Business Management · Using data, company functions, people skills, influence and clear communication

Data Types, Sources and Quality in Business Decisions

Updated 11 October 2026 · Fact-checked

Data for business decisions is either structured (fixed fields, such as policy records) or unstructured (free text, images, call notes). It comes from internal or external sources. Before using it, check completeness, accuracy, consistency, timeliness, relevance and validity, then state the limitations and how they affect the decision.

Understand Data Types, Sources and Quality in Business Decisions

Data is recorded facts you use to understand a situation and choose an action. A decision is only as good as the data behind it. In CB3 you are tested on whether you can pick suitable data, judge it, and say honestly what it can and cannot tell you.

Structured data sits in fixed fields and formats, such as a policy database with age, sum assured, premium and claim amount. It is easy to sort, total and analyse. Unstructured data has no fixed layout. Examples are emails, customer complaints, call recordings, social media posts and scanned documents. It is rich in meaning but needs effort to turn into something you can count. Some data is semi-structured, such as forms with a free-text comment box.

Internal sources come from inside the organisation: policy administration systems, claims files, sales and finance records, customer service logs and past project results. They are specific to your business and usually cheap to get. External sources come from outside: regulator publications, industry bodies, census and government statistics, market research, rating agencies, competitor reports and bought datasets. They give context and benchmarks but may not match your own customers.

Good decisions need quality data. Quality has several dimensions: completeness (nothing missing), accuracy (values are correct), consistency (same item recorded the same way across systems), timeliness (up to date), relevance (fits the question) and validity (values are in the allowed range and format). Also ask where the data came from, who collected it, and whether it is large enough to rely on.

Every dataset has limitations. It may be old, biased, small, or collected for another purpose. It may show a link without proving a cause. In an answer, do not stop at listing checks. Say what you would do about each problem: clean the data, adjust for it, add a margin for uncertainty, collect more, or tell the decision maker how far to trust the result.

Key rules to remember

Completeness rate: Completeness = (number of records with the field filled ÷ total records) × 100%; A simple way to quantify missing data for one field. It says nothing about whether the filled values are correct.
Error rate from a sample: Error rate = (records found wrong in sample ÷ sample size) × 100%; Only an estimate for the whole dataset. It is reliable only if the sample is chosen without bias.
Quality dimensions checklist: Completeness, Accuracy, Consistency, Timeliness, Relevance, Validity; Use these as a memory list for any data quality question. Add source and sample size where relevant.
Data classification: Structured / semi-structured / unstructured; internal / external; Classify any dataset on both scales before judging it.

How to solve Data Types, Sources and Quality in Business Decisions questions

Use this order for any exam question on data types, sources and quality in a decision.

1Restate the decision and the question the data must help answer.
2Classify the data: structured or unstructured, internal or external. Link each type to what it can offer.
3Name the sources you would use and say why each is relevant to this decision.
4Check quality using the dimensions: completeness, accuracy, consistency, timeliness, relevance, validity. Apply each to the case facts given.
5Say how you would test or fix each problem, for example reconciliation, sampling, cleaning, or seeking a better source.
6State the limitations that remain: age, bias, small size, different purpose, correlation not cause.
7Conclude with a recommendation and say how much weight the decision maker should put on the data.

Quickest way: Type, Source, Quality, Limits (TSQL)

When to use it: Use in short written questions and when you have only a few minutes for a case-study part.

Type: write one line classifying the data as structured or unstructured.
Source: name one internal and one external source with a reason each.
Quality: list three or four dimensions that matter most in the case and tie each to a fact in the question.
Limits: give two limitations and one action for each.
Finish with one sentence on how far the data can be trusted for the decision.

Common mistakes in Data Types, Sources and Quality in Business Decisions

Listing the quality dimensions without applying them to the case.
Students memorise the list and write it as a definition.
Fix: For each dimension, link it to a specific fact in the scenario and say what problem it could cause there.
Treating unstructured data as useless because it is hard to count.
Numbers feel more reliable than text.
Fix: Explain that unstructured data, such as complaints, can reveal causes and customer views that numbers miss, and say how it can be coded or summarised.
Assuming internal data is always better than external data.
Internal data feels more relevant and trusted.
Fix: Note that internal data may be small, biased to existing customers or poorly recorded. External data adds benchmarks and wider context. Use both where possible.
Ignoring the purpose for which the data was originally collected.
Students focus on the data values and not on how they were gathered.
Fix: Ask why and how the data was collected. Data gathered for billing may not suit a study of customer behaviour.
Confusing correlation in the data with cause.
A clear pattern looks like an explanation.
Fix: Say that a pattern shows association only. Suggest further analysis or testing before acting on a causal claim.
Stopping at problems without saying what to do.
Students think identifying issues earns the marks.
Fix: For each issue, give an action: clean, reconcile, adjust, add a margin, collect more data, or report the limitation to the decision maker.

