Research Design: Types, Selection & Application

TAKEAWAYS:

• Research design is the blueprint of a study, guiding data collection, analysis, and interpretation.
• The four main types (exploratory, descriptive, explanatory, experimental) differ in purpose, methods, and rigor.
• Experimental designs (true experiments, quasi-experiments) allow causal inferences but require strict control.
• Non-experimental designs (descriptive, correlational) observe relationships without manipulation.
• Mixed-methods designs combine quantitative and qualitative approaches for richer insights.
• Real-world IT applications (e.g., A/B testing in apps, user behavior studies) rely on well-structured research designs.

1. What is Research Design?

Research design is the framework that structures how a study is conducted. It defines:

• Research questions to be answered.
• Data collection methods (surveys, experiments, observations).
• Sampling strategies (random, stratified, convenience).
• Data analysis techniques (statistical tests, thematic analysis).

A well-designed study ensures validity (measuring what it claims) and reliability (consistent results).

2. Types of Research Design

Research designs are categorized based on purpose, control, and time frame. The four primary types are:

3. Exploratory Research Design

Definition: Used when the researcher lacks clear understanding of a problem. It explores ideas, generates hypotheses, and identifies variables.

Methods Used:

• Literature reviews (secondary data).
• Pilot studies (small-scale tests).
• Qualitative techniques (interviews, focus groups, case studies).

Example: Daraz’s New Delivery Model

Problem: Daraz wants to test a same-day delivery feature in Kathmandu but lacks data on feasibility. Approach:

1. Conduct focus group discussions with 20 customers to gauge interest.
2. Analyze past delivery data to estimate costs.
3. Run a pilot in Thapathali with 100 users. Outcome: Identifies challenges (e.g., traffic delays) before full rollout.

4. Descriptive Research Design

Definition: Aims to describe characteristics of a population or phenomenon. It answers "what," "how much," or "how often."

Methods Used:

• Surveys (structured questionnaires).
• Observational studies (direct or indirect).
• Case studies (in-depth analysis of a single entity).

Example: NTC’s Internet Speed Study

Problem: NTC wants to measure average internet speed across Nepal. Approach:

1. Random sampling of 500 users in 10 districts.
2. Speed tests conducted at peak hours (6–9 PM).
3. Data analysis to report mean speed and variability. Outcome: Publishes a descriptive report on internet performance by region.

sequenceDiagram
    participant NTC as NTC Researcher
    participant User as Sampled User
    NTC->>User: Distributes speed-test link
    User->>NTC: Runs test (6-9 PM)
    NTC->>NTC: Records data (Mbps)
    NTC->>NTC: Analyzes mean/median speed
    NTC->>Public: Releases report

5. Explanatory Research Design

Definition: Investigates why or how variables are related. It seeks causal explanations (though not always conclusive).

Methods Used:

• Correlational studies (measuring relationships).
• Regression analysis (predicting outcomes).
• Longitudinal studies (tracking changes over time).

Example: Khalti’s User Retention Study

Problem: Khalti notices high dropout rates after the first transaction. Approach:

1. Survey 1,000 users on reasons for leaving (e.g., fees, app crashes).
2. Run a regression analysis to identify top predictors (e.g., "users who see ads are 30% more likely to leave").
3. Recommendations: Reduce ads, improve onboarding. Outcome: Explains why users leave and guides policy changes.

6. Experimental Research Design

Definition: The gold standard for causal inference. It involves manipulating an independent variable to observe its effect on a dependent variable.

Types of Experimental Designs:

Key Components:

1. Independent Variable (IV): The factor being changed (e.g., ad color in an app).
2. Dependent Variable (DV): The outcome measured (e.g., click-through rate).
3. Control Group: Receives no treatment (baseline for comparison).
4. Experimental Group: Receives the treatment.

Example: eSewa’s Discount Incentive Problem: eSewa wants to know if discounts increase app usage. Design:

• IV: 20% discount on bills for Group A; no discount for Group B.
• DV: Number of logins in 30 days.
• Result: Group A shows a 25% increase in logins (causal link established).

flowchart TD
    A["Random Assignment"] --> B["Group A: 20% Discount"]
    A --> C["Group B: No Discount"]
    B --> D["Measure Logins"]
    C --> D
    D --> E["Compare DV: Group A > Group B"]

7. Non-Experimental vs. Experimental Designs

8. Mixed-Methods Research Design

Definition: Combines quantitative (numbers, stats) and qualitative (words, themes) methods to provide a holistic view.

