Survey Research — A Comprehensive Guide
About This Chapter
This chapter provides a comprehensive foundation in survey research methodology, designed to serve students across both business/social science (BBA) and computer science (BCA) disciplines. Survey research is one of the most widely used quantitative methods in capstone projects — yet it is also one of the most frequently misapplied. This guide covers the complete survey research lifecycle: from topic selection and population definition through sampling design, questionnaire construction, data collection, statistical analysis, and reporting. It includes 20 sample research topics — 10 from business and 10 from computer science — to illustrate the range of questions that survey methods can address across disciplines.
Whether you are studying consumer behaviour, employee engagement, technology adoption, or developer practices, the principles and procedures in this chapter will help you design and execute a rigorous survey-based capstone.
1. Foundations of Survey Research
Survey research is a systematic method for collecting standardised information from a sample of individuals drawn from a defined population. It uses structured questionnaires or interview schedules to measure attitudes, beliefs, behaviours, characteristics, and experiences. The defining features of survey research are: (a) standardisation — every respondent receives the same questions in the same way, enabling comparison and aggregation; (b) sampling — data is collected from a subset of the population, with the goal of drawing inferences about the population; and (c) quantification — responses are coded numerically, enabling statistical analysis of patterns, relationships, and differences.
1.1 When Survey Research is the Right Method
| Survey Research is Appropriate When... | Survey Research is NOT Appropriate When... |
|---|---|
| Your RQs ask about attitudes, beliefs, perceptions, self-reported behaviours, or characteristics that can be measured through standardised questions | Your RQs require deep exploration of meaning, lived experience, or processes that unfold over time — these require qualitative methods (interviews, ethnography) |
| You need to describe the prevalence, distribution, or relationships among variables in a defined population | You need to establish causality with high internal validity — surveys can identify associations but cannot manipulate variables. Experiments are required for causal claims. |
| You can access a sufficient sample of your target population and can achieve an adequate response rate | Your population is inaccessible, hidden, or unwilling to respond to surveys — alternative methods (secondary data, observation) may be more feasible |
| You need data that is comparable across respondents, enabling statistical generalisation (with appropriate sampling) | Your constructs are poorly understood and need exploratory qualitative work before they can be measured — consider an exploratory sequential mixed-methods design |
| You can use or adapt validated measurement instruments (scales, indices) for your key constructs | No validated instruments exist for your constructs and you lack the time/resources to develop and validate a new scale — consider qualitative exploration instead |
1.2 Survey Research Across Disciplines
| Dimension | Business / Social Science (BBA) | Computer Science / IT (BCA) |
|---|---|---|
| Typical RQs | What factors influence consumer purchase intention? How does leadership style affect employee engagement? What is the relationship between CSR perception and brand loyalty? | What factors affect user adoption of a new software system? How do developers perceive the usefulness of AI coding assistants? What are the barriers to cloud migration among IT managers? |
| Common Constructs | Satisfaction, trust, loyalty, engagement, motivation, perceived value, brand attitude, organisational commitment, job performance | Perceived usefulness, perceived ease of use (TAM), technology acceptance, system usability (SUS), trust in automation, privacy concern, self-efficacy, technology readiness |
| Typical Populations | Consumers, employees, managers, students, citizens, patients, investors, entrepreneurs | Software developers, IT professionals, system users, technology adopters, online community members, data scientists, cybersecurity professionals |
| Common Theoretical Frameworks | TAM, UTAUT/UTAUT2, TRA, TPB, SERVQUAL, Source Credibility Theory, Social Exchange Theory, Stakeholder Theory | TAM, UTAUT, IS Success Model (DeLone & McLean), Task-Technology Fit, Diffusion of Innovation, Technology Readiness Index, Trust in Automation |
2. Population, Sample, and Sampling in Survey Research
The relationship between population and sample is the conceptual foundation of survey research. Every survey study is an exercise in inference: you collect data from a sample and use it to draw conclusions about a population. The quality of that inference depends entirely on how the sample was selected and how well it represents the population.
2.1 Defining the Population — Precision is Everything
Target Population: The complete set of individuals, organisations, or entities to whom you wish to generalise your findings. Must be defined precisely in terms of: (a) who (demographic, professional, or behavioural characteristics), (b) where (geographic or institutional boundaries), and (c) when (time period). Sampling Frame: The list or mechanism from which the sample is actually drawn — the operationalisation of the population. The sampling frame is almost always imperfect (some population members missing, some non-members included). Sample: The subset of the sampling frame from whom data is actually collected. The achieved sample is almost always smaller than the planned sample due to non-response.
| Element | Vague (Inadequate) | Precise (Required) |
|---|---|---|
| Target Population | "Indian consumers" | "Smartphone users aged 18–45 in Tier-1 and Tier-2 Indian cities who have made at least one online purchase in the preceding 6 months" |
| Sampling Frame | "People I can reach online" | "Registered users of [specific online consumer panel] who meet the screening criteria; supplemented by social media recruitment through LinkedIn and Instagram ads targeting the specified demographic" |
| Planned Sample | "About 200 respondents" | "Target n = 250 (accounting for ~20% expected incomplete/unusable responses, yielding ~200 complete responses for analysis). Minimum required for planned regression with 5 predictors: n = 98 (Green, 1991); target n = 200 ensures adequate power for medium effect sizes." |
| Achieved Sample | "187 responses were collected" | "Of 250 initiated responses, 218 were completed (87.2% completion rate). After screening for quality (completion time, attention checks, straightlining), 12 were excluded, yielding a final analytical sample of n = 206. Mean age 31.4 years (SD = 8.2); 52% female; 68% from Tier-1 cities." |
2.2 Probability Sampling — The Gold Standard
Probability sampling gives every element in the population a known, non-zero chance of selection. It is the only sampling approach that supports statistical generalisation from sample to population with quantifiable error. When feasible, it should be your first choice.
