Auto-schedule - Design Picker

Inline primary

A filled primary button between the counters and the view toggle.

What you are looking at

The button carries its own count, so the header sentence reads "13 scheduled, 9 open slots, auto-schedule 5 waiting" left to right without a legend.

Disabled, it drops the count and says why underneath. The reason names both excluded rows so nobody hunts for a fifth delivery that was never eligible.

Why it might win

• Impossible to miss on the first visit, which matters for a control nobody has used before.
• One control, one label, no state where the affordance is ambiguous.

The tradeoff

• A filled primary in the header competes with Apply in the rail, and Apply is the button that actually writes.
• It is loud on every month you open, including the eleven months where nothing is waiting and it is dead weight.

Quiet until waiting

A secondary button that only takes primary weight when the inbox is not empty.

What you are looking at

With five waiting it borrows the primary tint and a count pill. With nothing waiting it is a flat, disabled secondary that says nothing until you read the line beneath it.

The count is a pill rather than words, so the control keeps one width as the number changes.

Why it might win

• The header is loud exactly when there is work and silent otherwise, which is most of the year.
• Never competes with Apply, because Apply only exists once a plan is staged and by then the header button has done its job.

The tradeoff

• Quieter means less discoverable. Somebody who has never used auto-schedule may not see it on the first pass.
• Two visual weights for one control is a rule the rest of the admin panel does not use anywhere else.

Counter as affordance

Scheduled, open slots and waiting rendered as three statistics; the third one is clickable.

What you are looking at

The existing "13 scheduled, 9 open slots" text becomes a three cell statistic strip and the waiting count joins it. The waiting cell is tinted and carries an arrow, which is the whole affordance.

At zero it greys out in place. The strip never changes width, so the header does not reflow between the two states.

Why it might win

• The number and the action are the same object, so there is no way to see 5 waiting and not see how to place them.
• It replaces a plain text counter with something that answers "and what do I do about it".

The tradeoff

• A statistic that is also a button is a pattern that has to be learned; two of the three cells do nothing when clicked.
• It rewrites the header row the calendar already ships, so this is the only Module A variant that is not additive.

Stacked planner

Every option stacked full width in the 280px rail the component already ships.

What you are looking at

Spacing is three stacked buttons, each showing the dates it would produce, so the choice is made by reading outcomes rather than guessing at a label.

Start from defaults to tomorrow, 2026-08-30. That is a Sunday, so the first placement is Tue Sep 1 whatever you do; the field exists for the case where you want to skip a week.

Every plan row carries its own section select, defaulted to resources because all five deliveries target resources slugs.

Why it might win

• Fits 280px exactly, so the shipped grid split at component line 124 does not change.
• The spacing hints mean the dates are visible before you commit to a strategy, not after.
• One scroll gets you from strategy to the five dates to Stage, with no expanding or hiding.

The tradeoff

• Tall. With five rows and their section selects the rail runs past the fold on a 900px viewport, so Stage needs a scroll.
• The dates in the spacing hints have to be recomputed as you change Start from, or they go stale and lie.

Segmented compact

Options as two segmented rows, plan as a one line per placement list.

What you are looking at

Spacing and Order become segmented controls, which costs two lines instead of eight and puts all five dates above the fold.

The plan list shows ordinal, date and title. A row expands in place when you need the per row section select, shown here on row 3.

Why it might win

• The entire panel, options and plan and actions, fits without scrolling.
• Reads like a schedule rather than a form, which is what it is.

The tradeoff

• Three segments in 280px means the labels have to shorten to Next open, One a week, One a fortnight, and the dates each produces are no longer visible.
• The per row section select is one click away rather than present, so the default of resources is more likely to go unexamined.
• Truncated titles: "Radiation dose monitoring: a practical program guide" does not fit one line at 280px.

Wide planner drawer

The rail widens from 280px to 340px while a plan is being built.

What you are looking at

Same options as B1, but each plan row also shows words, images, arrival date and the warning count from the last gate run. Radiation dose monitoring is 4043 words against a 2000 to 3900 band, and that is the kind of thing you want to see before you put it on a Tuesday.

The second state changes two things at once, spacing to one a week and order to newest first, so Double reading radiology, which arrived this morning, leads.

Why it might win

• The only variant where the decision to place and the decision to hold back can be made from the same screen.
• At 340px the titles fit on two lines without truncation.

The tradeoff

• It needs 340px, not the 280px the component ships. The grid loses 60px, which at a 1280px viewport is roughly nine pixels off each day column.
• A rail that changes width when you press a button is a layout shift, and the month reflows under you at the moment you are reading it.
• More numbers per row is more to read for a task that is mostly about dates.

Dashed slot, filled in

A planned card keeps the dashed edge of the open slot it landed on.

What you are looking at

The dashed border says the same thing it said when the day was empty: this is not on disk. It is now tinted and holds a title, so the day reads as spoken for without reading as scheduled.

Sep 17 carries the amber card, which is the existing pending move treatment, and Sep 11 shows the vacated marker it left behind. Blue dashed is planned, amber solid is a move you made.

Every card has its own undo, and every card is draggable: dragging a planned card to another open day rewrites that one placement and leaves the other four alone.

Why it might win

• Continuous with what the day looked like a second earlier, so nothing jumps.
• The dashed edge is already the codebase vocabulary for "not real yet".

The tradeoff

• A dashed border also means "open slot", so at a glance a planned day and an empty day share an outline.
• Nothing on the card says where it sits in the plan, so reordering has no visible handle.

Numbered placement card

Solid primary edge, an ordinal badge and a grip.

What you are looking at

The badge is the one fact a planned card carries that a pending move never does: its position in a plan of five. Drag a card and the numbers renumber.

Solid blue against amber is unambiguous at a glance across a full month, which is the state that actually has to be readable.

The grip in the top right says draggable without a hover, which matters because these cards appear five at a time.

Why it might win

• Distinct from both the open slot and the amber move at any zoom.
• Makes reorder a visible, direct manipulation rather than a hidden option in the rail.

The tradeoff

• A solid border reads as committed, and nothing here is committed until Apply. The tint has to carry all of the "not yet" meaning.
• The badge and the grip cost a line of vertical space in a 112px cell that already holds a title and a section chip.

Accent bar

No fill, a 3px primary bar down the left edge and the position in the footer.

What you are looking at

The lightest of the three. A planned card looks like a normal scheduled card wearing a stripe, so a month with five placements still reads as a month rather than a plan.

Position and undo sit in the footer row next to the section chip, so the card keeps the same height as a booked one.

Why it might win

• Least visual noise across a full month, and the booked articles stay readable.
• The stripe scales: ten placements do not turn the grid into a blue wall.

The tradeoff

• It is the least distinguishable from a scheduled card, and the whole point is that these are not scheduled yet.
• A 3px stripe is easy to miss on a laptop screen at 1280px, where each day column is about 150px wide.
• Depends on the legend more than the other two do.