Sampling Design, Steps in Sampling Design, Criteria of Selecting a Sampling Procedure, Characteristics of Good Sample Design, Types of Sample Design

Sampling design is a blueprint or plan that defines how a researcher selects a subset (sample) from a larger population for the purpose of making generalizations. Since studying an entire population is often impossible due to time, cost, and accessibility constraints, sampling provides a practical alternative. A good sampling design ensures that the selected sample accurately represents the population, minimizing bias and sampling error. It specifies the target population, sampling frame, sampling technique (probability or non probability), and sample size. In Indian business research, sampling design is critical when studying large populations such as ecommerce users, bank customers, or factory workers. Poor sampling design leads to invalid conclusions regardless of how well other steps are executed. Thus, sampling design is the foundation of reliable business research.

Steps in Sampling Design:

1. Define the Target Population

The first step is to clearly define the target population, meaning the entire group of individuals, objects, or events that possess the characteristics relevant to the research problem. The definition must specify inclusion criteria (who belongs) and exclusion criteria (who does not). For example, a study on ebanking adoption might define target population as “Indian adults aged 18 to 60 who have at least one bank account and internet access.” In Indian business research, vague population definitions lead to sampling errors. If you cannot specify exactly who your population is, you cannot draw a proper sample. The target population determines everything that follows: sampling frame, technique, and sample size. Spend adequate time on this step. A well defined population is the foundation of valid generalization.

2. Identify the Sampling Frame

The sampling frame is the actual list or source from which the sample will be drawn. Ideally, the sampling frame should match the target population perfectly. In practice, mismatches occur. For example, if your target population is “all ecommerce users in India,” the sampling frame might be “email list of registered users of one ecommerce platform.” This frame excludes users of other platforms. In Indian business research, common sampling frames include telephone directories, voter lists, customer databases, employee rosters, and school registers. Frame errors occur when the frame underrepresents (missing some population members), overrepresents (includes ineligible members), or is outdated. Always assess frame quality before proceeding. A poor frame cannot produce a good sample, regardless of sampling technique.

3. Select the Sampling Technique

The sampling technique is the method used to select elements from the sampling frame. Choose between probability sampling (every element has known, nonzero chance of selection) and non probability sampling (chance is unknown or zero). Probability techniques include simple random, stratified, cluster, and systematic sampling. Non probability techniques include convenience, purposive, quota, and snowball sampling. In Indian business research, probability sampling is preferred for generalizable conclusions, but non probability sampling is common due to time and budget constraints. The choice depends on research objectives, available resources, and access to the sampling frame. For example, a PhD thesis requiring statistical inference needs probability sampling. An exploratory study with a hard to reach population (e.g., informal sector workers) may use snowball sampling. Document your justification clearly.

4. Determine the Sample Size

Sample size refers to the number of elements selected from the sampling frame. Too small a sample lacks precision and statistical power. Too large a sample wastes resources and may detect trivial differences as statistically significant. In Indian business research, sample size can be determined using formulas based on population size, desired confidence level (typically 95 percent), margin of error (typically 3 to 5 percent), and expected effect size. For unknown populations, rules of thumb exist: minimum 30 for central limit theorem, 100 for correlational studies, 200 for structural equation modeling. Online calculators and software (G*Power, SPSS SamplePower) simplify the task. Never arbitrarily pick a sample size (e.g., “I will take 50 respondents”). Justify your sample size statistically or with published precedents from similar Indian studies.

5. Specify Sampling Unit

The sampling unit is the individual element or group of elements selected at each stage of sampling. It may be the same as the target population element or different. For example, in a study of ecommerce customers, the sampling unit could be an individual customer. In cluster sampling of Indian villages, the primary sampling unit might be a village, and the secondary sampling unit (within selected villages) might be households. In Indian business research, clearly specify the sampling unit for each stage. Ambiguity leads to unit of analysis errors, where you draw conclusions about individuals from data collected at group level (ecological fallacy) or vice versa. The sampling unit must match the unit of analysis stated in your research objectives. Consistency across these units ensures logical validity.

