Sumary finds every anomaly. One decision clears twenty.
Sumary already finds them: the rule fires, the correction is proposed. But Anomalies is still a list — sorted by date, one row at a time, no amounts. The finding scales with the ledger. The clearing scales with her.
I’m Carl, I design fintech products. I rebuilt Anomalies on your own nav tree and status names — no brief, no access. Then drawing it turned up the case a bulk button lives or dies on: the ledger writes 17 of 20, and the queue has to say so. It’s live below.
The concept, running
It’s live — pick a cause, expand it, clear it.
From finding the work to finishing it.
Same nav tree, same statuses, same words — nothing here is a new information architecture. What changed is what a row means: dates group into causes, each cause carries the money behind it, and the verb sits where the evidence is. An anomaly isn’t finished when it’s found. It’s finished when it’s posted.
Four decisions, and why:
Grouped by cause, not by date
Anomalies arrive date-sorted — the one order in which no two rows share a decision. Grouped by cause, 47 entries become 6 decisions, each priced by what it puts at risk.
2 entries dated before the locked period stay open.
The decision lives where the evidence is
The money at stake, the field it changes and the verb, in one panel. Two of the 22 sit in a period that is already locked, so the button says 20 before she presses it, not after.
One decision becomes a rule
Clearing a whole cause at once is a person saying something general. That’s the moment a rule is worth writing, so it is written then — and only then. Three cleared causes, 27 entries that don’t come back.
A bulk that half-fails says so
20 go to the ledger; the ledger writes 17. The 3 it refused don’t leave the queue — they stay, with the reason on the row, and the rule waits until they land.
Any tool can find the anomaly. Only a queue lets her finish it, and says so when it can’t.
A bit about me.
I’m Carl. I run Sunday, a product-design studio for fintech. Before this, Swedbank. I work embedded, like part of the team, from first research to the final interface.
The clients and vouchers in the demo are invented, modelled on a Danish ledger. Wired to real ones instead, this gets a lot sharper.
“He champions user-centered design without ever losing sight of how it drives real business outcomes. That balance is rare.”
That’s the idea.
Your product already won the hard argument — the anomalies find themselves. The next one is won by finishing. Grab 30 minutes and I’ll walk you through where I’d take it, or a two-week sprint makes the clearing flow shippable in your app.