The listing is a deal to judge. You scroll a single scannable report — verdict → fair-value band → the comps → fit → area — then act: Pass / Ask / Shortlist. No chat front door; "Ask the agent" is scoped to this verdict.
Leads with the moat: a comps-derived verdict and fair-value band the portals can't compute. Every number is cited (the 6 comps each link to the agency-direct source). The agent writes the verdict and explains it; it never invents the figure.
Blue-edged blocks = computed, auditable facts. The purple panel = the agent's read (interpretation), visually softer. A buyer always knows which is which.
This is the conservative / shippable take — a familiar report layout, low-risk, high-trust. Variant C turns the same verdict object into a swipe deck.
Layout, interactions & the shape of the data are designed. Numbers are realistic Eastern-Suburbs placeholders; images via picsum.photos. In production every cited figure is live D1 data (sold comps, SEIFA, catchments, GTFS transit).