6. Evaluate Sampling Frame Adequacy

Before drawing the sample, assess whether the sampling frame is adequate in terms of coverage, currency, and accuracy. Coverage means the frame includes all population elements. Currency means the frame is up to date. Accuracy means the frame contains correct information. For example, a customer database from 2019 is inadequate for a 2024 study because many customers may have changed addresses or stopped using the service. In Indian business research, common frame problems include outdated telephone directories, incomplete employee rosters (missing contract workers), and voter lists that exclude migrants. If the frame is inadequate, consider alternative frames or acknowledge the limitation. If no adequate frame exists, you may need to use non probability sampling. Document all frame limitations in your research report for transparency.

7. Draw the Sample

This step involves physically selecting elements from the sampling frame according to the chosen sampling technique. For probability sampling, use random number generators, lottery method, or systematic selection with a random start. For non probability sampling, follow the rules of the chosen method (e.g., select first 100 available participants for convenience sampling). In Indian business research, maintain a sampling log documenting every selection decision. This log allows others to replicate your procedure. Avoid substituting selected elements who are unavailable with convenient replacements, as this introduces bias. If a selected element refuses or cannot be contacted, follow a predetermined protocol (e.g., make three attempts at different times before replacing). Never arbitrarily choose replacements. The integrity of your sampling design is judged by how faithfully you execute this step.

8. Assess Non Response and Bias

After data collection, assess the extent of non response (selected elements who did not participate). High non response rates (above 20 to 30 percent) threaten generalizability because non respondents may differ systematically from respondents. For example, in a survey of Indian factory workers, non respondents might be those with low job satisfaction who fear speaking out. Compare respondents and non respondents on available characteristics (age, gender, location) to detect bias. In Indian business research, common non response reasons include lack of time, privacy concerns, and survey fatigue. Attempt to contact a subsample of non respondents with a shorter questionnaire to estimate differences. If non response bias is significant, apply statistical weighting adjustments or acknowledge the limitation in your report. Sampling does not end with selection; it includes handling missing data.

9. Pilot Test the Sampling Procedure

Before full scale data collection, conduct a pilot test of your sampling procedure on a small scale (typically 5 to 10 percent of intended sample size). The pilot reveals practical problems: inaccessible sampling frames, incorrect contact information, high refusal rates, or ambiguous inclusion criteria. For example, a pilot might show that your sampling frame of ecommerce users contains many inactive accounts. In Indian business research, pilot testing saves enormous time and money. It allows you to revise the sampling design before committing to full data collection. During the pilot, document every obstacle encountered and how you resolved it. After the pilot, evaluate whether the sample obtained matches the target population on key demographics. If not, adjust your sampling frame or technique. A pilot tested design is a reliable design.

10. Document Sampling Design Decisions

The final step is to comprehensively document every sampling decision made throughout the process. This includes: target population definition, sampling frame description and limitations, sampling technique justification, sample size calculation, sampling unit specification, non response handling protocol, and any deviations from the original plan. In Indian business research, this documentation belongs in the methodology chapter of your thesis or report. Clear documentation allows readers to assess the generalizability of your findings and enables replication by other researchers. It also demonstrates your competence as a researcher. Examiners and journal reviewers scrutinize sampling documentation carefully. Vague statements like “a random sample was taken” are insufficient. Provide exact details: which random method, which software, what seed value, what dates, what response rate. Transparency is the hallmark of quality sampling.

Criteria of Selecting a Sampling Procedure:

1. Research Objectives

The primary criterion is the research objective. If the goal is to generalize findings to a larger population, probability sampling is necessary. If the goal is exploratory or illustrative, non probability sampling may suffice. For example, a study measuring the percentage of Indian ecommerce users satisfied with delivery requires probability sampling for accurate estimation. A study exploring reasons for cart abandonment among 20 heavy users can use purposive sampling. In Indian business research, clearly state whether you need statistical inference or just directional insights. Misalignment between objective and sampling procedure invalidates conclusions. Always ask: “Do I need to claim that my findings represent the entire population?” If yes, choose probability sampling. If no, non probability methods are acceptable.