| Method | How It Works | Best For | Capstone Feasibility | Business Example | Computer Science Example |
|---|---|---|---|---|---|
| Simple Random Sampling (SRS) | Every element has an equal probability of selection. Random number generator selects from complete frame. | When you have a complete, accurate sampling frame and want the simplest probability design | High if frame exists (employee database, customer list, student registry). Low if no frame. | Randomly selecting 200 employees from an HR database of 2,500 | Randomly selecting 150 registered users from an app's user database |
| Stratified Random Sampling | Divide population into strata (groups), then randomly sample within each stratum. Ensures representation of subgroups. | When subgroups are important to your RQs and you need to ensure their representation; when you want to compare groups | Moderate — requires knowledge of strata proportions. More complex to execute than SRS. | Stratifying by department (Marketing, Finance, IT, HR) and sampling proportionally within each | Stratifying by user type (free, premium, enterprise) and sampling proportionally |
| Cluster Sampling | Divide population into clusters (natural groups), randomly select clusters, then sample within selected clusters. | When the population is geographically dispersed; when a complete frame is unavailable but clusters can be identified | Moderate — reduces travel/admin costs but increases sampling error (clusters tend to be internally homogeneous) | Randomly selecting 8 of 40 retail stores, then surveying all employees in selected stores | Randomly selecting 5 of 25 software teams, then surveying all developers in selected teams |
| Systematic Sampling | Select every k-th element from a list after a random start. | When you have a list and want a simpler alternative to SRS; ensures spread across the frame | High if list exists. Caution: if the list has periodic patterns, bias can be introduced. | Selecting every 12th customer from a CRM database of 3,000 | Selecting every 8th commit author from a repository's contribution log |
2.3 Non-Probability Sampling — When Probability Sampling is Not Feasible
Non-probability sampling does not give every element a known chance of selection, and findings cannot be statistically generalised to the population. However, it is often the only feasible approach in capstone research — and when executed well and reported honestly, it can produce valuable evidence.
| Method | How It Works | When to Use | Limitation to Acknowledge |
|---|---|---|---|
| Purposive Sampling | Deliberately select participants who meet specific, predefined criteria relevant to your RQs | When you need participants with specific characteristics, experiences, or expertise; when the population is too specific for random sampling to efficiently reach | "Participants were purposively selected. Findings reflect the perspectives of this specific group and may not generalise to [broader population]." |
| Quota Sampling | Set quotas for subgroups (e.g., 50 male, 50 female) and use convenience to fill each quota | When you want to ensure representation of specific subgroups but cannot use probability methods; common in market research | "Quota sampling ensured demographic diversity, but within quotas, participants were convenience-sampled. Findings are indicative, not representative." |
| Snowball Sampling | Start with a few participants who meet criteria; ask them to refer others | When the population is hidden, hard to access, or lacks a sampling frame (e.g., gig workers, niche technology users, entrepreneurs) | "Snowball sampling may over-represent individuals with larger professional networks. Findings are exploratory and not generalisable." |
| Convenience Sampling | Select participants based on accessibility — students, social media followers, professional network | When no other feasible approach exists within capstone constraints. Acceptable for exploratory or pilot research. | "This study used a convenience sample. Generalisability to the broader population is limited. Findings should be interpreted as exploratory." |
2.4 Sample Size Determination
| Analytical Method | Minimum Recommendation | Calculation Method | Notes |
|---|---|---|---|
| Descriptive statistics only | n ≥ 100 (reasonable precision); n ≥ 385 (population estimate, 5% margin of error, 95% CI) | Sample size formula for proportions: n = Z²p(1−p)/e² where Z = 1.96 (95% CI), p = 0.5, e = margin of error | Add 20% for expected non-response |
| Correlation | n ≥ 80–100 for detecting medium effects (r = 0.30) with 80% power | Power analysis (G*Power): test = correlation, effect = 0.30, α = .05, power = .80 | Larger samples needed for small effects |
| Multiple Regression | n ≥ 50 + 8k where k = number of predictors (Green, 1991) | For 5 predictors: n ≥ 90. Power analysis recommended: f² = 0.15 (medium), α = .05, power = .80 | Green's rule is a minimum. Larger samples improve stability of estimates. |
| Independent t-test | n ≥ 64 per group (medium effect d = 0.50, power = .80) | Power analysis (G*Power): test = independent t-test, d = 0.50, α = .05, power = .80 | n ≥ 30 per group allows CLT to apply for normality |
| One-way ANOVA (3 groups) | n ≥ 159 total (~53 per group for medium effect f = 0.25) | Power analysis (G*Power): test = ANOVA, f = 0.25, α = .05, power = .80 | More groups → larger total sample required |
| Factor Analysis / SEM | n ≥ 200 minimum; n ≥ 10 per indicator; n ≥ 20 per construct | More complex — rules of thumb vary; Monte Carlo simulation preferred | SEM is sample-hungry; ensure your capstone can achieve the required n |
Most capstone samples will not meet the ideal requirements for probability sampling and may fall short of formal sample size recommendations. This is not disqualifying — but it must be acknowledged. A convenience sample of 120 analysed with regression, where you honestly state "the non-probability sample limits generalisability, and findings should be considered exploratory," is methodologically honest. The same sample presented as representative of "Indian consumers" is not. The evaluation committee accepts practical constraints. They do not accept misrepresentation.
