Methodology
Useful stock data starts with knowing when to say “we don’t know.”
AC Radar reports availability published by stores. It does not estimate inventory, infer a nearby branch, or turn a failed check into a reassuring answer.
01 · Collection
Small, polite and store-specific
We monitor selected portable air-conditioner listings at stores. Collection uses an isolated adapter, schedule and request limits. We use store-specific stock responses and only attribute a reading when the source supplies an identifiable branch.
A blocked request, unknown store, wrong product identity or unrecognised response becomes no reading. It never becomes an invented stock status.
02 · Confirmation
One flicker is not a change
The first valid observation establishes a baseline. Later changes must appear on two distinct consecutive collection cycles before they are confirmed. A no-reading cycle breaks a pending change.
Prices are shown only when they fall inside the curated product’s plausible price band. Missing or implausible prices are omitted while a valid stock status may remain.
03 · Public delay
Confirmed first, published later
Confirmed changes receive a publication time. The current public delay is configured to 30 minutes. The map and list query only the already-published history; a newer private transition cannot leak through the current-status table.
It means there are no reliable, published in-stock or low-stock readings in that radius. It does not prove every branch is out of stock.