2. Population Characteristics

The nature of the target population influences sampling procedure selection. Consider population homogeneity (how similar members are), geographic dispersion, and accessibility. A homogeneous population (e.g., employees in a single call center) can be studied with smaller samples and simpler methods. A heterogeneous population (e.g., all Indian ebanking users across 28 states) requires stratified sampling to ensure representation of subgroups. In Indian business research, populations with hard to reach members (e.g., informal workers, tribal communities, homeless individuals) may force use of snowball or purposive sampling. Also consider population size. Very small populations (under 200) may be studied entirely (census) rather than sampled. Match your sampling procedure to the practical realities of your population.

3. Availability of Sampling Frame

A sampling frame is the list from which the sample is drawn. Probability sampling requires a complete, accurate, and accessible sampling frame. If such a frame exists (e.g., employee roster, customer database, voter list), probability methods are feasible. If no frame exists or the frame is severely flawed (outdated, incomplete, inaccessible), probability sampling becomes impossible. In Indian business research, many target populations lack proper frames. For example, there is no complete list of all street vendors in Mumbai or all users of local kirana stores. In such cases, non probability methods like convenience or quota sampling are the only practical options. Never pretend to use probability sampling without a genuine sampling frame. Transparently acknowledge frame limitations.

4. Available Time and Budget

Sampling procedures vary enormously in time and cost. Simple random sampling requires listing all population elements (costly for large populations). Stratified sampling requires detailed population data before sampling. Cluster sampling reduces travel costs but increases statistical complexity. Non probability methods are generally faster and cheaper. In Indian business research, students and small organizations often have severe constraints. A PhD scholar with limited funding may use convenience sampling from easily accessible participants. A multinational corporation can afford complex probability designs. Be realistic. Do not propose a sampling procedure you cannot execute given your resources. However, also recognize the trade off: cheaper methods produce weaker evidence. Document your constraints and their implications for generalizability.

5. Required Precision and Confidence

Precision refers to how close sample estimates are to true population values. Confidence refers to the probability that the sample estimate falls within a specified margin of error. Higher precision and confidence require probability sampling with larger sample sizes. If your research requires precise estimates (e.g., “55 percent of Indian urban consumers use ecommerce, plus or minus 2 percent”), you must use probability sampling with rigorous sample size calculation. If rough estimates are acceptable (e.g., “approximately half of consumers use ecommerce”), non probability methods may suffice. In Indian business research, regulatory or policy studies demand high precision. Academic theses vary: PhD requires higher precision than undergraduate projects. Specify your required confidence level (typically 95 percent) and margin of error before selecting sampling procedure.

6. Nature of the Study (Exploratory vs Descriptive vs Causal)

Exploratory studies (generating hypotheses, understanding new phenomena) do not require probability sampling. Convenience or purposive sampling is acceptable. Descriptive studies (estimating population parameters, measuring frequencies) require probability sampling for valid generalization. Causal studies (testing cause effect relationships) may use either, but probability sampling strengthens internal validity. In Indian business research, match the sampling procedure to the study type. A master’s dissertation exploring “reasons for low ebanking adoption in rural Bihar” can use purposive sampling of 30 villagers. A survey estimating “percentage of rural households with ebanking accounts” must use probability sampling. Clearly classify your study type before selecting sampling procedure. Misclassification leads to inappropriate sampling and invalid conclusions.