Approaches:

1. Sequential: Quantitative → Qualitative (or vice versa).
   • Example: Survey 1,000 users (quantitative) → Interview 10 (qualitative) to explain survey results.
2. Convergent: Run both simultaneously and compare.
   • Example: Track NEPSE stock trends (quantitative) + interview traders (qualitative) to explain volatility.
3. Embedded: One method dominates; the other supports it.
   • Example: Main study = survey (quantitative); embedded = case studies of top-performing stocks.

Example: Pathao’s Driver Satisfaction Study

1. Quantitative: Survey 5,000 drivers on income, hours worked.
2. Qualitative: Interview 50 drivers to understand why some earn more.
3. Findings: Low earnings linked to lack of training (qualitative insight) + traffic delays (quantitative data).

9. Selecting the Right Research Design

Choosing a design depends on:

• Research question: Exploratory? Descriptive? Causal?
• Resources: Time, budget, expertise.
• Ethics: Can you manipulate variables? (e.g., testing a drug vs. studying user behavior).
• Context: Real-world vs. lab setting.

Decision Tree:

10. Real-World Applications in Nepal & Globally

In Nepal:

1. eSewa’s Fraud Detection

   • Design: Experimental (A/B test).
   • How: Random users get biometric login; others use passwords. Compare fraud rates.
   • Outcome: Biometric login reduces fraud by 40%.

2. NTC’s Smart Grid Pilot

   • Design: Quasi-experimental.
   • How: Install smart meters in Bhaktapur (treatment) vs. Kathmandu (control). Compare energy savings.
   • Outcome: Smart meters cut waste by 15% in Bhaktapur.

3. Daraz’s Warehouse Optimization

   • Design: Mixed-methods.
   • How:
     • Quantitative: Track order fulfillment times.
     • Qualitative: Interview warehouse staff on bottlenecks.
   • Outcome: Redesigns layout, reducing delays by 30%.

Global Examples:

1. Google’s PageRank Algorithm

   • Design: Experimental (tested on live search data).
   • How: Compared ranking methods to see which predicted user clicks best.

2. WhatsApp’s End-to-End Encryption

   • Design: Quasi-experimental.
   • How: Rolled out in phases; monitored message security vs. unencrypted chats.

3. Netflix’s Recommendation System

   • Design: Mixed-methods.
   • How:
     • Quantitative: Tracks watch time, clicks.
     • Qualitative: Surveys users on why they like/dislike recommendations.

11. Common Pitfalls & How to Avoid Them

12. Step-by-Step: Designing a Research Study

Example: "Does using dark mode in a mobile app reduce eye strain?"

1. Define the Problem:

   • Research Question: Does dark mode reduce eye strain in low-light conditions?

2. Choose Design:

   • Experimental (manipulate screen mode; measure eye strain).

3. Operationalize Variables:

   • IV: Screen mode (light vs. dark).
   • DV: Eye strain (measured via survey scale or device sensor data).

4. Select Participants:

   • Sample: 200 users aged 18–35, with normal vision.
   • Random assignment: Half get dark mode; half get light mode.

5. Data Collection:

   • Tool: App with built-in eye strain tracker + post-use survey.
   • Duration: 1 week of usage.

6. Analyze Data:

   • Quantitative: Compare mean eye strain scores (t-test).
   • Qualitative: User feedback on comfort.

7. Draw Conclusions:

   • If dark mode group reports 20% less strain, conclude causal link.

Exam Tip

How This Unit is Tested:

1. Definitions & Differences:

   • Expect questions like "Distinguish between exploratory and descriptive research designs" or "When would you use a quasi-experiment?"
   • Tip: Memorize the purpose, methods, and examples of each design type.

2. Application-Based Questions:

   • "How would you design a study to test if Khalti’s new OTP system reduces fraud?"
   • Tip: Use the step-by-step design process (problem → design → variables → sampling → analysis).

3. Critical Analysis:

   • "A study claims that using WhatsApp reduces stress. Identify the design flaws."
   • Tip: Look for lack of control group, poor sampling, or correlation ≠ causation.

4. Diagrams & Tables:

   • You may be asked to draw a research design flowchart or compare two designs in a table.
   • Tip: Practice sketching experimental vs. non-experimental setups quickly.

5. Real-World Scenarios:

   • "How does Daraz use research design to improve delivery times?"
   • Tip: Link to mixed-methods (quantitative data + qualitative feedback).

Marks Distribution:

• Definitions: 20%
• Examples/Applications: 30%
• Comparison Tables/Diagrams: 20%
• Critical Analysis: 30%

Final Checklist Before Exam: ✅ Can you name and explain the 4 main research designs? ✅ Can you pick the right design for a given scenario (e.g., "testing a new feature")? ✅ Can you identify flaws in a poorly designed study? ✅ Can you draw a simple experimental setup (IV, DV, groups)? ✅ Do you know 2 real-world IT/business examples for each design type?