How we count.
Every rule. In the open.
A number you can't check is an opinion wearing a suit. This page exists so you can check ours.
An event we retrieved from a named public source and stored — one row per event. Pull the same feeds for the same window and you get the same numbers.
An estimate — our math on top of the counts. We publish none yet. The first will be the Demand Index, after 60–90 days of calibration, labeled and kept separate.
Official public warnings and advisories for the six counties we watch: Los Angeles, Orange, Riverside, San Bernardino, San Diego, Ventura.
The public earthquake feed, magnitude 2.5 and above, inside the same region.
Source health is monitored. If a feed goes quiet, the gap stays a gap — we never backfill missing hours into existence.
It pressures four territories, so it counts once in each. Every count states its window.
Retrieved twice, stored once — deduplicated on the source's own identifier. “Active” means the source's own expiry hasn't passed.
Can't be honestly pinned to one county (an offshore quake)? Off the map — never out of the totals. We don't guess events into a territory.
Rules a working technician would agree with out loud. An event that matches no rule is stored and shown with no trade tag — never forced into one.
No private or scraped personal data — named public feeds only.
No pay-to-play — no one can buy a better-looking county.
No backdating, no silent edits — history is append-only, corrections are open.
No modeled figure dressed as a count — ever.
No fake “live” — nothing animates a freshness that doesn’t exist.
Every counted aggregate on the dashboard is also published as machine-readable JSON — for contractors, reporters, and researchers. Same numbers, same labels: the counted block is recountable, the modeled block says exactly why it's empty. No API keys, no tracking, refreshed about every 5 minutes.
An industry standard is a number everyone can pull, check, and cite. This endpoint is that promise in production — years before anyone asks for it.