7. Statistical Analysis Requirements

The statistical tests you plan to use impose requirements on the sampling procedure. Parametric tests (t test, ANOVA, regression) assume random sampling from normally distributed populations. If you plan parametric tests, you need probability sampling. Non parametric tests (chi square, Mann Whitney) are more flexible but less powerful. Additionally, advanced techniques like structural equation modeling require large sample sizes (200 to 500) obtained through probability methods for generalizable results. In Indian business research, many students plan parametric tests but use convenience sampling. This violates test assumptions and produces questionable findings. Before finalizing sampling procedure, list your planned statistical tests and check their assumptions. If assumptions cannot be met, either change your tests or change your sampling procedure. Alignment is essential.

8. Degree of Heterogeneity in Population

Heterogeneous populations (members differ significantly on key variables) require more sophisticated sampling procedures. Simple random sampling may miss rare subgroups entirely. For example, a study of Indian ecommerce users must represent users from different age groups, income levels, states, and languages. Simple random sampling might accidentally exclude elderly users or rural users. Stratified sampling ensures each subgroup is adequately represented. If the population is homogeneous (all members are similar), simple random or even convenience sampling may suffice. In Indian business research, most populations are heterogeneous due to cultural, linguistic, economic, and geographic diversity. Assess your population’s heterogeneity before choosing sampling procedure. Do not use simple random sampling for a diverse population without stratification.

9. Risk of Sampling Bias

Sampling bias occurs when some population members are systematically more likely to be selected than others. Probability sampling minimizes bias by design. Non probability sampling inherently contains bias. The acceptable level of bias depends on the research purpose. For academic publication or policy making, bias must be minimized, favoring probability sampling. For internal business reports or student exercises, some bias may be tolerable. In Indian business research, common bias sources include: using only urban samples to represent national populations, using only online samples excluding offline populations, and using only willing volunteers. Assess your bias risk honestly. If bias would invalidate your conclusions, choose a procedure that reduces it. If bias is unavoidable, disclose it clearly in limitations.

10. Replicability Requirements

If other researchers need to replicate your study, probability sampling is strongly preferred because it provides transparent, reproducible selection procedures. Non probability sampling relies on researcher judgment, making exact replication impossible. In Indian business research, PhD theses and journal articles should prioritize replicability. The sampling procedure must be described in sufficient detail that another researcher could independently draw a sample from the same population using the same method. Simple random sampling (with documented random seed) is highly replicable. Purposive sampling (e.g., “select experts with at least 10 years experience”) is replicable in principle. Convenience sampling (e.g., “survey available students”) is not replicable because availability changes. Consider whether your research will be used as evidence for policy or practice. If yes, replicability is essential.

Characteristics of Good Sample Design:

1. Representativeness

A good sample design must produce a sample that accurately represents the target population on all relevant characteristics. Representativeness means that the distribution of key variables (age, gender, income, location, etc.) in the sample closely matches their distribution in the population. For example, if 40 percent of Indian ecommerce users are women, a representative sample should have approximately 40 percent women. Without representativeness, generalizations from sample to population are invalid. Probability sampling techniques (random, stratified) maximize representativeness. Non probability techniques risk unrepresentative samples. In Indian business research, check representativeness by comparing sample demographics with known population parameters from census or industry reports. If significant mismatches exist, apply weighting adjustments or acknowledge the limitation. Representativeness is the most important characteristic of any sample design.

2. Absence of Sampling Bias

Sampling bias occurs when some population elements have a systematically higher or lower chance of being selected than others. A good sample design eliminates or minimizes all sources of sampling bias. Common biases include selection bias (excluding certain groups), non response bias (selected individuals refuse to participate), and frame bias (sampling frame does not match population). In Indian business research, bias often enters through convenience sampling (e.g., surveying only urban college students to represent all youth). Probability sampling techniques are designed to eliminate selection bias. However, bias can still occur during execution (e.g., substituting unavailable respondents with convenient alternatives). A good design anticipates bias sources and includes protocols to prevent them. Document and report any remaining bias transparently.

