A dashboard with forty charts and no recommendation is a reporting failure, not an analytics success. The exam tests whether you can name the technique that fits a question, read a regression output, and pick a chart that does not distort.
Big Data is data too large, too fast, or too varied for traditional tools. Volume is sheer size. Velocity is the speed of arrival, such as streaming point-of-sale feeds. Variety is the mix of formats: numbers, text, images, sensor logs. Veracity is trustworthiness, meaning accuracy, completeness, and freedom from bias.
TRAP: Veracity is the V candidates forget. Data that is huge, fast, and varied but wrong is worse than no data, because it carries false authority.
- Structured data: fixed rows and columns in a relational database. General ledger entries, sales orders, payroll records.
- Semi-structured data: self-describing tags without a fixed schema. Extensible Markup Language (XML) and JavaScript Object Notation (JSON) files, electronic data interchange (EDI) messages, email headers.
- Unstructured data: no predefined model. Contracts, call recordings, images, social posts. This is the majority of enterprise data and the hardest to analyze.
Common mistakes
- Reading as accuracy. An of 0.87 explains variation; it says nothing about bias or whether the relationship holds outside the observed range. Pair it with the standard error of the estimate.
- Using clustering when labels already exist. If the fact pattern names the outcome categories, the tool is classification. Clustering is only for discovering groups you did not define.
- Choosing logistic regression for a dollar amount. Logistic predicts a probability of a binary event. A predicted cost or revenue in dollars calls for multiple regression.
Bottom line
- Four Vs: volume (size), velocity (speed), variety (formats), veracity (trustworthiness)
- Data types: structured (rows and columns), semi-structured (XML, JSON, tagged), unstructured (text, audio, images)
- Progression: data to information to knowledge to insight to action, and value arrives only at action
- Analytics types: descriptive (what happened), diagnostic (why), predictive (what will happen), prescriptive (what to do)
Exam shortcut
When a question names the outcome categories in advance, answer classification; when it asks you to discover segments that were never defined, answer clustering. The presence or absence of labels decides it before you read the rest of the stem.
The full lesson (about 2,379 words, 16 min read) adds 2 worked examples, all 6 common mistakes, a self-check, free in the app.
Learning objectives
- 1F4
Browse all free CMA Part 1 lessons or jump into free CMA Part 1 practice questions.