3. Survey Research Methods and Designs
3.1 Survey Administration Modes
| Mode | Description | Strengths | Limitations | Response Rate (Typical) | Best For |
|---|---|---|---|---|---|
| Online / Web Survey | Self-administered via platforms (Google Forms, Qualtrics, SurveyMonkey, LimeSurvey). Distributed by email, social media, or embedded in websites. | Low cost; fast; large geographic reach; automated data capture; skip logic and randomisation; anonymity easy to ensure; multimedia possible | Sampling frame often absent; self-selection bias; lower response rates than interviewer-administered; digital divide excludes non-users; survey fatigue | 10–30% (external); 30–60% (internal/organisational) | Most capstone surveys. Default recommendation for BBA and BCA unless there is a reason to use another mode. |
| Face-to-Face Interview | Trained interviewer administers questionnaire in person. Can be structured (fixed questions) or semi-structured. | Highest response rates; can clarify questions; can use visual aids; longer questionnaires possible; richer data | Expensive; time-consuming; geographically limited; interviewer effects (social desirability bias); harder to ensure anonymity | 50–80% | When the population is accessible in one location; when questionnaire is complex; when high response rate is essential |
| Telephone Survey | Interviewer-administered by phone. CATI (Computer-Assisted Telephone Interviewing) systems manage the process. | Faster than face-to-face; broader geographic reach; can clarify questions; CATI ensures skip logic | Declining response rates (robocalls, spam); limited to shorter questionnaires (15–20 min max); no visual aids; mobile-only households challenging | 10–30% | When phone numbers are available and the population is accessible by phone; declining in relevance for capstone research |
| Mail / Postal Survey | Paper questionnaire mailed to respondents with return envelope. | Can reach populations without internet; perceived as more legitimate by some populations; no interviewer bias | Slow; expensive (printing, postage); low response rates without incentives; no skip logic; manual data entry required | 5–20% (without follow-up); 20–50% (with follow-up) | Rarely used in capstone research. May be appropriate for older or rural populations with limited internet access. |
3.2 Cross-Sectional vs. Longitudinal Survey Designs
| Dimension | Cross-Sectional Design | Longitudinal Design |
|---|---|---|
| What It Is | Data collected at ONE point in time from a sample of the population | Data collected from the SAME sample at MULTIPLE time points |
| What It Measures | Prevalence, relationships, and differences at a single time point — a snapshot | Change, stability, and trends over time — a moving picture |
| Causal Claims | Cannot establish causality — temporal precedence cannot be determined. "X is associated with Y" — not "X causes Y." | Can establish temporal precedence (X precedes Y), strengthening (though not proving) causal inference |
| Feasibility for Capstone | High — the default for capstone survey research. Data collection completed in 2–6 weeks. | Low — requires multiple data collection waves months apart. Rarely feasible within a capstone timeline unless using existing panel data. |
3.3 Survey Design Typology — Choosing Your Design
| Design Type | Purpose | Example RQ (Business) | Example RQ (CS) |
|---|---|---|---|
| Descriptive Survey | Describe the characteristics, attitudes, or behaviours of a population | "What is the prevalence of mobile payment adoption among small retailers in Tier-2 Indian cities?" | "What proportion of software developers in Indian startups use AI-assisted coding tools, and for which tasks?" |
| Correlational Survey | Examine relationships among two or more variables without manipulation | "What is the relationship between employee engagement, perceived organisational support, and turnover intention among IT professionals?" | "How is perceived code review quality related to developer job satisfaction and team trust?" |
| Comparative Survey | Compare two or more groups on one or more variables | "Do male and female entrepreneurs differ in their perception of institutional barriers to business growth?" | "Do developers working in agile teams report different levels of burnout compared to those in waterfall teams?" |
| Predictive Survey | Identify which variables predict an outcome, and with what relative strength | "To what extent do trust, perceived value, and website quality predict online purchase intention among older consumers?" | "What factors (training, documentation quality, community support) predict continued use of open-source tools among enterprise developers?" |
4. Statistical Tools for Survey Data Analysis
4.1 The Survey Analysis Workflow
Import raw data → Screen for errors (impossible values, straightlining, speeders) → Handle missing data → Reverse-code negatively worded items → Compute scale totals → Check reliability (Cronbach's α ≥ .70 per scale) → Create dummy variables for categorical predictors
Compute Ms, SDs, frequencies, percentages → Assess distributions (skewness, kurtosis, normality) → Generate correlation matrix → Create demographic profile of sample → Compare achieved sample to target population
Check assumptions for each planned test: normality (Shapiro-Wilk, skewness/kurtosis), linearity (scatterplots), homoscedasticity (Levene's, Breusch-Pagan), multicollinearity (VIF, tolerance), independence (Durbin-Watson). Document all checks and any violations.
Run the analyses specified in your methodology: t-tests, ANOVA, correlation, regression, factor analysis, etc. Report exact statistics with effect sizes and confidence intervals. Report ALL planned analyses — not just significant ones.
Format results in APA style. Create properly formatted tables and figures. Write the results narrative organised by RQ. Connect findings back to RQs and hypotheses. Save interpretation for the discussion chapter.
