Research Design: Types, Models & Applications in Business

What is Research Design?

Research design is the blueprint that guides how a study will be conducted. It determines:

• What data to collect (quantitative, qualitative, or mixed).
• How to collect it (surveys, experiments, observations).
• When and where to collect it (field vs. lab, cross-sectional vs. longitudinal).
• How to analyze it to answer the research question.

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

Types of Research Design

Research designs are broadly categorized into three main types, each serving different research objectives:

1. Exploratory Design

Purpose: To gain insight, understand phenomena, or identify new research problems. When to use: When the research problem is vague or not well-defined (e.g., "Why are customers switching from Ncell to NTC?"). Methods:

• Literature reviews
• Case studies
• Pilot studies
• Focus group discussions

Example: A startup like Pathao might use exploratory design to understand why riders abandon the app mid-trip. They could conduct interviews with 20-30 users to identify common pain points (e.g., unclear pricing, long wait times).

2. Descriptive Design

Purpose: To describe characteristics of a population or phenomenon (who, what, where, when, how much). When to use: When the goal is to measure or profile (e.g., "What is the average monthly spending of Daraz customers in Kathmandu?"). Methods:

• Surveys
• Observations
• Archival data analysis

Example: Khalti might use descriptive design to profile its user base:

• Demographics: Age, gender, location.
• Behavior: Frequency of transactions, preferred payment methods.
• Satisfaction: Net Promoter Score (NPS) via surveys.

3. Causal (Explanatory) Design

Purpose: To explain cause-and-effect relationships (e.g., "Does increasing ad spend on Facebook boost eSewa’s user sign-ups?"). When to use: When testing hypotheses about interventions or treatments. Methods:

• Experimental designs (randomized control trials)
• Quasi-experimental designs (no random assignment)
• Time-series analysis

Example: A bank like Nabil Bank might test whether sending personalized SMS reminders reduces loan default rates. They could:

1. Randomly assign 100 loan customers to a control group (no SMS).
2. Assign 100 others to a treatment group (weekly SMS reminders).
3. Compare default rates after 6 months.

Visual: Research Design Classification

mindmap
  root((Research Design Types))
    Exploratory
      "Purpose: Gain insight"
      "Methods: Literature review, case studies, pilot studies"
      "Example: Pathao rider drop-off analysis"
    Descriptive
      "Purpose: Describe characteristics"
      "Methods: Surveys, observations, archival data"
      "Example: Khalti user profiling"
    Causal
      "Purpose: Explain cause-effect"
      "Methods: Experiments, quasi-experiments"
      "Example: Nabil Bank loan default study"

Experimental vs. Non-Experimental Designs

Key Research Design Models

1. Cross-Sectional vs. Longitudinal Design

Example Trace (Cross-Sectional): Suppose Daraz wants to know why 30% of orders are abandoned at checkout. They design a one-time survey sent to 1,000 users who abandoned carts. The survey asks:

• Did you face technical issues?
• Was the shipping cost too high?
• Were you comparing prices elsewhere?

Analysis: 60% cite "high shipping costs" → Daraz introduces free shipping over $500.

2. Qualitative vs. Quantitative Design

Example Trace (Qualitative): Nabil Bank wants to improve its mobile app. They conduct 5 focus group discussions with:

• Young professionals (25–35 years)
• Retirees (60+ years)
• Rural users

Findings:

• Young users want faster transactions.
• Rural users struggle with low internet connectivity.
• Retirees prefer voice-based navigation.

Action: Bank redesigns app with offline mode and voice commands.

In the Real World

1. Daraz’s Order Fulfillment Design

   • Problem: High customer complaints about delayed deliveries.
   • Design Used: Mixed-methods (quantitative survey + qualitative interviews).
   • How:
     • Quantitative: Surveyed 5,000 customers on delivery times.
     • Qualitative: Interviewed warehouse staff to identify bottlenecks.
   • Outcome: Redesigned last-mile delivery routes using data analytics.

2. Khalti’s Fraud Detection System

   • Problem: Rising fraudulent transactions.
   • Design Used: Longitudinal experimental design.
   • How:
     • Tested real-time transaction alerts vs. no alerts for 1,000 users.
     • Found alerts reduced fraud by 40% → implemented system-wide.

3. NTC’s Customer Satisfaction Index (CSI)

   • Problem: Declining CSI scores.
   • Design Used: Cross-sectional survey with weighted scoring.
   • How:
     • Surveyed 10,000 customers on service speed, billing accuracy, complaint resolution.
     • Weighed responses by customer segment (residential vs. corporate).
   • Outcome: Launched 24/7 chat support and automated billing checks.