3. Precision and Accuracy

Precision refers to the degree of closeness between repeated sample estimates from the same population. Accuracy (unbiasedness) refers to how close the sample estimate is to the true population value. A good sample design achieves both. Precision is measured by standard error or margin of error; smaller values indicate higher precision. Precision improves with larger sample sizes and more efficient sampling techniques (stratified sampling often gives higher precision than simple random sampling for the same sample size). In Indian business research, specify desired precision before collecting data. For example, “I want to estimate average monthly ecommerce spending within plus or minus ₹200.” The sample design must achieve this precision. Accuracy requires unbiased selection procedures. A precise but biased sample (e.g., consistently overestimating spending) is worthless.

4. Adequate Sample Size

A good sample design includes a sample size that is neither too small nor unnecessarily large. Too small a sample lacks statistical power, produces wide confidence intervals, and cannot detect meaningful differences. Too large a sample wastes resources and may detect trivial differences as statistically significant. Sample size must be determined scientifically using formulas based on population size, desired confidence level, margin of error, and expected effect size. In Indian business research, common errors include arbitrary sample sizes (“I will take 100 respondents”) or grossly inadequate sizes (e.g., 30 respondents for a study requiring 200). Use software (G*Power, SPSS SamplePower) or published tables to determine minimum required sample size. Always justify your sample size in the methodology section. Adequacy is a mathematical, not a subjective, criterion.

5. Feasibility and Practicality

A good sample design must be feasible to execute given available time, budget, personnel, and access. A theoretically perfect design (e.g., stratified random sample of all Indian households) is useless if you cannot afford it or cannot access the sampling frame. Practicality means the design can be implemented without unrealistic assumptions. In Indian business research, feasibility constraints often force trade offs. For example, cluster sampling may be chosen over simple random sampling to reduce travel costs. Non probability sampling may be chosen when no sampling frame exists. A good design is not the absolute best in theory; it is the best that can actually be executed. However, do not sacrifice representativeness and precision beyond acceptable limits. Document all feasibility constraints and their impact on your findings.

6. Efficiency

Efficiency means achieving maximum precision for a given sample size (or minimum sample size for a given precision). An efficient sample design produces smaller sampling errors than alternative designs with the same cost. Stratified sampling is often more efficient than simple random sampling because it reduces variability within strata. Cluster sampling is less efficient (produces larger sampling errors) but may be more cost effective. In Indian business research, efficiency matters because resources are limited. Compare alternative sampling designs on both precision and cost. For example, a stratified design might require a sample of 400 to achieve the same precision as a simple random sample of 500. That is greater efficiency. Use design effect calculations (available in survey software) to quantify efficiency. Choose the design that gives you the most information per rupee spent.

7. Replicability

A good sample design must be described in sufficient detail that another researcher can exactly replicate the sampling procedure and obtain comparable results. Replicability requires transparency: document the sampling frame, selection method, random seed (if used), substitution rules, and response rate calculation. In Indian business research, replicability is often neglected. Vague descriptions like “a random sample was drawn” are insufficient. Specify: “Using SPSS random number generator with seed 12345, we selected 400 respondents from the customer database of XYZ ecommerce platform dated January 2024.” Replicability allows others to verify your findings and build upon your work. It also protects against fraud because fabricated data cannot be replicated. For probability sampling, replicability is straightforward. For non probability sampling, provide detailed inclusion criteria so others can apply the same judgment rules.

8. Flexibility

A good sample design anticipates practical problems and includes flexibility to handle them without introducing bias. For example, a flexible design includes a protocol for handling non response (e.g., three callbacks at different times before substitution). It includes contingency plans if the sampling frame is incomplete or outdated. In Indian business research, field conditions are often unpredictable: respondents may refuse, addresses may be wrong, or security issues may block access. A rigid design collapses under these conditions, forcing researchers to make ad hoc decisions that introduce bias. A flexible design builds in predetermined alternatives. However, flexibility does not mean changing the design after seeing data. All flexibility must be specified in advance in the sampling plan. Good design balances rigor with realistic adaptability to field conditions.