4.2 Selecting the Right Statistical Test
| Your RQ | Independent Variable(s) | Dependent Variable | Test | Non-Parametric Alternative |
|---|---|---|---|---|
| Compare means between TWO groups | 1 categorical (2 groups) | 1 continuous | Independent t-test | Mann-Whitney U |
| Compare means before/after (same people) | 1 categorical (2 time points, within-subjects) | 1 continuous | Paired t-test | Wilcoxon Signed-Rank |
| Compare means across 3+ groups | 1 categorical (3+ groups) | 1 continuous | One-way ANOVA | Kruskal-Wallis H |
| Compare means across 2+ IVs | 2+ categorical IVs | 1 continuous | Factorial ANOVA | Aligned Rank Transform ANOVA |
| Relationship between 2 continuous variables | 1 continuous | 1 continuous | Pearson correlation (r) | Spearman's ρ (rho) |
| Predict DV from multiple IVs | 2+ continuous and/or categorical | 1 continuous | Multiple Linear Regression | Robust regression (bootstrapped SEs) |
| Predict binary outcome | 2+ continuous and/or categorical | 1 binary (0/1) | Binary Logistic Regression | N/A — logistic regression is robust |
| Identify underlying factor structure | Multiple continuous items | Latent factors | Exploratory Factor Analysis (EFA) | Principal Component Analysis (PCA) |
| Test mediation (X → M → Y) | 1 IV, 1 mediator, 1 DV | 1 continuous | Mediation Analysis (PROCESS / lavaan) | Bootstrapped indirect effects |
| Test moderation (interaction) | 2 IVs + interaction term | 1 continuous | Moderated Regression | Simple slopes analysis |
4.3 Software Options for Survey Analysis
| Software | Type | Learning Curve | Best For | Reproducibility | Cost |
|---|---|---|---|---|---|
| SPSS | Menu-driven + syntax | Low | Standard analyses (t-tests, ANOVA, regression); widely taught; APA output options | Moderate (syntax mode); poor (GUI mode) | Institutional licence |
| R (tidyverse, psych, lavaan) | Programming language | Moderate-High | All analyses; publication-quality graphics (ggplot2); reproducible (RMarkdown); cutting-edge methods | Excellent — scripts are inherently reproducible | Free |
| Python (pandas, scipy, statsmodels) | Programming language | Moderate | ML integration; data manipulation at scale; Jupyter notebooks for literate programming | Excellent — scripts + notebooks | Free |
| JASP / jamovi | Menu-driven (GUI) | Very Low | Standard analyses; APA-ready output; Bayesian alternatives; free and open-source | Moderate | Free |
5. How to Choose a Survey Research Topic
Topic selection is the most consequential decision in survey research. A well-chosen topic is researchable through survey methods, feasible within capstone constraints, significant enough to justify the research, and personally engaging enough to sustain motivation across months of work. This section provides a structured process for topic selection.
5.1 The Topic Selection Framework — Five Filters
Does your question ask about attitudes, beliefs, perceptions, self-reported behaviours, or demographic/contextual characteristics that respondents can report? If your question requires observation, experimentation, or deep interpretive exploration, a survey is the wrong method. Test: "Could I write 25–35 survey items that would answer this question?" If not, reconsider the method.
Who is your target population? Can you reach them? How? Do you have access to a sampling frame? What is a realistic sample size given your timeline, resources, and network? A study requiring 200 senior executives will be infeasible unless you have exceptional access. Test: List 5 specific channels through which you will recruit respondents. If you cannot list 5, your access is uncertain. If 3 of 5 channels are "social media," your access is optimistic.
Does your question address a genuine gap in the literature? Does it have practical implications — for organisations, policymakers, practitioners, or technology designers? Will anyone care about the answer? Test: Complete the sentence: "This study matters because if we don't answer this question, [consequence]." If you cannot complete this sentence convincingly, the topic needs refinement.
Is there existing theory that can frame your study and generate testable hypotheses? Are there established constructs with validated measurement instruments? Test: Name the theory. Name the key constructs. Identify at least one validated scale for each construct. If you cannot, your study may lack theoretical grounding — a common weakness in student survey research.
You will spend 6–12 months on this topic — reading literature, designing instruments, collecting data, analysing results, and writing. Does the topic genuinely interest you? Is it connected to your career aspirations? Test: Can you explain your topic to a non-specialist in 60 seconds with genuine enthusiasm? If you sound bored explaining it, you will be bored researching it.
5.2 The Topic Refinement Process
- Start broad, then narrow. "Digital transformation" → "Technology adoption in Indian SMEs" → "Factors influencing cloud computing adoption among manufacturing SMEs in Pune." Each refinement makes the topic more researchable.
- Identify your constructs. What are the key variables you will measure? For a study of cloud adoption: adoption intention (DV), perceived usefulness, perceived ease of use, organisational readiness, competitive pressure (IVs). Each construct needs a validated scale.
- Check for existing literature. Search your constructs in Google Scholar, Scopus, or your institutional database. Are there 20+ relevant papers? If fewer than 10, the topic may be too niche. Are there 500+? The topic may be saturated — find a specific angle or context that differentiates your study.
- Assess respondent access. Who is your population? How will you reach them? What is a realistic sample size? If you cannot access at least 100 respondents, survey research may not be viable.
