Adoption

1Round 1 · Round 3Decide, then what changed, then detail. The alert leads the page, names its count and its amount so the reader can weigh it, and links to the evidence. It uses Frost's warning surface with its own surface-foreground, the triangle-alert glyph and role="status". It deliberately has no edge treatment, because Frost rule K4 is still open.

2Round 1 · Round 2One hero, the rest recede. Adoption rate is the number this page is about, so it is the only colored numeral, at the ladder's 54px in --primary. The other three sit at 34px in foreground ink. The qualifier appears once for the whole band rather than four times. Numerals are Barlow Semi Condensed, chosen in round 1.

3Round 3Direction is declared per metric, never inferred. Unused spend rising is bad, adoption rate rising is good, and both are marked accordingly. A dashboard that lets one color logic run across a mixed set of metrics produces confident misreads. Every delta carries its glyph, so direction never rides on color alone.

4Round 3 · build 6The gaps are the finding, not the totals. This is the pattern the library names ClassLink Analytics inside, and round 1's cost view was a crude version of it: totals loud, drop-off invisible. Here the losses between stages carry the emphasis and the stage figures recede. It is a budget display as much as a usage one.

5Round 3"Active" is defined on the face of the component. Not in a tooltip, because the person defending a renewal has to quote the definition. Without it the number means nothing.

6Round 1 · Round 2A bar never travels without its number, and the header names the maximum. The adoption column is the inline-bar recipe from round 1. The "vs district" column uses round 2's validated diverging ramp, whose warm pole is deliberately not a status color: below the district average is not an error. Click a school to go deeper.

7Round 3Any matrix is read as a gradebook. Rows here are schools, not students, so the first column says so and the title is explicit. Sortable by any column; the sort is stated in the finding line above so the table always says what it is showing.

8Round 3 · build 5Suppression is expected, not optional. Groups below the reporting minimum render as a dash and never as a number or a zero, the threshold sits in the legend on the face of the table, and the count of suppressed groups is visible. A table without this reads as naive or non-compliant and erodes trust in everything else the product says. It opens at 30, a minimum several states use for federal reporting, which already hides three groups. Drop it to 10 and they come back. That trade between privacy and visibility is the real subject of the table, and it should be arguable on the surface rather than settled silently in a query.