13  Data Collection Methods

Every analysis — descriptive, diagnostic, predictive, or prescriptive — is only as trustworthy as the data behind it. Data collection is the process of gathering information in a systematic way so that it can be measured, analysed, and used to answer a business question. Before a single chart is drawn or a single test is run, an analyst has to decide what to measure, from whom, and how — and those choices shape everything that follows.

13.1 Why Data Collection Matters

  • Poor data collection produces poor analysis, no matter how sophisticated the statistical technique.
  • The method chosen determines the type of data obtained (numeric, categorical, text, behavioural) and therefore which analyses are even possible.
  • Well-planned collection reduces bias, missing values, and rework later in the analytics pipeline.
  • It connects the business question directly to the evidence that will be used to answer it.

13.2 Primary vs Secondary Data

Primary Data

Data collected first-hand by the researcher or organisation for a specific purpose. Example: A retailer surveys its own customers about a new loyalty programme.

Secondary Data

Data that already exists, having been collected by someone else for a different original purpose, and is reused for the current analysis. Example: Using government census figures or a published industry report to estimate market size.

Primary data is more expensive and time-consuming to collect but is tailored exactly to the question at hand; secondary data is faster and cheaper but may not fit perfectly.

13.3 Methods of Primary Data Collection

Surveys and Questionnaires

Structured sets of questions administered to a sample of respondents, in person, by phone, or online. Example: A customer-satisfaction survey sent by email after a purchase.

Interviews

One-on-one or small-group conversations that allow deeper, open-ended exploration of a topic. Example: Exit interviews with employees leaving an organisation to understand attrition drivers.

Observation

Recording behaviour as it naturally occurs, without direct interaction with the subject. Example: Tracking how long shoppers linger in different aisles using in-store cameras.

Experiments

Manipulating one or more variables under controlled conditions to measure their effect on an outcome. Example: An A/B test comparing conversion rates between two website designs.

Focus Groups

Guided group discussions used to gather qualitative insight on attitudes, perceptions, and reactions. Example: Testing reactions to a new product concept with a panel of target consumers.

13.4 Methods of Secondary Data Collection

Internal Records

Data an organisation already holds as part of its normal operations, such as sales transactions, CRM logs, or financial statements.

Government and Institutional Publications

Census data, labour statistics, and regulatory filings published by government bodies or international organisations (for example, the World Bank or the RBI).

Published Research and Industry Reports

Academic journals, market-research reports, and white papers that summarise data already collected and analysed by others.

Online and Public Databases

Open datasets, APIs, and web-scraped data available through platforms such as data.gov, Kaggle, or company APIs.

13.5 Choosing the Right Method

The choice between methods — and between primary and secondary data — depends on several practical factors:

  • Research objective: Does the question require attitudes and opinions (survey/interview) or actual behaviour (observation/experiment)?
  • Cost and time: Secondary data and online surveys are usually cheaper and faster than experiments or field interviews.
  • Accuracy and control: Experiments give the most control over cause-and-effect but are the hardest to run at scale.
  • Sample access: Some populations (senior executives, rare medical conditions) are only reachable through targeted interviews rather than mass surveys.
  • Ethical and privacy constraints: Observation and data reuse must respect consent and data-protection regulations.

13.6 Common Challenges and Best Practices

Challenges
  • Non-response and self-selection bias in surveys.
  • Observer effects, where people behave differently because they know they are being watched.
  • Secondary data that does not exactly match the current definitions or time period needed.
  • Cost and logistics of large-scale primary data collection.
Best Practices
  • Pilot-test instruments (questionnaires, interview guides) before full deployment.
  • Clearly define the target population and sampling frame before collecting data.
  • Document the source, date, and method of every dataset used, primary or secondary.
  • Combine methods (triangulation) where possible to cross-check findings.

Transition to Sampling

Deciding how to collect data is only half the problem — an analyst rarely has the time or budget to collect data from an entire population. The next chapter addresses who to collect it from: the sampling techniques used to select a representative subset of a population.

Summary

Concept Description
Foundations
Data Collection The systematic process of gathering information so it can be measured and analysed to answer a business question
Primary Data Data collected first-hand by the researcher for the specific purpose at hand
Secondary Data Existing data originally collected for another purpose and reused for the current analysis
Primary Methods
Surveys and Questionnaires Structured question sets administered to a sample, in person, by phone, or online
Interviews One-on-one or small-group conversations that allow deeper, open-ended exploration
Observation Recording naturally occurring behaviour without direct interaction with the subject
Experiments Manipulating variables under controlled conditions to measure their effect on an outcome
Focus Groups Guided group discussions used to gather qualitative attitudes and reactions
Secondary Methods
Internal Records Data an organisation already holds from its normal operations, such as sales or CRM logs
Government Publications Census, labour, and regulatory data published by government or institutional bodies
Published Research Academic journals, market-research reports, and white papers summarising existing findings
Online/Public Databases Open datasets, APIs, and web-sourced data available through public platforms
Choosing a Method
Research Objective Whether the question needs attitudes/opinions or actual behaviour drives the method choice
Cost and Time Secondary data and online surveys are typically cheaper and faster than experiments or field work
Accuracy and Control Experiments offer the most control over cause-and-effect but are hardest to scale
Ethical Constraints Consent and data-protection regulations must be respected in observation and data reuse
Challenges/Practices
Non-response Bias Bias introduced when certain types of respondents are less likely to participate in a survey
Observer Effect People behaving differently because they know they are being observed
Pilot Testing Testing an instrument on a small group before full deployment to catch design flaws
Triangulation Cross-checking findings by combining more than one data-collection method