Unified QR: one code,
every bank.

One entry point, branching after type selection. Each type sees only the fields its legal status requires; the flow adapts instead of the merchant.

A single universal form covering all three types — “simpler to build”, one spec, one screen.

In prototype tests the universal form stalled self-employed users on statutory-document fields they don’t legally have. The branched version passed without a single clarifying question.

Hide multi-bank routing behind one QR. The receiving bank is chosen once, in settings — never at the moment of payment. The counter shows one code, always.

Exposing bank choice per transaction — technically “more transparent”, and the banks initially wanted their brand visible at pay time.

Interviews at the counter: merchants refuse to “think about banks” with a queue in front of them. Per-payment choice produced hesitation, mis-taps, and a slower line — the exact failure mode for retail.

Split verification into steps with checkpoints between them, and open a limited mode early: merchants accept payments under a capped limit while full checks run in the background.

A full KYB gate before any access — everything up front in one heavy form. Safest on paper, and the default the compliance team started from.

When compliance added three fields two weeks before launch, I modelled the funnel in three days: a projected −40% completion. The phased checkpoint plan was approved in a single meeting — argued with data, not taste.

Default to a single location; “add a location” lives in the dashboard, with a QR generated per location. Onboarding never asks about it.

Asking “how many locations?” during onboarding for everyone — the classic “collect it while we have them” instinct.

20+ discovery interviews: the overwhelming majority of SMBs start with one point of sale. The question up front was pure cognitive tax on the 90% to serve the 10% — who found “add location” later without help.

Promote confirmation to a full-screen receipt — amount, source bank, settlement time. Processing is an explicit named state, never an anonymous spinner.

A toast notification over the transaction list — the standard, “lightweight” pattern every wallet uses.

The discovery insight (below) reframed this: merchants re-checked history repeatedly to confirm money arrived. The full-screen receipt with settlement time removed the re-checking loop in follow-up tests.

Every decline carries a reason and a next action — retry the QR or switch method — phrased for the merchant, who is standing in front of a live customer.

A bare “Payment declined” status. Honest, minimal — and useless at the counter.

In usability sessions, declines without a next step ended in “I’d call support”. With reason + action, participants resolved the situation at the counter without leaving the screen.