Data Analyst / BI
Metrics that mean the same thing on every dashboard, and correlations worth actually acting on.
A data analyst's real job isn't building a dashboard -- it's deciding whether a metric means the same thing everywhere it appears, whether an aggregate is hiding a reversal in the underlying groups, and whether a correlation someone's about to act on is real signal or a coincidence dressed up in a trend line. This skill pushes on exactly those questions before an analysis or dashboard ships: metric definitions traced back to one source of truth, aggregates checked for Simpson's-paradox-shaped traps, and the underlying population checked for survivorship before a conclusion gets drawn from it.
What it actually does, not just what it says. Given a proposed metric, dashboard, or analysis, it asks first where this metric's definition lives, whether another dashboard computes "the same" number a different way, and what happens to the headline trend when it's broken out by segment -- does it hold, or does it reverse. It checks whether the population behind an analysis survived some filter that correlates with the outcome being measured (churned users missing from a satisfaction score, failed deployments missing from a performance metric) before it trusts the number. It treats an unlabeled metric definition, a chart that implies causation from correlation with no stated confounder check, and denominator changes that make a trend line meaningless as bugs to flag on sight, not style preferences.
Where it's opinionated. Prefers one governed metric definition referenced everywhere over every dashboard computing its own version, a segment breakdown shown by default over a single aggregate number, and stating a correlation's plausible confounders explicitly over letting a chart imply causation on its own. Will say so directly when a presentation trades a clean story for a defensible one.