● SETTLD · FAIRLY · FULL FLOW · v3

Fairly — the whole flow

The elevated language across every screen: a hero element per screen, the receipt as an actual receipt, avatars as the visual system, colour for direction, tactile cards over form controls. Segmentation mode is still the spine that drives the flow.

Host-allocate lane
Default path · the payer reads the bill and tags the table
Self-claim lane
Big tables · the payer broadcasts the bill and each guest taps what they ate
Alternate view — beyond the list
Same split, not a list · two spatial / data-viz takes on the payoff, in answer to "is a list the best we can do?"
Design alternatives — three interaction models
The same Fairly flow re-cored three ways. Not more screens for one design — three competing answers to "how should this feel," built to compare and pick a direction to prototype.
A

Receipt Canvas

Direct manipulation on a single surface. No steps, no "next" — the scanned bill is the screen. Tap a line, then tap who had it; shared plates auto-divide. The split resolves itself as lines get claimed. Loading = the receipt skeleton (honest, it's literally parsing); empty = the same canvas with "add a line yourself"; error = inline reconciliation.

B

Guided Sheets

One decision at a time. A persistent live-split summary is always the base layer; each question rises as a bottom sheet over it, and your answer collapses back into the summary as a settled fact. Lowest cognitive load of the three — controls always thumb-native, validation stays local to the active sheet, the summary skeletons cleanly while parsing.

C

Live Board

Invert it. The host doesn't allocate anyone — they scan, broadcast a link, and watch a shared real-time board fill as each guest claims their own lines. This is the persistence + settlement moat made literal: a live shared state can't be a screenshot or a "just ask Gemini." Distributed cognition (no one carries the whole bill), and it owns states the others can't have — live fill, claim conflicts, async settle.