It wasn't sparse. It was small.
You were right to not be sold, and it turns out not to be a taste problem — it is a measurement problem I introduced. Every mock so far was about 500px wide with a 142px menu, roughly a third of a real admin screen. At that size a 15px heading surrounded by white space reads as clean. At 1440px with a 160px menu the identical proportions read as empty — the content is floating in three times the room and none of the type grew to meet it.
So: House palette locked, nothing recoloured, drawn at 1440. Five devices become possible once there is room, and every one of them is something we were already doing at the wrong size. That is the test — if "spazz it up" had needed new concepts, the system would be too thin to ship.
The sheet, full bleed
Admin menu hidden — the convention 4 of 4 LMS plugins follow. We own the viewport, so this is where the crop marks finally mean something and where a 96px numeral fits.
step 3 of 5 · search index
Where should search actually run?
This decides where the index lives, not what gets indexed. You can change it later without losing anything — reindexing takes about a minute for a site this size.
what you are choosing
Hairlines and tiles only — no illustration and no icon set, so it costs nothing to ship and nothing to maintain. It also does the one thing a paragraph cannot: shows that the choice sits in the middle of a pipeline the reader already owns both ends of.
pick one
this site
- posts
- 842
- pages
- 67
- courses
- 339
- total
- 1,248
- largest
- 41 kB
Read from your database just now. It is the number the recommendation is based on — under ~10,000 items, This site is genuinely fine and cheaper.
searchsync 2.1.0 · sheet 03 / 05
The three additions that carry it: the ghost 03 gives the page a focal point that costs nothing and cannot be mistaken for advertising; the schematic turns a settings choice into an explanation; and the right-hand aside only exists because 1080px of sheet leaves room for it — and it is the most persuasive thing on screen, because it argues against the pre-selected option using the reader's own data.
The panel, same devices, real menu
The in-place case, with the actual 160px menu and a hostile notice above. The tile ladder replaces the numbered rail here, and at this width the rail can afford 246px.
The tile ladder is the change worth looking at twice: each step's marker is the mark itself at that step's fill, so the rail reads as one object assembling downward rather than as a list of numbers. It also makes the skipped step legible without a special case — that row's mark is simply empty.
Five devices, no new concepts
| Device | Was | Now | Why it only works at scale |
|---|---|---|---|
| Ghost numeral | 11px step number | 96px, mark colour, 50% | At 500px it would be the whole screen. At 1080px it is a focal point that cannot be mistaken for advertising. |
| Schematic | a sentence | 4-node hairline diagram | Needs ~700px to lay out horizontally without wrapping into nonsense. |
| Crop marks | panel corners, 7px | sheet corners, 14px | On a small panel they read as a tic. On a sheet they read as a drawing, which is the intent. |
| Tile ladder | mono numbers in the rail | the mark per step | Needs a 246px rail. At 190px the mark and the label fight. |
| Type scale | 15px headings | 28px headings, 14px body | The single largest contributor to "not sold". 15px in a 1080px box is a caption, not a heading. |
What I deliberately did not add
1440px of room is exactly how a setup screen turns into a marketing page. Each of these was available and declined, and the reason matters more than the decision.
The schematic is hairlines, borders and mono labels. Nothing to draw, nothing to maintain, nothing to redraw per plugin — and it stays legible in greyscale and at any admin colour scheme.
The panel version would have it destroyed by the notice zone, and the sheet version does not need one — the ghost numeral already gives the page a top-left anchor.
The mark encodes progress and the label says step 3 of 5. A third indicator would be the same information a third time, which reads as anxiety rather than clarity.
Tutor LMS ends its two-step wizard on a 7,900px wall with six Get Pro buttons. The one place we mention money is the free-tier note on the Algolia card, and that is there because it is useful.
Four purple stops with four jobs, and the salmon tile still appears exactly once per screen. The temptation at this size is a fifth colour for the aside; the aside uses a hairline instead.
The inputs and both buttons are core's, including
the blue focus ring. .button-hero is core's own size modifier,
not ours.
Two, and the first is new
The right-hand aside is the most persuasive element on the sheet and it needs roughly 260px it does not always have — the admin menu collapses at 960px and users run plugins in 1280px windows. It has to degrade to a block below the cards, and at that point it stops being an aside and becomes more scrolling. Worth deciding whether it is load-bearing or a wide-screen bonus.
The tile ladder makes this more pressing, not less: now every rail row draws a six-cell mark, so a six-step flow has nowhere to go. Cap the flow at five, or accept that the tiles show proportion rather than an exact step.