- Draft your RQs and hypotheses. Write 1–2 primary RQs and 3–5 hypotheses grounded in your theoretical framework. If you cannot articulate clear, testable hypotheses, your topic needs further development.
6. Sample Survey Research Topics — 10 Business + 10 Computer Science
Each topic below is presented as a complete research idea: the research questions, the theoretical framework, the target population, the key constructs to measure, and the sampling strategy. These are starting points — adapt, refine, and make them your own.
6.1 Ten Business / Social Science Survey Research Topics
RQs: (1) To what extent do seller verification badges and cash-on-delivery availability predict consumer trust in Indian e-commerce platforms? (2) Does the relative importance of these trust signals differ between experienced and novice online shoppers? Theory: Source Credibility Theory (Hovland); Trust Transfer Theory. Population: Online shoppers aged 18–55 in Indian Tier-1 and Tier-2 cities who have made at least 3 online purchases in the last 12 months. Constructs: Trust in platform, trust in seller, perceived risk, purchase intention, seller verification perception, COD preference. Sampling: Stratified quota sampling — 150 respondents, stratified by shopping experience (novice: ≤1 year; experienced: ≥3 years). Online survey distributed through consumer panels and social media. Analysis: Hierarchical multiple regression; moderation analysis (experience as moderator of trust → purchase relationship).
RQs: (1) What is the relationship between remote work intensity (hours/week), manager communication quality, and employee engagement among IT professionals? (2) Does manager communication quality mediate the relationship between remote work intensity and engagement? Theory: Social Exchange Theory; Leader-Member Exchange (LMX). Population: Full-time IT professionals in Indian companies who work remotely at least 2 days per week. Constructs: Remote work intensity, manager communication quality, employee engagement (Utrecht Work Engagement Scale — UWES), perceived organisational support, turnover intention. Sampling: Purposive sampling through LinkedIn and professional networks. Target n = 250. Analysis: Mediation analysis (PROCESS Model 4); hierarchical regression.
RQs: (1) What factors predict fintech adoption (digital payment acceptance, digital ledger use) among small retailers in Tier-2 Indian cities? (2) Does the UTAUT2 model adequately explain adoption in this context, or do context-specific factors (digital literacy, trust in technology providers) add explanatory power? Theory: UTAUT2 (Venkatesh et al., 2012) extended with context-specific constructs. Population: Owners/managers of small retail businesses (kirana stores, mobile shops, pharmacies) in Tier-2 cities in one Indian state. Constructs: Performance expectancy, effort expectancy, social influence, facilitating conditions, hedonic motivation, price value, habit, digital literacy, trust in technology providers, adoption intention, actual use. Sampling: Cluster sampling — randomly select 5 Tier-2 cities, then purposively sample 40 retailers per city. Total target n = 200. Face-to-face survey administration recommended for this population. Analysis: Hierarchical regression comparing UTAUT2-only model with extended model (ΔR² test).
RQs: (1) To what extent do environmental values, environmental knowledge, and perceived product effectiveness predict green purchase intention among Indian millennials? (2) Does perceived greenwashing moderate the relationship between environmental values and purchase intention? Theory: Value-Belief-Norm Theory (Stern); Theory of Planned Behaviour (TPB). Population: Indian consumers aged 25–40 (millennials) who have purchased at least one product marketed as "sustainable" or "eco-friendly" in the last 12 months. Constructs: Environmental values (NEP scale), environmental knowledge, perceived consumer effectiveness, perceived greenwashing, green purchase intention, willingness to pay premium. Sampling: Online survey via social media (Instagram, LinkedIn, WhatsApp). Target n = 300. Quota sampling for gender balance. Analysis: Moderated regression; structural equation modelling (if n allows).
RQs: (1) Do consumers perceive micro-influencers (10K–100K followers) as more credible, authentic, and persuasive than macro-influencers (1M+ followers)? (2) Which source characteristics (expertise, trustworthiness, attractiveness, similarity) most strongly predict purchase intention for each influencer type? Theory: Source Credibility Theory; Parasocial Interaction Theory. Population: Instagram users aged 18–35 in urban India who follow at least one influencer and have made a purchase based on an influencer recommendation. Constructs: Perceived expertise, trustworthiness, attractiveness, similarity (PSI), authenticity, purchase intention, brand attitude. Sampling: Purposive sampling through Instagram ads and influencer partner posts. Target n = 250. Analysis: Independent t-test (micro vs. macro on credibility); multi-group regression (predictors of purchase intention by influencer type).
RQs: (1) What is the relationship between organisational culture types (Competing Values Framework) and perceived organisational innovativeness? (2) Does psychological safety mediate the relationship between culture type and innovativeness? Theory: Competing Values Framework (Cameron & Quinn); Psychological Safety (Edmondson). Population: Middle managers and team leaders in Indian companies with 100+ employees across IT, manufacturing, and financial services. Constructs: Organisational culture (OCAI — clan, adhocracy, market, hierarchy), psychological safety (Edmondson's scale), organisational innovativeness, employee creativity. Sampling: Purposive + snowball sampling through professional networks and alumni databases. Target n = 200. Analysis: Mediation analysis (PROCESS); ANOVA comparing culture types on innovativeness.
