Dated 2026-08-22. Written against version 0.1.0 and the 2026-08-17 pins. This
document states what this project will not do, who could be harmed by it being misread,
and what risks remain after the mitigations. It is reviewed whenever a measurement is
added; the date above moves when it is.
Unofficial. Not affiliated with, endorsed by, or approved by CAL FIRE, the California Energy Commission, or any electric utility.
It measures how much of California's published wildfire damage-inspection record set can be attributed to a published electric service territory using public data alone, and publishes what cannot be attributed along with what can. The finding is a property of two public datasets. It is descriptive geography.
- No tie-breaking of contested records. Awarding a burned structure to a named company on the strength of a rule of thumb would be publishing a guess about somebody's service obligation as if it were data.
- No damage rate per territory or per county. Where one fire supplies most of an area's records, that arithmetic describes the fire, not the territory. Publishing it next to another area's number invites a comparison the data cannot support.
- No ranking, scoring or grading of any named company or place.
- Nothing about infrastructure. No pole, conductor, substation or circuit position is read or inferred from any source. Asset locations are not public, should not be, and nothing here approximates them.
- No collection beyond the published fields needed. Addresses, parcels and assessed values are never requested; coordinates are requested only to answer which polygon a point is in and are never republished.
- No routing around access controls. A blocked request stops the acquisition.
- No silent refreshes. The figures move when a person deliberately re-measures and writes down what moved.
| Risk | Who it lands on | What limits it | What does not limit it |
|---|---|---|---|
| The headline is read as a utility safety rating | Any named utility, especially ones with high contested shares through no act of their own | The report says on every page that it ranks nobody; contested share measures boundary overlap, not performance | A reader who sees only a table row with a company name on it |
| The repair sensitivity is read as doubt about specific companies | Utilities whose polygons arrive invalid | Both repairs are named, both are run, and the disagreement size is published rather than hidden | The fact that two of the repaired polygons are the largest territories in the state |
| The county cut is read as a statement about who serves a county | Residents and county governments | The section says it locates overlaps, not service | The words "county" and "territory" sitting near each other |
| Figures are quoted years later without their retrieval date | Whoever acts on the quote | Every page carries its retrieval date; staleness triggers are defined | Screenshots, which carry none of it |
| The inclusion rule misclassifies an entity type | Entities misread into or out of the index | The rule's cost is measured; outlines holding no record provably move nothing | The 24 indexed outlines that do hold records, whose classification remains unreviewed by anybody with sector expertise |
Quote the denominator with the figure. Name the retrieval date. Do not turn a contested share into a claim about a company. That is the whole list.