I designed a legal-entity dashboard for Google Finance so reviewers can find an entity, see its posture, and trace a figure to its source.

The problem was ordinary and costly: entities lived in different systems with different labels, rollups bent on edge cases, and shadow spreadsheets tried to make sense of it. The aim was a page you could trust at close. One canonical ID. A header that tells you what you’re looking at, what’s missing, and when it last refreshed. If a rollup is partial, it says so.

Movement stays narrow. Type to find an entity, pick the period, apply a small set of tested facets, and read a table you can explain out loud. Drilldowns open in place so context holds. Lineage opens at the side so you can see where a figure comes from and keep your spot.

Variance needs a story, so we gave it one. The explainer ties change to the usual causes in this system: FX, mapping changes, one-time adjustments, timing. Notes can be pinned to a row so the next reviewer doesn’t start from zero. I prototyped with sample data and sat with finance to see whether “what changed and why” could be answered without leaving the page. When it couldn’t, we fixed the control or the copy.

We measured the basics with data science: filter apply, drill-down, variance explainer open, lineage view, export, notes. Signals were time to first answer and how often someone bounced to a spreadsheet. Guardrails were completeness and freshness. If a source failed, the page said which one. The result is quiet: a name you can anchor to, a straight path to evidence, and less second-guessing on routine days.