Built from the live dashboard, not a mockup. Each option below uses Site Intel's own tokens and real data from the bbb.org run. Nothing here is implemented — pick a direction and I'll build it.
Screenshotting the live dashboard surfaced a real defect. The action row cannot shrink below its own content, so it claims the whole row and the page's <h1> — the URL being audited — collapses to zero width and renders one character per line.
Measured at a 1600px viewport. Playwright reports that heading as hidden.
The eight buttons are all the same weight, so nothing tells you which one you came here to press — and together they starve the title. This is a one-line CSS fix, but it is worth fixing alongside whichever direction you choose.
The smallest change that works. The title gets a guaranteed minimum width, one action is primary, the rest collapse into a menu. Everything below the header stays exactly as it is today.
Risk: very low. One CSS rule plus moving five buttons into the existing MoreSheet. Nothing else on the screen moves.
Today the score, coverage and provenance sit in one tall stack above the content, so once you scroll into the findings you lose them. This moves them into a sticky rail: the numbers stay put while you read, and the content column gets the full width.
Risk: medium. A real layout change to /analyses/[id], but additive — the tabs and their content are untouched. Below 860px the rail simply stacks back on top.
The most opinionated option, and the one the original plan called b-queue. The screen opens on a sentence, then an ordered list of work ranked by how many pages each issue touches — so a warning on 240 pages outranks a lone critical. The score becomes a caption.
Shown here with a site that has real problems, since bbb.org scores 100 and would make an empty queue.
Risk: high. This replaces what the screen is for. It is also the only option that answers "what should I do next" without you reading a chart first. Severity is carried by shape as well as colour, so it survives colour-blindness and forced-colours mode.
It fixes a defect that is live right now, on the screen you use most, and it costs almost nothing. There is no argument for leaving a zero-width <h1> in production while we debate layout.
B is comfort — it makes a screen you already know how to read less annoying to scroll. It does not change what the screen tells you.
C changes the job of the screen: from reporting a score to assigning work. That is the bigger win and the bigger risk, and it is worth doing only if the reach-ranked order genuinely matches how you decide what to sell and fix. You are the one who knows that — I would not ship C on my own judgement.
B is the one I would skip. It spends a real layout change on a comfort improvement, and if C lands afterwards most of that work gets thrown away.