Business Research Methods: Sampling, Analysis, and Ethics
Q1: Sampling Strategy and Sequential Design
Q1a: Sampling Strategy for the Proposed Study
Context: The study involves 10 lakh traders across diverse geographic locations and languages (English, Hindi, Marathi, Bengali, Kannada, Tamil, and Telugu), indicating a highly heterogeneous population.
Strategy: Stratified Sampling
Justification: Stratified sampling divides the population into meaningful subgroups (strata) based on characteristics like language, location, or business size. This ensures every element belongs to one stratum only. Stratification yields greater accuracy and is suitable when the population has variations that require representation. For this study, strata could be defined by regional language or trade association to ensure the sample accurately reflects the diverse group of traders.
Q1b: Sequential Sampling Design
Sequential sampling, also known as Double Sampling, involves drawing multiple stages of sampling.
Design/Process
A small initial sample is drawn to gather preliminary data. This first sample can be used to:
- Collect background material (pre-tasking).
- Determine eligibility (incidence rate).
- Reduce the cost of locating hard-to-find participants.
The information obtained from the first sample guides the design and selection criteria for the larger, second sample.
Justification for AIT/Meta Study
Sequential sampling could be used to:
- Determine the incidence rate (the proportion of the contacted population who are registered traders using WhatsApp Business and eligible for the MSBA program).
- Use a smaller, cheaper initial survey to refine the complex questionnaire or identify the optimal sampling frame before rolling out the full study to the large sample of 10 lakh traders.
Q2: MBajobs Chatbot Study – Data Analysis Strategy
Context
MBajobs transcribes 50 daily audio calls (5–10 minutes each) between interns and job seekers to develop a chatbot. This audio and transcript data is a source for qualitative research, focusing on capturing behaviors and communication patterns.
Data Analysis Strategy
Linguistic and Extra-Linguistic Analysis
The transcripts and audio recordings are analyzed for how people speak and what they say.
- Linguistic Analysis: Focus on the nature of information exchanged and interactions, possibly quantifying the frequency of specific words or phrases.
- Extra-linguistic Analysis: Study vocal features and temporal aspects (e.g., speed, pauses, interruptions).
Qualitative Coding and Thematic Analysis
To develop an effective chatbot, the researcher must understand the meaning and process of the conversations. This involves organizing textual data to identify key themes, challenges (critical incidents), or typical questions asked by job seekers, which will inform the chatbot’s design.
Non-verbal/Non-behavioral Analysis
While the calls are audio, the analysis of the records (transcripts, diaries, or Twitter data, if applicable) constitutes non-behavioral observation.
Q3: Mr. Chandler’s Database – Variables and T-Test Use
Context
Preparing a database from a hardcopy questionnaire, listing variables, their levels of measurement, and determining the appropriate use of a t-test.
Variables and Levels of Measurement
Variables are concrete, observable indicators, distinct from constructs (concepts assumed to have meaning, like “Best B School”).
1. Variable: A1: Ability to use WhatsApp to create groups
Levels of Measurement: 1. Don’t know | 2. I can manage | 3. Know very well. This is Ordinal measurement, as the numbers indicate a rank or order of ability, but the distance between the levels may not be equal.
2. Variable: Q1: Main purpose of using WhatsApp groups (A1, third column)
Levels of Measurement: “What is the main purpose?” This is likely a Nominal measure (Open-ended), providing a category or label without inherent order.
3. Variable: Q2/A2: Ranking of WhatsApp usefulness for contacting customers
Levels of Measurement: Rank 1 to 3 (or 1 to 3 in importance). This is Ordinal measurement, indicating relative preference (e.g., 1st, 2nd, 3rd).
How to Use a T-Test in This Data
A t-test is a statistical tool used to compare the means of two groups using quantifiable data. For appropriate use, the data analyzed should generally be derived from Interval or Ratio scales (or treated as such, like aggregated scores or means).
Application to the Data
If the researcher converts the Ordinal scale data (e.g., the 1–3 scores in A1 or the rankings in Q2) into a numerical average or a derived score, a t-test can be used to compare two distinct independent groups. Examples include:
- Comparing the mean ability score (A1) of male traders versus female traders.
- Comparing the average rank of usefulness (Q2) between traders in urban areas versus those in rural areas.
I. Business Research (BR) Fundamentals
BR is defined as the process of determining, acquiring, analyzing, synthesizing, and disseminating relevant business data, information, and insights to decision-makers, aiming to maximize business performance. The core sequence of BR is to Collect → Analyze → Insights → Business Performance.
The need for BR includes troubleshooting market and organizational problems, facing competition, sustaining business, retaining employees, growing into new geographies/products/services, and measuring the impact/ROI of CSR and Marketing campaigns.
Research may be conducted:
- In-House (Marketing, HR, R&D).
