Concept B · order-first

2 paid prescriptions cannot be ordered Rx #7 has no catalog item pinned · Rx #482 has two that disagree. Oldest waiting 6 days. No repair screen exists yet

Fulfillment

Seven orders, newest first.

New order

All orders

Newest first

Order	Preparations	Status	Created	Sent	Fills
Order #14	4 preparations · Semaglutide / B12, NAD+, BPC-157, PT-141	Draft	4 days ago · Casey Review	Not sent	9
Order #13	7 preparations · Tirzepatide / B6, Sermorelin Acetate, Glutathione and 4 more	Sent	7 days ago · Avery Review	6 days ago · Avery Review	22
Order #12	3 preparations · Semaglutide / B12, NAD+, BPC-157	Partially received	9 days ago · Casey Review	8 days ago · Avery Review	12
Order #11	5 preparations · Semaglutide / B12, Tirzepatide / B6, NAD+ and 2 more	Received	17 days ago · Casey Review	16 days ago · Casey Review	18
Order #10	4 preparations · BPC-157, Sermorelin Acetate, GHK-Cu, Ipamorelin	Received	24 days ago · Avery Review	23 days ago · Avery Review	14
Order #9	6 preparations · Semaglutide / B12, Retatrutide, Glutathione and 3 more	Received	32 days ago · Casey Review	31 days ago · Avery Review	20
Order #8	3 preparations · Tesamorelin, Methylcobalamin, PT-141	Received	39 days ago · Avery Review	38 days ago · Avery Review	11

What this page assumes

This is the same page as the desktop index, at 390 px. Same H1, same find box, same attention strip, same seven tabs — To order, Drafts, Sent, Labels, To ship, Stock, All orders — same seven rows, same order. Nothing was removed, reworded or reordered for the phone — the only difference is the width, so the comparison measures the architecture rather than a bespoke mobile build.
The find box sits in the header rather than in a card of its own, and goes full width below 560 px. It resolves any identifier — Rx number, patient, order, item, lot — to its object, which is the single gap that blocked every concept in scenario S3.
Concept B is the worst mobile case of the three, by construction. Its primary object is a six-column row. The measured build shows the cost exactly: on an order page at 390 px the visible table container is 316 px against a 720 px table, so an operator sees 44% of it and the quantity — the number they opened the page for — is off screen (finding F4). A queue of short list rows and a ranked list of stalls do not have this problem; a ledger does.
The shared card collapse is the fix, and it is in the shared stylesheet on purpose. Below 560 px every row becomes a labelled card with the column name printed beside each value, so nothing sits past a horizontal scroll. All three concepts inherit it equally, which is what keeps the mobile comparison about structure rather than about who wrote better CSS.
The fix converts a width problem into a height problem. Six columns become six stacked lines, so seven orders are seven tall cards. Scanning the ledger on a phone is now scrolling, not swiping — better, but still the longest of the three home screens.
The seven tabs replace seven full-width saved-view buttons. To order, Drafts, Sent, Labels, To ship, Stock and All orders wrap to two rows at 390 px, and the active All orders tab stays fully visible. The wrap costs two lines of height and hides nothing.
The 96 waiting fills are the first tab now, not a banner. In the pure order-first version they were a full phone screen of banner that opened nothing before a single order came into view — the concept's characteristic failure, louder at 390 px than on a desktop. As the leftmost tab they cost no vertical space, sit above the fold, and open a real surface. What the tab still does not do is say how long they have waited: 96 is a count, not a queue.
The exceptions strip renders at the top of the page content, directly under the tab row, on the order-ledger surfaces. The pile surfaces — Labels, To ship and Stock — do not carry it, so a stall is only visible from the ledger surfaces. On the untabbed version the two blocked prescriptions were a card at the very bottom, below seven tall order cards — reachable only by an operator who had already scrolled past everything they came for. A stall you have to scroll to find is a stall nobody finds.
Scenario S10 in practice. An operator at the bench asking "what did we order?" gets there in one click and two screens, and then reads a card per order. The answer is available; it is simply further down the page than in either other concept.
The tabs make the work reachable; they do not rank it. Order #14 has been an untransmitted draft for 4 days and, as a card on a phone, looks exactly like a healthy one — you learn it is stalled by reading the Created line most of the way down the card. The Drafts tab is one tap away and its count reads 1, but this ledger sorts by recency rather than staleness, so nothing here escalates on its own. That is the residual cost of order-first, and tabs do not pay it off.
This page is a prototype device only. It is the same page as the desktop index with a forced 390 px viewport, and it must not become a route in the built console.