Learning Objectives
By the end of this lesson, learners should be able to:
- Explain the purpose and principles of business data collection.
- Distinguish primary and secondary data collection.
- Evaluate major data collection methods.
- Assess data sources for relevance, reliability and suitability.
- Identify potential sources of bias and error during data collection.
- Select appropriate data collection approaches for different business problems.
1. Meaning of Data Collection
Data collection is the systematic process of obtaining information required for a defined business, analytical or research purpose.
Effective collection begins with a clear understanding of:
- The business objective.
- The population or subject being studied.
- The variables required.
- The required level of accuracy.
- The timeframe.
- Ethical and legal requirements.
Data should therefore not be collected simply because it is available.
2. Primary Data
Primary data is collected specifically for the current analytical purpose.
Common methods include:
- Surveys.
- Interviews.
- Observations.
- Experiments.
- Focus groups.
- Direct measurements.
Advantages
Primary data can be designed specifically around the research question.
Limitations
It can require substantial:
- Time.
- Cost.
- Planning.
- Human resources.
Poorly designed primary-data collection can also introduce significant bias.
3. Secondary Data
Secondary data is data originally collected for another purpose and subsequently used for a new analytical objective.
Examples include:
- Published industry statistics.
- Government datasets.
- Historical company records.
- Academic research.
- Commercial databases.
Secondary data can reduce collection costs, but analysts must determine whether its original definitions and collection methods are suitable for the current problem.
4. Surveys
Surveys collect information from respondents using structured questions.
They can be conducted through:
- Online forms.
- Telephone interviews.
- Paper questionnaires.
- Digital applications.
The quality of survey results depends heavily on:
- Question design.
- Sampling.
- Response rates.
- Respondent understanding.
- Data recording.
Leading or ambiguous questions can systematically distort results.
5. Interviews
Interviews involve direct interaction with respondents.
They may be:
- Structured.
- Semi-structured.
- Unstructured.
Interviews can provide detailed contextual information that may not be captured through standardized questionnaires.
However, interviewer influence and respondent interpretation can introduce bias.
6. Observation
Observation involves systematically recording behaviors, events or processes.
It may be useful where actual behavior differs from reported behavior.
For example, an organization may observe how customers navigate a digital platform rather than relying entirely on customers’ descriptions of their behavior.
7. Experiments
An experiment involves deliberately changing one or more conditions and examining the resulting effect.
Controlled experiments can be useful for assessing causal relationships.
For example, an organization may test two versions of a digital interface and compare user outcomes.
Experiments require careful design to ensure that differences in outcomes can reasonably be attributed to the factor being tested.
8. Administrative and Transactional Data
Organizations generate large volumes of data through routine operations.
Examples include:
- Sales transactions.
- Payments.
- Customer service records.
- Employee records.
- Inventory movements.
- Website activity.
These datasets can be valuable for analytics because they capture actual organizational activity.
However, they were often created for operational rather than analytical purposes and may therefore contain limitations.
9. External Data Sources
External data can provide information about factors outside an organization.
Examples include:
- Economic indicators.
- Industry statistics.
- Market research.
- Public datasets.
- Commercial information providers.
External data can help place internal performance into a broader context.
10. Sampling
When it is impractical to collect data from an entire population, analysts may use a sample.
A sample should be selected in a manner that supports valid conclusions about the population of interest.
Important concepts include:
- Population.
- Sample.
- Sampling frame.
- Sampling method.
- Sampling error.
Poor sampling can produce misleading conclusions even when the collected observations are accurately recorded.
11. Sources of Data Collection Bias
Bias can arise at multiple stages.
Examples include:
- Selection bias: Certain groups are systematically more likely to be included.
- Non-response bias: People who do not respond differ meaningfully from respondents.
- Measurement bias: The collection method systematically produces inaccurate measurements.
- Interviewer bias: The interviewer influences responses.
- Recall bias: Respondents inaccurately remember past events.
Recognizing potential bias is essential for trustworthy analytics.
12. Evaluating a Data Source
Before using a dataset, analysts should ask:
- Who collected it?
- Why was it collected?
- When was it collected?
- How was it collected?
- Who was included?
- What definitions were used?
- How complete is it?
- Are there known limitations?
- Is its use legally and ethically appropriate?
The credibility of a source depends not only on who provides it but also on how the data was generated.
Lesson Summary
Data collection is the systematic acquisition of information for a defined analytical purpose.
Primary data is collected specifically for the current purpose, while secondary data was originally collected for another purpose.
Common collection methods include surveys, interviews, observation, experiments and operational data capture.
Effective data collection requires careful consideration of sampling, measurement, bias, relevance, reliability, ethics and the intended analytical use.
References
- DAMA International — DAMA-DMBOK
DAMA International - International Organization for Standardization — ISO 8000 Data Quality
ISO 8000 Data Quality - OECD — Data and Digital Policy
OECD Data and Digital Policy
Review Questions
- What is data collection?
- Why should data collection begin with a clearly defined analytical objective?
- What distinguishes primary from secondary data?
- What are the strengths and limitations of surveys?
- Why can interviews introduce bias?
- When might observation be preferable to self-reported information?
- Why are experiments useful for investigating causal relationships?
- What distinguishes operational data from purpose-collected research data?
- Why is sampling necessary in many analytical projects?
- What are the major sources of data collection bias?
- What factors should an analyst consider before using an external dataset?
- Why does the original purpose of data collection matter when evaluating secondary data?