9. Operational Clarity

A good sample design defines every term and procedure operationally, leaving no ambiguity for the person executing the sampling. Operational clarity answers: Exactly who is included? Exactly how are they selected? Exactly what happens if a selected person is not available? In Indian business research, operational ambiguity is common. For example, “randomly select customers” is ambiguous. Does “customer” mean anyone who ever bought once, or only those who bought in the last 3 months? Does “randomly” mean using a random number generator or just picking arbitrarily? Write your sampling procedure as a step by step recipe that a research assistant could follow without asking clarifying questions. Operational clarity reduces researcher discretion, which reduces bias. It also simplifies replication and audit. Every sampling decision must be specified before execution, not decided during.

10. Alignment with Unit of Analysis

A good sample design ensures that the sampling unit matches the unit of analysis (the entity about which conclusions will be drawn). If your research conclusions are about individual customers, your sampling unit must be individual customers, not households or villages. In cluster sampling, this alignment requires careful attention. For example, a study of Indian ecommerce behavior might randomly select villages (primary sampling unit) and then randomly select households (secondary sampling unit), but the unit of analysis is individual adults within households. The sampling design must account for clustering and ensure that each individual has a known probability of selection. Misalignment leads to ecological fallacy (drawing individual conclusions from group data) or atomistic fallacy (drawing group conclusions from individual data). In Indian business research, clearly state your unit of analysis before designing sampling procedures.

Types of Sample Design:

Probability sampling is a technique where every element in the target population has a known, nonzero, and equal chance of being selected. This randomness eliminates selection bias and allows researchers to use statistical inference to estimate population parameters with quantifiable confidence levels. Probability sampling is preferred when generalizability is essential, such as in PhD research, policy studies, or large scale business surveys. In Indian business research, probability sampling is used when a complete sampling frame (e.g., customer database, voter list, employee roster) is available. The main drawback is higher cost and time compared to non probability methods.

Types of Probability Sampling:

1. Simple Random Sampling

In simple random sampling, every element in the population has an equal and independent chance of selection. Selection is done using lottery method, random number tables, or software generated random numbers. For example, to study customer satisfaction at an Indian ecommerce platform, you assign a unique number to each of 10,000 customers and use a random number generator to select 400. This method is unbiased and easy to understand. However, it requires a complete sampling frame and may miss small subgroups. It is inefficient for geographically dispersed populations. Simple random sampling is the foundation against which other probability methods are compared.

2. Stratified Random Sampling

Stratified random sampling divides the population into homogeneous subgroups (strata) based on a characteristic such as age, income, or location. Then simple random sampling is applied within each stratum. For example, an Indian ebanking study might stratify by urban and rural users, then randomly sample from each. This ensures representation of all subgroups and increases precision. Stratification requires accurate population data for each stratum. Sample sizes within strata can be proportional (same sampling fraction) or disproportional (oversampling small groups). In Indian business research, stratification by caste, language, or state is common to ensure diverse representation.

3. Cluster Sampling

Cluster sampling divides the population into naturally occurring groups (clusters) such as villages, schools, or city blocks. Randomly select clusters, then sample all elements within selected clusters (single stage) or randomly sample within clusters (two stage). For example, studying Indian rural ecommerce users, randomly select 20 villages from 500, then survey all households in those villages. Cluster sampling reduces travel and listing costs. However, it produces larger sampling error than simple random sampling because elements within a cluster tend to be similar (design effect). Cluster sampling is practical when no complete population list exists but geographic clusters are identifiable.

4. Systematic Sampling

Systematic sampling selects every kth element from a sampling frame after a random start. Calculate sampling interval k = population size divided by desired sample size. For example, from a list of 5,000 Indian ebanking customers, choose a random start between 1 and 10 (k = 5000/500 = 10), then select every 10th customer. This method is easier than simple random sampling and works well when the sampling frame has no periodic pattern. However, if the frame has hidden periodicity (e.g., every 10th customer is from the same city), bias occurs. Systematic sampling is widely used in Indian business research due to its simplicity.