- Outsourced to private consulting/market research firms.
- By the Government or by Trade Associations.
The decision on who conducts the research is determined by Time, Money, and Personnel.
II. The Research Process and Design
The research process begins by Defining the need (Information needed? Solution required?), deciding on Primary & Secondary, Qualitative & Quantitative approaches, Sampling, and Tools. Next is Design the method, choosing Secondary data, Survey, Observation/fieldwork, or Experiments. This is followed by Collect Data, then Analyze (Tabulations, Interpretations), and finally Present (Reports, PPTs).
The Research Design is the blueprint for data collection, measurement, and analysis. It functions like a research proposal, covering the Question, Nature of data, Collection & analysis methods, Time, Budget, and People involved.
III. Core Concepts and Approaches
- Constructs: Concepts with assumed meanings (e.g., Best B School, Ghosting).
- Variables: Concrete, observable indicators (e.g., Smoking related to cancer).
- Inductive Research: Moves from specifics (Quotes) to general theory.
- Deductive Research: Tests theory using specifics (Pie charts).
- Qualitative Research (QR): Focuses on the how (process) and why (meaning), seeking in-depth understanding.
- Quantitative Research: Focuses on what and how often happened.
IV. Ethics in Research
Ethical requirements mandate no deception; if deception is necessary, debriefing must be used. Informed consent and protection of the Right to Privacy are mandatory.
Researchers must maintain Sponsors’ confidentiality (e.g., not revealing the name of an MNC client). It is unethical to engage in Cooking data for or by clients, such as manipulating data regarding CSR impact by mining firms.
V. Secondary Data
Secondary data is data not collected by the researcher for this specific purpose. It is used in the Exploratory phase to:
- Expand understanding.
- Look for proven solutions.
- Refine research questions.
- Save cost and time.
Data Hierarchy:
- Primary sources: Raw data with no inferences (e.g., NSSO Data).
- Secondary sources: Interpretations of data (e.g., Tourism report).
- Tertiary sources: Interpretations of secondary sources (e.g., Ranking of best companies).
VI. Qualitative Research (QR) Methods
QR provides in-depth understanding of process and meaning and can reinforce quantitative studies in Mixed designs (Triangulation).
- Interviews: A two-way process used for probing respondents for insights. Types include Unstructured, Semi-structured, and Structured, as well as in-depth methods like life histories and the critical incident technique.
- Focus Group Discussions (FGDs): Involve about 10 participants, where the researcher acts as the moderator, providing lead questions while participants speak and dominate the discussion.
- Projective Techniques: Use imagination to uncover attitudes, such as Word or picture association, Sentence completion, or the Imagination exercise.
- Participant Observation: The researcher is in the field, performing either Direct observation (just seeing/noting) or Participate & Observe (deciding whether to reveal or hide identity).
VII. Observation Studies
Observation is useful because it is Not dependent on the respondent, captures mundane but important data, and allows for Real-time capture in a natural environment.
Observation includes:
- Behavioral Observations: Non-verbal analysis (body language), Linguistic analysis (nature of information), Extra-linguistic analysis (vocal/temporal features), and Spatial analysis (proxemics/distance).
- Non-Behavioral Data: Record analysis (Twitter & diaries), Physical condition analysis (Store audits), and Process analysis (File movement).
Unobtrusive measures, like video cameras or looking for residues, are used to avoid the reactivity response (participants changing behavior when watched).
VIII. Survey Methods and Errors
Surveys move research from exploratory to descriptive, analytical, and explanatory phases. They use a structured tool and are suitable for Large-scale, quantitative data collection.
Key errors include:
- Interviewer error: Inaccurate recording or inappropriate influences.
- Participant error: Not having information or not understanding their role.
- Non-Response error: Poor response rate.
- Response error: Incorrect responses or social desirability bias.
IX. Sampling Techniques
Sampling is preferred because it provides Greater accuracy, Greater speed, and Lower cost compared to studying the entire population.
- Probability Sampling Techniques (Random Selection): Simple Random Sampling, Systematic Sampling, Stratified Sampling, Cluster (area) Sampling, and Double (Sequential) Sampling.
- Non-Probability Sampling Techniques (Non-Random Selection): Convenience Sampling, Judgmental Sampling, Quota Sampling, and Snowball Sampling.
X. Case Specifics (AIT/Meta Study)
The AIT/Meta partnership aims to upskill 10 lakh traders on WhatsApp Business over three years, covering language modules and examinations in English, Hindi, Marathi, Bengali, Kannada, Tamil, and Telugu.
The research should aim to understand the program’s impact and whether MSBA certification will particularly benefit entrepreneurs who are just starting out. The requirements necessitate recommendations for Sampling Strategy (e.g., Stratified Sampling for the heterogeneous population) and Sequential sampling design (Double Sampling).