RQs: (1) How do dimensions of omnichannel integration (service consistency, channel transparency, information synchronisation) predict customer satisfaction and loyalty? (2) Is the effect of integration quality on loyalty mediated by customer satisfaction? Theory: SERVQUAL adapted for omnichannel; Expectation-Confirmation Theory. Population: Consumers who have used both online and offline channels of the same retailer in the last 6 months. Constructs: Omnichannel integration quality, perceived seamlessness, customer satisfaction, customer loyalty (repurchase intention, word-of-mouth), brand trust. Sampling: Online survey distributed through retailer customer databases or consumer panels. Target n = 300. Analysis: Structural equation modelling (SEM) or path analysis; mediation analysis.
RQs: (1) What is the relative importance of perceived autonomy, algorithmic management intensity, and social support in predicting gig worker job satisfaction and well-being? (2) Does social support moderate the negative effect of algorithmic management on well-being? Theory: Self-Determination Theory (SDT); Job Demands-Resources (JD-R) Model. Population: Platform-based gig workers in Indian metros — food delivery, ride-hailing, and task-based platforms. Constructs: Perceived autonomy, algorithmic management perception, social support (peer and family), job satisfaction, psychological well-being (WHO-5), continuance intention. Sampling: Snowball sampling starting from gig worker forums, WhatsApp groups, and platform worker associations. Target n = 150–200. Analysis: Hierarchical regression; moderation analysis.
RQs: (1) How do consumers respond to brand activism on social issues — does perceived sincerity of the brand's stance predict brand attitude and purchase intention? (2) Does consumer-brand value congruence moderate this relationship? Theory: Attribution Theory; Social Identity Theory. Population: Socially conscious consumers aged 18–40 who follow brands on social media and are aware of brand activism campaigns. Constructs: Perceived brand activism sincerity, consumer-brand value congruence, brand attitude, purchase intention, brand trust, political ideology (control). Sampling: Online survey through social media and consumer panels. Target n = 300. Quota for gender and age. Analysis: Moderated regression; multi-group analysis by value congruence level.
RQs: (1) To what extent do financial literacy, risk tolerance, and social influence predict investment participation (mutual funds, stocks, crypto) among young Indian professionals? (2) Does financial literacy moderate the relationship between risk tolerance and investment diversification? Theory: Theory of Planned Behaviour (TPB); Behavioural Finance. Population: Employed professionals aged 22–35 in urban India with monthly income ≥ ₹30,000. Constructs: Financial literacy (objective — quiz-based), subjective financial knowledge, risk tolerance, social influence, investment self-efficacy, investment participation, investment diversification. Sampling: Online survey distributed through professional networks, LinkedIn, and finance-focused communities. Target n = 350. Analysis: Logistic regression (investment participation — binary); moderated regression (diversification).
6.2 Ten Computer Science / IT Survey Research Topics
RQs: (1) What factors predict software developers' adoption of AI-assisted coding tools (GitHub Copilot, ChatGPT, CodeWhisperer)? (2) How do perceived productivity gains, trust in AI-generated code, and concerns about code quality predict continued use intention? Theory: UTAUT2 extended with AI-specific constructs (trust in automation, algorithmic transparency perception). Population: Professional software developers in India who have used at least one AI coding assistant. Constructs: Performance expectancy, effort expectancy, social influence, facilitating conditions, trust in AI-generated code, code quality concern, AI self-efficacy, continued use intention. Sampling: Online survey distributed through Stack Overflow, GitHub, LinkedIn developer communities, and tech company newsletters. Target n = 250. Analysis: Hierarchical regression; SEM (if n allows).
RQs: (1) What is the level of cybersecurity awareness and practice adoption among Indian SMEs? (2) What factors — perceived threat severity, cost, technical expertise, regulatory pressure — predict cybersecurity investment and practice implementation? Theory: Protection Motivation Theory (PMT); Technology-Organisation-Environment (TOE) Framework. Population: IT managers, CTOs, or business owners in Indian SMEs (10–250 employees) across sectors. Constructs: Perceived threat severity, perceived vulnerability, response efficacy, self-efficacy, cost barrier, technical expertise, regulatory pressure, cybersecurity practice adoption. Sampling: Purposive sampling through industry associations (NASSCOM, CII), LinkedIn, and business networks. Target n = 200. Analysis: Multiple regression; cluster analysis to identify SME segments by cybersecurity maturity.
RQs: (1) How do professional developers and citizen developers differ in their perceptions of low-code/no-code (LCNC) platform usefulness, ease of use, and limitations? (2) What factors predict satisfaction with LCNC platforms and intention to continue using them? Theory: TAM; Task-Technology Fit. Population: Users of LCNC platforms (professional developers and business users/"citizen developers") in Indian organisations. Constructs: Perceived usefulness, perceived ease of use, perceived limitations (customisation, scalability, integration), task-technology fit, developer satisfaction, continued use intention. Sampling: Online survey distributed through LCNC platform user communities (Microsoft Power Platform, OutSystems, Mendix forums), LinkedIn, and tech communities. Target n = 200. Analysis: Independent t-test (professional vs. citizen developers); multi-group regression.
RQs: (1) What intrinsic and extrinsic motivations drive Indian developers to contribute to open-source software (OSS) projects? (2) What barriers — time constraints, lack of mentorship, imposter syndrome, unclear contribution guidelines — most strongly inhibit sustained OSS contribution? Theory: Self-Determination Theory (SDT); Social Cognitive Career Theory. Population: Indian software developers who have made at least one contribution (code, documentation, issue reporting) to an OSS project in the last 2 years. Constructs: Intrinsic motivation (enjoyment, learning), extrinsic motivation (reputation, career benefit), internalised motivation (community identification), contribution barriers, OSS self-efficacy, contribution frequency, sustained contribution intention. Sampling: Online survey distributed through GitHub, GitLab, OSS community forums, and Indian developer communities. Target n = 300. Analysis: Multiple regression; structural equation modelling.
