honest-datavizlisted

# Honest dataviz A chart is read by people and executed by you. This skill turns "make it look good" into decisions with reasons, so the result is right by construction. Every rule here was bought with a measured failure in a production reporting engine — none is taste. ## First decision: is it even a chart? | The data is… | Use | Not | |---|---|---| | One current value (+ trend) | Stat tile: label, hero figure, semantic delta, small sparkline | A one-bar bar chart | | A handful of headline numbers | A KPI row of stat tiles | A grouped bar chart | | A ratio against a limit | A meter with the limit marked | A two-slice pie | | More than ~7 categories that all matter | A table (or table + chart) | More colours | | A verification story (checks, statuses) | A tracker strip: one cell per check | Prose | If a chart is right, the data's JOB picks the form: magnitude → bars; trend → line/area; part-to-whole → stacked bar (horizontal when names are long) or a two-segment donut at most; above/below a baseline → diverging bars around zero; one series against context → emphasis (one hue, grey rest). The most underused honest form is **emphasis**; the most overused dishonest one is categorical rainbow. ## Typography on charts — the two highest-leverage fixes 1. **At most three font sizes per view.** Generated dashboards carry five to eight; consolidating to three is the single biggest "suddenly looks designed" move. 2. **Pull the weights back.** Nothing above semibold. Bold ev