Data Analytics Techniques and Tools
What you must know
Understand the analytics maturity ladder: descriptive (what happened), diagnostic (why), predictive (what will happen) and prescriptive (what should we do). Know core techniques - aggregation, filtering, joins, sorting, trend analysis, variance analysis, ratio analysis, correlation versus causation, regression, clustering, classification and anomaly detection. Be familiar with tool categories: spreadsheets, SQL databases, ETL/ELT tools, BI platforms, statistical languages (Python, R) and visualisation tools. Understand the difference between population testing (full-data analytics) and sampling, and between deterministic rules (e.g. duplicate payment matching) and probabilistic models.
Traps in questions
Correlation is routinely confused with causation in distractors. Another trap is treating a predictive model's output as fact rather than a probability with error rates. Questions may test whether a technique suits the objective: clustering is unsupervised segmentation, classification is supervised prediction of a known label. Watch for confusion between ETL (transform before load) and ELT (load then transform in the target). Also remember that a larger sample reduces sampling risk but not non-sampling risk such as flawed logic.
Rule of thumb
Match the technique to the question type: comparison → aggregation and variance; relationship → correlation or regression; grouping → clustering; prediction of a known outcome → classification; unusual items → anomaly detection.
Worked example
An auditor wants to identify duplicate supplier payments. The right approach is a deterministic rule: match on supplier ID, invoice number, amount and date within a tolerance window across the full population. Regression would be the wrong tool because the objective is exact rule-based identification, not prediction.