Inline panel

The panel opens inside the card that prompted it.

Why this shape

The evidence stays on screen while you decide. The message, the phone and the volume are still right there above the choice, which is what tells you this is a bot rather than a buyer.

The narrowest option that catches more than one lead is preselected, so the common case is one click.

Looks like

• Radio list, one row per field
• Live match count on every row
• Over the ceiling turns red and says so

Trade-offs

• Card grows tall, pushing other leads down
• Long message text can crowd the value column

Blast-radius table

The same choice, read as a comparison rather than a list.

Why this shape

A bar makes the difference between 6 and 41 immediate rather than something you read and convert. The gmail.com row is visibly the wrong answer before you have finished reading it.

Columns line up, so long values and long messages do not distort the layout the way they can in a sentence.

Looks like

• Table, one row per field
• Proportional bar, green to red
• Panel sits below the card, not inside it

Trade-offs

• Heavier, more chrome for one decision
• Bar is proportional to your biggest option, which can flatter a mid-sized one

Two steps, named leads

Choose, then see the actual leads before confirming.

Why this shape

A count tells you how many. This tells you which. If a real hospital were about to be swept up by a careless rule, this is the only variant where you would read its name before pressing the button.

That matters most for the rules you will be tempted by later, on message text or domain, where the match is fuzzier than an exact organization.

Looks like

• Numbered two-step inside the card
• Named preview list with reference codes
• Destructive button only on step two

Trade-offs

• Two clicks instead of one, every time
• Preview list needs its own scroll past ~10 leads

Same cards, greyed

A blocked lead looks like the lead it is, only quieter.

Why this shape

Nothing new to learn. The message, the volume and the address are all still visible, which is exactly what you need to judge whether a block was wrong.

Search and a date filter are here because the list only grows, and a false positive that has fallen off the first page might as well not exist.

Looks like

• Live card at reduced contrast
• Rule, time and author on every card
• Restore says what it will and will not send

Trade-offs

• Tall. Six spam leads fill the screen
• Repeats the rule line on every card

Dense table

Built for the day there are two hundred of these.

Why this shape

Six rows fit without scrolling and the repeated organization becomes obvious as a column. Filtering by rule answers the question that actually matters when something looks wrong: what else did rule 3 take?

Bulk restore is the recovery path for a rule that turned out to be too broad, and it is only practical in a table.

Looks like

• Checkbox column with a bulk action bar
• Filter by rule and by date
• Message truncated to one line

Trade-offs

• Truncated message hides the thing that proves it is spam
• Reads as data, less like a person who wrote to you

Grouped by rule

Organised by the decision, not by the date.

Why this shape

The question you will actually arrive with is "what did this rule take", and this answers it without a filter. One rule, one group, one Undo this rule that restores the whole group at once.

Rows archived by hand, with no rule, get their own group rather than looking like they came from somewhere.

Looks like

• One panel per rule with a header count
• Group-level undo next to per-row restore
• Compact single-line rows

Trade-offs

• Chronology is lost, which is how you would spot a new wave
• Many small rules means many small groups

Table with hit bars

Everything on one row, sorted by what is working.

Why this shape

Sorted by hits, so the rule doing the work is first and the dead one is last and visibly greyed. A rule with zero hits after a month says "no hits yet" in plain words rather than leaving you to interpret a zero.

Looks like

• Hit count as the visual anchor
• Inactive rules dimmed in place, not hidden
• Turn off before delete, in that order

Trade-offs

• Three buttons per row is a lot of chrome
• No sense of what the rule actually caught

Rule cards

Hits as the hero number, one card per decision.

Why this shape

Matches the leads list you already read, so the whole page is one visual language. The hit count is large enough to scan down the page without reading anything else.

Room for the full action set in words, including Undo this rule, which is hard to fit in a table row and is the action you want most when something has gone wrong.

Looks like

• Same card shell as a lead
• Big count, right aligned
• Off rules dimmed with an explicit chip

Trade-offs

• Only three rules fit on a screen
• Harder to compare hits across many rules

Table with evidence

Every rule carries the last thing it actually caught.

Why this shape

A hit count tells you a rule is firing. The most recent catch tells you whether it is firing at the right thing, which is the question that matters and the one no count can answer.

The dead rule uses that column to explain itself: it caught a real hospital once, that lead was restored, and it was turned off. That is the story you would otherwise have to reconstruct.

Looks like

• Four columns, evidence given the widest
• Two visible actions, rest behind More
• Off rules keep their explanation

Trade-offs

• Widest layout of the three
• More hides Apply and Undo one click deep