Review of Doppio (brand, style guide, landing, dashboard)

Method: a rendered lint pass on the static pages, real-browser measurements and screenshots (desktop, 390px mobile, dark mode) of all three deliverables, and a read of each against PLAN.md and against each other. Findings are in priority order; each one has a fix status in CHANGES.md.

P1: broken or misleading

  1. Landing hero headline is fragile. The fitted Six Caps line is sized once, from a measurement taken when document.fonts.ready fires. If the webfont lands after that (slow network, cache miss) the headline is measured in a fallback face and then overflows the viewport, which is exactly what the lint render captured at 1280px ("BOOK A BARISTA. BRIN..." cut off). On a 375px phone the single fitted line is also only about 30px tall, so the strongest brand element is the weakest on the device customers will use most.
  2. The dashboard shows a problem it cannot solve. "Need a barista: 3" is a dead stat, and the booking detail panel says "Assign one from the roster" but there is no way to assign anyone. Filling short-staffed bookings is the central job of a roster business, so the dashboard reads as a mock rather than a tool.
  3. Sample data contradicts itself. Booking 2038 (140 guests) is "Confirmed" with 2 baristas, but the rule everywhere (PLAN.md, landing, panel) is one barista per 60 guests, so it needs 3. The same booking, #2041, is $1,340 in the style guide ticket but $1,100 in the dashboard and by the price maths. The style guide stats (87%, $18.4k, 42) do not match the dashboard (30%, $21,860, 10).

P2: quality, accessibility, coherence

  1. Auto-looping hero ticket has no pause (WCAG 2.2.2: moving or updating content over 5 seconds needs a control). It also keeps looping after the user opens the booking preview.
  2. Weak focus and hit targets in the dashboard. Clickable table rows and schedule blocks rely on shadcn's 50% ring outline, which is hard to see on buff paper. Filter buttons and rows work by keyboard but the indicator is too faint to trust.
  3. Brand leaks. The dashboard tab says "vite-app" with the Vite favicon; the landing page and style guide have no favicon. The first thing a customer sees in a tab is not the brand.
  4. Mobile landing details. The pricing table headers wrap into three-line cells; the nav has no links below 60rem at all; the interactive ticket reserves empty space for a stamp that is not showing; the booking form sits outside <main>.
  5. Mobile dashboard. Four full-width stat cards push the chart and the bookings below three screens of scrolling.
  6. Booking form validation is partial. Only guests is checked; an empty date or malformed email "succeeds". The Reschedule toast claims "Moved to Thu 8:00am" even when nothing changed.
  7. The landing page persuades too thinly. It shows how to book but not what arrives (machine, milks, cups, clean-down). The brand promise "bring the bar" is never made concrete. No photographs exist, so concrete particulars have to carry it.

P3: polish and documentation

  1. Schedule does not show who is working (PLAN.md section 3 promised barista lanes); blocks only show client and setup.
  2. Naming drift. "Tasting" (landing, segmented) vs "Tasting table" (pricing, dashboard).
  3. Style guide gaps. It does not document the app density (14px UI vs 16px body), the dark/inked variant used by the dashboard, or the focus ring spec. Its hero uses the same one-shot fitting as the landing. One lint FAIL remains (word density on the colour screen, a reference sheet by nature).
  4. PLAN.md drift. It promises an FAQ and a 3-anchor nav; the build has no FAQ and four anchors.
  5. Generated shadcn hook use-mobile.ts fails the React-hooks lint rule (set state in effect). Not ours, left as is.