descriptive (cross-sectional survey). Steps:

1. Sample: 1000 users (500 urban, 500 rural).
2. Data: Age, device type, daily usage hours, primary use (social media, work, etc.).
3. Findings:
   • Urban: 60% use smartphones; 70% for social media.
   • Rural: 40% use feature phones; 60% for education/work.
4. Action: NTC prioritized cheaper data plans for rural areas.

Visual: Descriptive Research Structure

mindmap
  root((Descriptive Design))
    Population["Define Target Group\n(e.g., NTC users)"]
    Data["Collect Data\n(Surveys, Observations)"]
    Analysis["Summarize Statistics\n(Averages, Percentages)"]
    Reporting["Present Findings\n(Charts, Tables)"]

Real-World Tie:

• Khalti uses descriptive surveys to track transaction volumes by age group, helping them design targeted promotions (e.g., discounts for students).

3. Explanatory/Causal Design

Goal: Prove cause-and-effect (e.g., "Does X lead to Y?"). Methods:

• Experiments (controlled tests).
• Quasi-experiments (real-world interventions).
• Longitudinal studies (tracking over time).

Key Components

Worked Example: Nabil Bank’s Loan Default Study

Research Question: Does providing financial literacy training reduce loan defaults? Design: Quasi-experimental (no random assignment; uses existing groups). Steps:

1. Group 1 (Treatment): 500 borrowers given 3-month financial training.
2. Group 2 (Control): 500 borrowers (no training).
3. Measure: Default rates after 12 months.
4. Result: Group 1 had 15% defaults vs. Group 2’s 28%.
5. Conclusion: Training reduces defaults by 13 percentage points.

Visual: Causal Design Framework

flowchart LR
    A["Independent Variable\n(e.g., Training Program)"] -->|"Leads To"| B["Dependent Variable\n(e.g., Loan Default Rate)"]
    C["Control Variables\n(Same loan terms, borrower profile)"] --> B
    D["Random Assignment\n(If possible)"] --> A & B

Real-World Tie:

• Pathao tested whether dynamic pricing (raising fares during peak hours) reduced driver shortages. They found a 20% increase in driver availability during surges.

Time Dimension in Research Design

Designs also differ by when data is collected:

Experimental vs. Non-Experimental Designs

Visual: Experimental vs. Non-Experimental

flowchart TD
    subgraph Experimental
        A["Manipulate IV\n(e.g., New App Feature)"] --> B["Random Assignment"] --> C["Measure DV\n(e.g., User Retention)"]
    end
    subgraph Non-Experimental
        D["Observe IV & DV\n(e.g., Ice Cream Sales vs. Temperature)"] --> E["Analyze Correlation"]
    end

In the Real World

1. eSewa’s Fraud Detection

   • Design Used: Explanatory (Quasi-Experimental)
   • How: eSewa tested biometric verification (fingerprint + OTP) vs. traditional PIN-only logins. They found fraud cases dropped by 40% in biometric groups, leading to mandatory biometric checks.

2. Daraz’s Warehouse Location Strategy

   • Design Used: Descriptive + Exploratory
   • How: Daraz first mapped demand hotspots (descriptive) in Kathmandu, then tested a new warehouse in Bhaktapur (exploratory). Sales in Bhaktapur rose 30%, confirming the strategy.

3. NTC’s Fiber Optic Expansion

   • Design Used: Longitudinal Study
   • How: NTC tracked internet speed and adoption in 10 districts over 3 years. They found speeds doubled where fiber was rolled out, justifying nationwide expansion.

Choosing the Right Design: Decision Tree

flowchart TD
    A["Research Goal"] --> B{"Is the goal to explore?"}
    B -->|"Yes"| C["Use Exploratory\n(Interviews, Case Studies)"]
    B -->|"No"| D{"Is the goal descriptive?"}
    D -->|"Yes"| E["Use Descriptive\n(Surveys, Observations)"]
    D -->|"No"| F{"Can you manipulate variables?"}
    F -->|"Yes"| G["Use Experimental\n(A/B Tests, Control Groups)"]
    F -->|"No"| H["Use Non-Experimental\n(Correlation Studies)"]

Case Study: Himalayan Java’s Market Entry

Problem: Himalayan Java wanted to expand into instant coffee but lacked data on consumer preferences. Research Design:

1. Exploratory Phase:
   • Conducted focus groups in Kathmandu and Pokhara.
   • Found consumers preferred less sugar and organic ingredients.
2. Descriptive Phase:
   • Surveyed 1000 urban professionals on coffee habits.
   • 60% drank coffee 3+ times a week; 70% used instant coffee.
3. Explanatory Phase:
   • Tested two flavors (caramel vs. vanilla) in a quasi-experiment (sold in two different stores).
   • Vanilla outsold caramel by 25%, leading to its launch.

Outcome: Himalayan Java’s instant coffee line grew 40% in the first year.

Exam Tip

What Examiners Look For

1. Matching Design to Purpose:

   • Exploratory → "Investigate why..."
   • Descriptive → "Profile X population..."
   • Explanatory → "Test if Y causes Z..."
   • Example: If the question asks "Analyze the impact of social media ads on Pathao’s bookings," experimental or quasi-experimental is expected.

2. Strengths/Weaknesses:

   • Always discuss limitations (e.g., "Cross-sectional data cannot show trends").
   • Example Answer Snippet:

     "While the descriptive survey provided a snapshot of Khalti users’ demographics, it failed to capture how preferences change over time, limiting the study’s long-term applicability."

3. Real-World Application:

   • Link theories to companies. For example:

     "Like Ncell’s 4G rollout study, businesses should use longitudinal designs to track technology adoption trends."

4. Diagrams:

   • Draw flowcharts for research processes (e.g., exploratory → descriptive → explanatory).
   • Label variables clearly in causal diagrams (IV → DV).

5. Common Pitfalls:

   • ❌ Confusing correlation (non-experimental) with causation (experimental).
   • ❌ Ignoring control variables in causal studies.
   • Fix: Always state "All other factors (e.g., seasonality) were controlled."

Sample Exam Question & Answer

Question: "A bank wants to study whether offering interest rate discounts to customers who pay bills online reduces loan defaults. Suggest a research design and justify your choice."

Model Answer: The bank should use a quasi-experimental design with the following structure:

1. Independent Variable (IV): Interest rate discount for online bill payers.
2. Dependent Variable (DV): Loan default rate after 12 months.
3. Groups:
   • Treatment Group: Customers eligible for the discount.
   • Control Group: Customers not eligible (existing loan terms).
4. Data Collection: Track defaults for both groups over 1 year.
5. Analysis: Compare default rates using statistical tests (e.g., chi-square).

Justification:

• Quasi-experimental is suitable because random assignment is impractical (customers self-select into online bill payment).
• Strengths: Provides stronger evidence than non-experimental designs by isolating the IV’s effect.
• Limitations: Confounding variables (e.g., customer income) may still influence results.
• Real-World Tie: Similar to Nabil Bank’s financial literacy study, this design helps quantify the discount’s impact on risk.

Visual for Answer:

flowchart LR
    A["Interest Discount\n(IV)"] -->|"Treatment Group"| B["Online Bill Payers"]
    A -->|"No Discount"| C["Control Group\n(Traditional Payments)"]
    B & C --> D["Measure Default Rates\n(DV)"]
    D --> E["Compare Groups\n(Statistical Test)"]