RQs: (1) To what extent do privacy concern, perceived app utility, and trust in app developer predict users' willingness to share personal data with mobile applications? (2) Does the type of data requested (location, contacts, health, browsing history) moderate this relationship? Theory: Privacy Calculus Theory; Communication Privacy Management Theory. Population: Smartphone users aged 18–55 in India who use 5+ mobile apps regularly. Constructs: Privacy concern (IUIPC), perceived app utility, trust in developer, data type sensitivity, willingness to share data, privacy-protective behaviour. Sampling: Online survey through social media, university networks, and consumer panels. Target n = 350. Quota for age groups. Analysis: Repeated-measures ANOVA (data type as within-subjects factor); moderated regression.
RQs: (1) How do specific agile practices (daily stand-ups, retrospectives, pair programming, sprint planning) relate to team satisfaction and perceived team effectiveness? (2) Does agile maturity (years of agile experience) moderate the relationship between practice adherence and satisfaction? Theory: Task-Technology Fit; Job Characteristics Model. Population: Software developers, Scrum Masters, and product owners working in agile teams in Indian IT companies. Constructs: Agile practice adherence (per practice), team satisfaction, perceived team effectiveness, psychological safety, agile maturity, organisational support for agile. Sampling: Purposive + snowball sampling through LinkedIn, agile community meetups, and tech company contacts. Target n = 250. Analysis: Multiple regression; moderation analysis; relative weight analysis to identify most impactful practices.
RQs: (1) What technological, organisational, and environmental factors predict cloud migration intention among Indian enterprises? (2) How do perceived benefits (scalability, cost, agility) compare with perceived barriers (security, vendor lock-in, compliance) in predicting migration decisions? Theory: TOE Framework; Diffusion of Innovation (DOI). Population: IT decision-makers (CTOs, IT managers, Heads of Engineering) in Indian companies with 50+ employees that have not fully migrated to the cloud. Constructs: Relative advantage, compatibility, complexity, security concern, vendor lock-in concern, regulatory compliance pressure, competitive pressure, top management support, cloud migration intention. Sampling: Purposive sampling through NASSCOM, CIO associations, LinkedIn, and professional networks. Target n = 180. Analysis: Logistic regression (migration intention — binary); multiple regression.
RQs: (1) What is the relationship between remote work technology quality (video conferencing, collaboration tools, VPN/remote access), technology overload, and digital burnout among IT professionals? (2) Does technology self-efficacy moderate the technology overload → burnout relationship? Theory: Technostress Framework (Tarafdar et al.); Job Demands-Resources Model. Population: IT professionals in India who work remotely or in hybrid arrangements at least 3 days per week. Constructs: Technology quality (reliability, usability, feature sufficiency), technology overload (information, communication, system feature overload), technology self-efficacy, digital burnout (exhaustion, cynicism, reduced efficacy), job satisfaction. Sampling: Online survey through LinkedIn, Slack/Discord developer communities, and HR contacts in IT companies. Target n = 300. Analysis: Structural equation modelling (SEM) or path analysis; moderation analysis.
RQs: (1) What factors predict user acceptance of AI-powered healthcare applications (symptom checkers, telemedicine triage, health monitoring) among Indian smartphone users? (2) Does trust in AI mediate the relationship between perceived AI competence and behavioural intention? Theory: UTAUT2; Trust in Automation. Population: Smartphone users aged 25–60 in urban India who have used or considered using a health-related mobile application. Constructs: Performance expectancy, effort expectancy, social influence, perceived AI competence, trust in AI, data privacy concern, health literacy, behavioural intention. Sampling: Online survey through health and wellness communities, social media, and consumer panels. Target n = 300. Quota for age and prior AI health app experience. Analysis: Mediation analysis (PROCESS); multi-group SEM (users vs. non-users of AI health apps).
RQs: (1) What factors predict the intention to adopt blockchain technology for supply chain management among Indian logistics and manufacturing firms? (2) How do perceived benefits (traceability, transparency, efficiency) compare with perceived barriers (complexity, cost, interoperability) in predicting adoption intention? Theory: TOE Framework; TAM extended for blockchain. Population: Supply chain managers, logistics heads, and IT decision-makers in Indian manufacturing and logistics companies. Constructs: Perceived traceability benefit, perceived transparency benefit, perceived complexity, perceived cost, interoperability concern, top management support, competitive pressure, regulatory uncertainty, blockchain adoption intention. Sampling: Purposive sampling through industry associations (CII, FICCI), logistics forums, LinkedIn, and professional networks. Target n = 150. Analysis: Multiple regression; importance-performance map analysis (IPMA).
7. Questionnaire Design — From Constructs to Items
7.1 The Construct-to-Item Mapping Process
Every item in your questionnaire should be traceable to a specific construct and a specific dimension of that construct. The construct-to-item map is your design document — it ensures that your questionnaire measures what you intend to measure, completely and without redundancy.
| Construct | Dimension | # Items | Sample Item | Source |
|---|---|---|---|---|
| Perceived Usefulness (PU) | Performance enhancement | 2 | "Using [system] improves my performance in [task]." | Davis (1989), PU2 |
| Productivity | 2 | "Using [system] increases my productivity." | Davis (1989), PU3 | |
| Effectiveness | 2 | "Using [system] enhances my effectiveness in [task]." | Davis (1989), PU5 |
7.2 Questionnaire Item Writing — 10 Rules
- One idea per item. Never combine two questions into one item. Split double-barrelled items.