Worked examples

Example 1

An Indian health insurer wants to decide whether to revise premiums for a group medical product. It has claim records from its policy system, free-text notes from claims handlers, and a published industry report on medical cost inflation. (a) Classify each data item. (b) Name two quality checks you would carry out on the claim records.

Show the solution

Claim records in the policy system are structured data and internal, because they sit in fixed fields and come from the insurer's own system.
Claims handlers' notes are unstructured data and internal, because they are free text with no fixed layout.
The industry report on medical cost inflation is external. Its figures are usually structured, though it may also contain narrative text.
Quality check 1, completeness: count the records where key fields such as age, diagnosis or claim amount are blank. For example, if 4,000 of 50,000 records lack diagnosis, completeness for that field is 46,000 ÷ 50,000 = 92%.
Quality check 2, validity and consistency: look for impossible values such as negative claim amounts or dates of claim before the policy start, and check that the same hospital or diagnosis is coded in the same way across the records.
Fix problems found: investigate blanks, correct or exclude invalid records, and recode inconsistent entries before using the data for the premium review.

Answer: Claim records: structured, internal. Handlers' notes: unstructured, internal. Industry report: external. Check completeness (e.g. 92% for diagnosis in the example) and validity and consistency, then clean the data before the premium decision.

Example 2

A life insurer plans to launch a product aimed at young professionals in a new city. It proposes to rely only on its existing customer database to estimate demand. Discuss the limitations of this data and suggest how to improve the evidence.

Show the solution

Decision: estimate demand for a new product in a new city. The data must represent the target market.
Relevance: the existing database covers current customers, who may be older, in other cities and with different incomes. They may not represent young professionals in the new city.
Bias: current customers are people who already chose this insurer. Non-customers and people who rejected it are missing, so demand may be overstated or misjudged.
Timeliness: customer data may be several years old and may not reflect current income levels or preferences.
Completeness: fields such as occupation or income may be missing, which limits analysis of the target group.
Improve the evidence: use external sources such as census and government statistics for age and income in the new city, market research or surveys of young professionals, and competitor product information.
Also use unstructured data, such as agent feedback and customer comments, to understand needs that the numbers cannot show.
Conclude that the database is a useful starting point but should not be the sole basis. Add a margin for uncertainty and consider a pilot launch.

Answer: The internal database is limited by poor relevance, selection bias, possible age and missing fields. Supplement it with external statistics, survey research, competitor information and agent feedback, and treat demand estimates with caution, for example by piloting first.

Exam tips

Always tie each point to a fact in the case. Generic lists score low in written parts.
Learn the quality dimensions as a list and apply at least three of them to the data in the question.
For MCQs, watch the exact wording: structured means fixed fields, unstructured means no fixed format, and internal or external depends on who holds the data.
Always add what you would do about a limitation, not only that it exists.
End with a clear view on how far the decision maker can rely on the data.

Practice questions from Using data, company functions, people skills, influence and clear communication

Data Types, Sources and Quality in Business Decisions in other exams

The same ground in other exams, if you are preparing for more than one or want another angle on it.

Data Types, Sources and Quality in Business Decisions: frequently asked questions

What is the difference between structured and unstructured data?

Structured data sits in fixed fields and formats, such as a policy database, so it is easy to count and analyse. Unstructured data has no fixed layout, such as emails, call notes or social media posts. It needs more effort to analyse but often holds useful detail.

What are internal and external data sources?

Internal sources are held inside the organisation, such as claims, sales and customer service records. External sources come from outside, such as regulator reports, government statistics, industry studies and market research. Good decisions usually use both.

How do I assess data quality in an exam answer?

Go through completeness, accuracy, consistency, timeliness, relevance and validity, and apply each to the facts given. Then say how you would check or fix each problem. Finish with the limitations that remain.

Does this topic need calculations?

Usually not much. You may be asked for a simple figure such as a completeness percentage or error rate from a sample. The marks mostly come from classifying data, applying quality checks and discussing limitations.