The 5-KPI journey

This is what BOSS actually checks before it will recommend acquiring a property. Each step below is a real KPI in decision.json, with the tool or data source that feeds it and its honest status — not every step resolves cleanly, and that's shown, not hidden.

Loading subject property…

When will this flat be rented out?

Reads price-vs-market positioning and nearby amenity density to give a directional letting-speed band. No live "days-to-let" data source exists for this micromarket yet, so this is a heuristic, not a measurement.

↓

Out-of-pocket cost to make it a Flent home

A photo-reading sub-agent (see agents/renovation-capex-brief.md) estimates furnishing + repair capex per room. It's gated on image evidence by design — it refuses to guess from the text description alone.

↓

Locality — what's nearby

Nearby hospitals, supermarkets, bars/pubs, schools, and restaurants, geocoded and radius-searched against OpenStreetMap (Nominatim + Overpass API) — a free, keyless substitute for the Google Maps Places API, same query shape a paid integration would use in production.

↓

Sqft vs. rooms vs. market price

Two reads combined: how the unit's size suits different tenant profiles, and how its asking rent compares to a market price built from the median of real scraped comparable listings. Broker-call price discovery (calling local brokers for a second, independent price signal) is designed but not built yet.

↓

Rent + deposit + Flent overhead → breakeven

Capital employed (furnishing capex + net deposit gap) divided by monthly contribution (tenant rent minus landlord rent and maintenance) gives months-to-breakeven. Capital employed now has a real bound; payback itself still needs two Flent-internal inputs.

Tools we used

Text and comp data came from Claude Code's built-in WebSearch/WebFetch tools; real gallery photos (JS-rendered, not visible to a plain fetch) came from a Firecrawl headless-browser scrape; locality data came from direct calls to OpenStreetMap's free, keyless APIs — see data/sources.md for the full lineage.

See the full analysis →