5. Multistage Sampling

Multistage sampling combines two or more probability sampling methods sequentially, typically used for large, geographically dispersed populations. First stage: randomly select clusters (e.g., Indian states). Second stage: randomly select districts within selected states. Third stage: randomly select households. This design is cost effective for national surveys. Each stage uses simple random, systematic, or stratified sampling. The Indian National Sample Survey Office uses multistage sampling extensively. The drawback is increased sampling error (design effect) and complex analysis requiring specialized software. Researchers must calculate probabilities at each stage to produce valid population estimates.

Introduction to Non Probability Sampling

Non probability sampling is a technique where elements are selected based on the researcher’s judgment, convenience, or specific criteria, not on random selection. The probability of selection is unknown, and statistical inference to the population is not justified. Non probability sampling is faster, cheaper, and practical when no sampling frame exists or when the population is hard to reach. In Indian business research, it is common in exploratory studies, student projects, and corporate research where generalizability is not the primary goal. However, findings cannot be generalized with statistical confidence, and bias is inherent.

Types of Non Probability Sampling

1. Convenience Sampling

Convenience sampling selects elements that are easiest to access, such as surveying customers at a single mall, employees in one office, or students in a university class. The researcher picks whoever is available and willing. For example, an Indian MBA student studying ecommerce habits might survey 100 fellow students in the library. This method is extremely cheap and fast. However, it has high bias because the sample is unlikely to represent the broader population. Convenience sampling is acceptable only for exploratory, pilot, or illustrative research. In formal business research, conclusions from convenience samples must be treated as tentative, not generalizable.

2. Purposive Sampling

Purposive sampling selects elements based on the researcher’s judgment about who can best provide the required information. The researcher deliberately chooses participants who possess specific characteristics or expertise. For example, a study on Indian ebanking security might purposively select only cybersecurity officers and fraud victims. This method is useful for qualitative research, case studies, and expert opinions. The drawback is high researcher bias because selection depends entirely on subjective judgment. There is no statistical basis for generalization. Purposive sampling is appropriate when the research goal is deep insight from specific informants, not statistical representation of a population.

3. Quota Sampling

Quota sampling selects a predetermined number (quota) of elements from various subgroups, but within subgroups selection is non random (often convenience based). For example, an Indian study of ecommerce users might set quotas: 100 urban men, 100 urban women, 100 rural men, 100 rural women. The researcher then finds anyone fitting each quota. Quota sampling ensures subgroup representation like stratified sampling but without random selection. It is cheaper and faster than stratified sampling. However, within quotas, bias remains because the researcher chooses convenient elements. Quota sampling is common in market research but cannot produce statistically generalizable results.

4. Snowball Sampling

Snowball sampling starts with a few initial participants (seeds) who then refer other eligible participants. This method is used when the population is hidden, hard to reach, or has no sampling frame, such as illegal workers, drug users, or informal sector employees in India. For example, studying migrant construction workers, the researcher finds three workers who then refer others. Snowball sampling is efficient for reaching networked populations. However, it has severe bias because the sample is limited to people connected to the seeds. It overrepresents well connected individuals. Snowball sampling is appropriate for exploratory qualitative research but not for population estimation.

5. Self Selection Sampling

Self selection sampling occurs when individuals volunteer to participate, often in response to an open invitation. For example, an Indian ecommerce company emails all customers a survey link, and those interested click and respond. The sample consists entirely of volunteers. This method is easy and inexpensive for online research. However, volunteers differ systematically from non volunteers (volunteer bias). They may be more opinionated, have more free time, or have stronger positive or negative experiences. Self selection sampling cannot produce generalizable results. It is acceptable for customer feedback or exploratory studies but not for estimating population parameters. Always disclose that the sample is self selected.