- Use simple, concrete language. Write for the reading level of your least educated respondent. Avoid jargon unless your population uses it.
- Avoid leading questions. Don't signal the desired answer. "Most experts agree..." biases responses.
- Avoid loaded terms. Emotionally charged words ("waste," "suffering," "unfair") trigger emotional rather than considered responses.
- Provide balanced response scales. For agreement: Strongly Disagree to Strongly Agree (not Disagree to Strongly Agree — missing the negative anchor).
- Include "Not Applicable" where needed. If a question may not apply to all respondents, provide an N/A option rather than forcing a response.
- Ensure mutually exclusive categories. Age: 18–25, 26–35 (not 18–25, 25–35 — 25 appears in both).
- Keep items short. Aim for ≤ 20 words. Long items confuse respondents and reduce data quality.
- Use established scales where possible. Adopting validated scales is methodologically superior to creating new items. Check the original reliability (Cronbach's α) and cite the source.
- Pilot test before deployment. Administer to 5–10 people similar to your target population. Ask them what they thought each item meant. Revise items where interpretation differs from intention.
8. Common Pitfalls in Survey Research — and How to Avoid Them
| Pitfall | What It Looks Like | Consequence | Prevention |
|---|---|---|---|
| Common Method Bias (CMB) | All data collected from the same source at the same time using the same method — especially when both IV and DV data come from the same respondents in the same survey | Inflated correlations between constructs; relationships appear stronger than they actually are because they share method variance, not just substantive variance | Procedural: separate IV and DV measurement (different sections, different response formats); ensure anonymity to reduce social desirability; use validated scales. Statistical: Harman's single-factor test; marker variable technique. Acknowledge CMB as a limitation. |
| Social Desirability Bias | Respondents answer in ways that make them look good rather than truthfully — over-reporting desirable behaviours, under-reporting undesirable ones | Systematic overestimation of positive behaviours and underestimation of negative ones; biased relationships between constructs | Ensure anonymity; use indirect questioning ("How common is it for your colleagues to..."); include social desirability scale (Marlowe-Crowne) as control; acknowledge in limitations |
| Non-Response Bias | Respondents differ systematically from non-respondents on variables relevant to your RQs — your sample is biased toward those who chose to respond | Findings may not represent the population; the achieved sample may be systematically different from the target population in unmeasured ways | Compare early vs. late respondents on key variables; compare achieved sample demographics with known population parameters (census data); report response rate; acknowledge as limitation |
| Straightlining / Satisficing | Respondents select the same response option repeatedly (e.g., all "3 — Neutral") without reading items — providing no actual information | Unreliable data; inflated or deflated scale reliabilities; meaningless patterns | Include reverse-coded items; include attention check items ("Please select 'Strongly Agree' for this item"); screen for straightlining during data cleaning (SD = 0 across scale items); exclude identified straightliners |
| Low Reliability | Cronbach's α for a scale is below .70 — the items are not consistently measuring the same construct | Unreliable measurement attenuates relationships; you may miss real effects because your instrument is too noisy | Use validated scales with established reliability; pilot test and check α before full deployment; if α < .70 in your data, check inter-item correlations, consider removing problematic items, and report both original and corrected α |
Think Deeper — Cross Questions
Discuss in pairs before sharing with the class.
You plan to survey 200 "Indian consumers" about their online shopping behaviour. A peer asks: "Which consumers? Where? Selected how?" You realise your population definition is too vague. Rewrite your population definition with sufficient precision that another researcher could replicate your sampling procedure. What specific criteria did you add? Why does precision at the population definition stage matter for everything that follows?
Your ideal sample size (from power analysis) is 250. Your realistic achievable sample (given your network, timeline, and resources) is approximately 100. What do you do? Consider: (a) adjusting your analytical approach (simpler tests require smaller samples), (b) reframing your study as exploratory, (c) supplementing with a secondary data source, (d) extending the data collection timeline. Which option is best for your specific situation — and why?
Review the 20 sample topics (Section 6). Select ONE topic from the business list and ONE from the CS list. For each: identify the strongest aspect of the proposed study design and one weakness or risk. How would you strengthen the design to address that weakness?
Reflect on your own survey research design (or plans). Apply the five filters (Section 5.1). Does your topic pass all five? Which filter is the weakest for your topic — and what specific action can you take to strengthen it before data collection begins?
Key Takeaways — Survey Research
A vaguely defined population produces vague conclusions. Define your population precisely — who, where, when — before designing your instrument or sampling strategy. Every design decision flows from the population definition. Imprecise populations produce uninterpretable findings.
Probability sampling supports statistical generalisation. Non-probability sampling does not — and presenting it as if it did is the most common error in capstone research. Be honest about your sample's limitations. The evaluation committee accepts acknowledged constraints; they reject misrepresentation.
A survey is not a collection of interesting questions. It is a measurement instrument. Every item must be justified, linked to a construct, and tested before deployment. Adopt validated scales. Pilot test. Check reliability before running substantive analyses.
The same methodological principles — precise population definition, rigorous sampling, validated measurement, systematic analysis — apply whether you study consumer behaviour or developer tool adoption. The contexts differ; the standards are the same. A well-designed survey in CS is methodologically equivalent to a well-designed survey in business.