Data Analytics and Decision Support
Data Analytics and Decision Support tests whether you can turn data into a defensible decision rather than a convincing chart. Questions cover analytics types, data quality, variation, sampling, visualization, decision models, governance, and bias.
Core rules and concepts
- Distinguish descriptive, diagnostic, predictive, and prescriptive analytics. Each answers a different question: what happened, why, what is likely, and what should be done.
- Validate data before analysis: source, extraction logic, completeness, definitions, refresh timing, and known gaps. Document transformations so results are reproducible.
- Separate common-cause variation from special-cause variation. A control chart distinguishes routine noise from a signal worth investigating; a run of points or a point beyond a limit is a signal, not proof of a cause.
- Use sampling that fits the decision. State the population, sample frame, size, and selection method. A biased sample cannot be rescued by a larger sample.
- Choose visuals that match the comparison: trend over time, distribution, relationship, or composition. Avoid dual axes and truncated scales that exaggerate differences.
- Apply data governance: ownership, access, privacy, retention, and quality rules. A model without governance is an unmanaged risk.
Worked example
A dashboard shows cycle time rising for three weeks. The operations lead wants to retrain the whole team. A control chart shows the process has routine variation within limits, plus one point above the upper control limit on a day when a system outage occurred. The reasoning is to investigate the special-cause day first, confirm the outage link, and avoid changing the standard process in response to common-cause noise. If the chart shows a sustained shift after the outage, the team reviews the recovery backlog and staffing. The decision rule is written in advance: special cause -> investigate the event; common cause -> change the system; no signal -> do not tamper.
Traps that cost marks
- Reacting to every up-and-down movement as if it were a trend.
- Cherry-picking a favorable period or subgroup after seeing the result.
- Using a precise model on data with unclear definitions or missing cases.
- Confusing statistical significance with practical importance.
- Presenting a chart that hides uncertainty, sample size, or comparison baseline.