Three pages, letter size, PDF. Here is every section and why it is there.
Key facts as short bullets, not a paragraph. Every figure stated here also appears in the table below it — if the two ever disagreed, the table would win, so we do not let them disagree.
The ten closest reports, tabled: local time, type and size with the hail object name, the reporting qualifier, distance, location, county, the source's own verification tier, and the reference ID that sourced the row.
Counts for the property's county and for each bordering county, broken out by peril and never blended, with the counties named explicitly. Population density is included because a report count reflects how many people were there to report, not only how hard it hailed.
An annotated schematic — markers numbered to match the table rows, hail as circles sized by magnitude, wind as diamonds, the loss location as a gold star, storm motion with a boxed callout, distance rings, and a boxed legend. It is our own drawing, not a reproduction of an agency graphic.
The character of the event, what the data shows, and what it does not.
Every source as R-01, R-02 and so on, with the exact query parameters used. This is what makes the search reproducible by someone who does not trust us.
Confirmation that every reference was audited — all of them, never a sample — that every citation carries the exact query used, and that any finding was resolved before release.
What the report is, what it is not, and the limits of what it establishes.
| Hail radius | 10 miles from the confirmed property coordinate |
| Wind radius | 15 miles |
| Sparse data | Fewer than five reports doubles both radii to 20 and 30 miles, and the report discloses that it happened |
| Displayed | The ten closest reports overall, tabled and mapped; the rest roll into the county counts |
| Wind threshold | 58 mph and above is primary; 50–58 mph is carried in a lower-confidence tier rather than dropped |
| Tropical systems | Tagged with a named-storm and wind-deductible caveat |
| Date tolerance | Reports are timestamped in UTC, so a late-evening local event can land on the next UTC day. We search one day either side and mark any row where that applied |
Every reference cited is checked on three things: does the URL resolve, does what comes back cover the queried date and area, and are the values we attributed to that source actually present. It runs against 100% of references — the report states the coverage explicitly, so no reader has to wonder whether the unchecked ones were the weak ones.
Anything that is not a clean pass goes into a remediation loop and blocks release until it is resolved. An audit that flags problems without fixing them is theater. If a row cannot be substantiated at all, it is withdrawn and everything derived from it is recalculated.
Corrections are published, not hidden. A reissued report gets a new report ID and its own audit; the superseded version is retained and marked as superseded, never edited in place.