Case Study: Toyota’s Kaizen Research Design

Problem: Toyota wanted to reduce defects in its production line. Research Design:

1. Exploratory Phase:
   • Observed workers to identify common mistakes (e.g., misaligned parts).
   • Conducted interviews with assembly line employees.
2. Descriptive Phase:
   • Collected data on defect rates per shift.
   • Mapped defect locations on the assembly line.
3. Causal Phase:
   • Tested new training methods (experimental group) vs. old methods (control group).
   • Found interactive training reduced defects by 30% → implemented globally.

Key Takeaway: Toyota used a sequential mixed-methods design to move from exploration → description → causation.

Step-by-Step: Designing a Research Study

Let’s design a study for: "How does WhatsApp Business API adoption affect SMEs’ customer response time in Nepal?"

1. Define Objective:

   • Exploratory: Understand current adoption levels.
   • Descriptive: Measure response time before/after API use.
   • Causal: Test if API reduces response time.

2. Choose Design:

   • Quasi-experimental (no random assignment; compare SMEs using API vs. not using it).

3. Data Collection:

   • Survey: 200 SMEs (100 using API, 100 not).
   • Metrics: Average response time (pre-API: 24h → post-API: 2h).

4. Analysis:

   • t-test to compare response times.
   • Regression analysis to control for company size.

5. Conclusion:

   • Hypothesis: "API adoption reduces response time by ≥50%."
   • Result: Confirmed (response time dropped by 60%).

Exam Tip

1. Define Clearly:

   • Examiners check if you distinguish between exploratory, descriptive, and causal designs. Always state the purpose of your chosen design.

2. Link to Real-World Examples:

   • 20% of exam marks often come from applying concepts to Nepali businesses (e.g., Daraz, Nabil Bank, NTC). Prepare 2–3 case studies in advance.

3. Diagrams > Text:

   • Draw flowcharts for research processes (e.g., "How Khalti designs a survey").
   • Use tables to compare designs (e.g., cross-sectional vs. longitudinal).

4. Common Pitfalls:

   • ❌ Saying "survey = descriptive design" (it can also be exploratory).
   • ❌ Ignoring ethical considerations (e.g., informed consent in experiments).
   • ❌ Overlooking sampling bias (e.g., surveying only urban Daraz users).

5. Shortcut for Full Marks:

   • For 5-mark questions, use the "Purpose-Methods-Example" (PME) formula:

     "This is a descriptive design because its purpose is to measure customer satisfaction. The method used is a cross-sectional survey of 500 eSewa users. For example, Nepal Telecom uses similar designs to track Net Promoter Scores quarterly."

Practice Question with Solution

Question: "Differentiate between experimental and non-experimental designs with a business example for each."

Solution:

Visual:

flowchart TD
    A["Research Question"] --> B{"Can IV be Manipulated?"}
    B -->|"Yes"| C["Experimental Design<br/>Example: A/B test on eSewa’s checkout button"]
    B -->|"No"| D["Non-Experimental Design<br/>Example: Surveying Pathao riders’ wait times"]
    C --> E["Random Assignment<br/>Control vs. Treatment Group"]
    D --> F["Observe Natural Behavior<br/>No Intervention"]

Key Formulas and Checklists

1. Validity Checklist:

   • Internal Validity: Did the design rule out alternative explanations? (Critical for causal designs.)
   • External Validity: Can results be generalized? (e.g., surveying only Kathmandu limits external validity for Nepal-wide claims.)
   • Reliability: Would the same design yield similar results if repeated?

2. Design Selection Flowchart:

flowchart LR
    A["Start"] --> B{"Is the problem well-defined?"}
    B -->|"No"| C["Use Exploratory Design<br/>Pilot study, interviews"]
    B -->|"Yes"| D{"Do you need to explain cause-effect?"}
    D -->|"Yes"| E["Use Causal Design<br/>Experiment or quasi-experiment"]
    D -->|"No"| F{"Do you need trends over time?"}
    F -->|"Yes"| G["Use Longitudinal Design<br/>Repeat surveys"]
    F -->|"No"| H["Use Cross-Sectional Design<br/>One-time survey"]

Final Summary Table

Exam-Ready Bullet Points

• Research design is the "skeleton" of a study—it determines what, how, and when data is collected.
• Exploratory = "What’s the problem?"; Descriptive = "What is it?"; Causal = "Why does it happen?"
• Experimental designs are gold standard for causality but require randomization and control groups.
• Non-experimental designs are cheaper but prone to confounding variables.
• Cross-sectional = one shot; Longitudinal = repeat over time.
• Qualitative = depth; Quantitative = scale.
• Real-world link: Daraz uses descriptive surveys for trends, Nabil Bank uses causal experiments for loan strategies, and Pathao uses exploratory interviews for app improvements.