Short-stay market

Each current rental is compared with up to 15 Airbnbs of the same bedroom count within 1.5 km (3 km if fewer than 5 are found), nearest weighted most. Their forward occupancy (next 45 nights) and Mon–Thu / Fri / Sat / Sun rates give a guest-paid income per week; the weighted median comp is taken and cut by 15% for a new listing with no reviews. The host keeps 85%, then rent, cleaning per turnover, fixed weekly costs and 3% maintenance come off to leave the margin. The Cleaning switch above re-costs every rental on the spot — self-clean charges only laundry, linen and consumables per turnover; paid cleaner charges a cleaner's fee plus an amount per bedroom — and the scores, verdicts and ranks on every tab follow it.

Strong 75+, Good 60–74, Marginal 45–59, Weak under 45. Confidence is separate from the score: how many comps, how close, how much they agree, and how many of their rates were real quotes rather than estimated. A Blocked rental has a landlord or strata “no” recorded in rental_flag. Tunable numbers live in score_config.

Every calendar scrape keeps a full availability snapshot per listing. When a night that was open in the previous snapshot is closed in the next, that run of nights is a booking; the reverse is a release (a cancellation, or the host opening dates up). The price is the last asking price we saw for those nights — a quote from the nightly price sweep when one covered them, otherwise the listing's Mon–Thu / Fri / Sat / Sun rate (dow, marked ~). Nothing here is confirmed revenue: a closed night can be a host block or a sync from another platform, so kind is a guess — gap for a stretch shorter than the minimum stay or a same-day single night, block for more than 21 nights, guest for the rest. Lead time is how far ahead the stay was when it was booked; with the daily feed running that is accurate to a day.

Stays that cover the chosen night, week or month — keyed on the dates guests are staying, not the day we spotted the booking. A stay straddling the edge of the range is counted only for the nights inside it, so a 6-night stay starting 30 Sept contributes 1 night to September and 5 to October. Guest bookings only; gaps and host blocks are excluded.

This can only show bookings made since detection started on 11 Sept 2026. A night that was already booked before then is invisible here — it never flipped from available to unavailable while we were watching. So early months look far emptier than they really are, and the figures are a floor, not a total. The Listings tab's occupancy percentages are the better guide to how full a month actually is.

Every rentals scrape writes one row per rental per day into rental_price. The primary key is (rental, date), so re-seeing the same rent costs nothing and only a genuine change adds to the story. Days listed is the span from the first time we saw the property to the most recent — it starts from when we first saw it, not when the agent listed it, so early figures understate the true time on market.

Before 12 Sep 2026 rent was a single value that each scrape overwrote, so there is no history further back than that. The one earlier drop that survived (10/20-22 Woomba Pl, $850 → $790) did so only because an address-format change had accidentally split it into two rows.