Hail screening usually fails in the same way. Someone pulls a regional storm summary, finds that a metro took several significant hail days in the last five years, and concludes that the whole territory is a target. That is true and useless. It puts every property at the same priority, which is the same as having no priority.
A portfolio-level answer needs the event attached to the asset. Two things have to be true of a property before it belongs near the top of the list: a documented event passed over that parcel, and the asset on that parcel was in a state where the event would matter.
Step one: get the event history over the parcel
NOAA storm event records are public and carry the fields you need. Each hail report has a date, a location, and a reported stone size. Wind reports carry a date, a path and a reported or estimated speed. The records are published monthly with a reporting lag of several weeks.
The work is in the join. A storm report is a point or a path, not a footprint, and the reported location is where an observer was, not the edge of the hail swath. Screening at parcel level means deciding on a spatial rule and stating it, then applying that rule identically to every property.
- Record the event date, the reported stone size or wind speed, and the distance from the report to the parcel, per property.
- Set one buffer distance for the whole run and write it down, so that a property's inclusion can be explained rather than argued.
- Keep every qualifying event, not just the largest. A roof that took three separate one-inch days is a different story from one that took a single one-inch day.
- Treat a report location as approximate. Reported hail is observed and phoned in, so both the coordinates and the size are estimates.
- Note when the nearest report is far enough away that the property is only in the list because the buffer put it there.
Step two: describe the asset
An event only matters against a specific roof. The assessor roll usually carries roof cover and year built, and in many counties an effective year or a last improvement date as well. Imagery carries what the roof looks like now.
Roof cover is the first sort. Different assemblies respond to impact differently, and the trade literature on each of them is better evidence than anything you can infer from above. What matters for screening is that the cover type is recorded on the row, so the priority order can be explained.
- Roof cover as coded by the assessor, with the code scheme named, since counties do not use the same one.
- Roof age, computed against the event date rather than against today, because a roof installed after the storm was not exposed to it.
- Any reroof, roof repair, or roof recover permit filed after the event date, with its issue date and status.
- Visible post-event evidence in current imagery: fresh patching, mismatched sections, replaced panels on a residential run, staged material, or a partially completed tear-off.
- Rooftop equipment and array presence, since a hail event over a parcel with an array raises a service question the same event over a bare roof does not.
- Imagery capture date relative to the event date, which decides whether the imagery can say anything about the aftermath at all.
That last item is the one most screens get wrong. If the most recent capture predates the storm, the imagery tells you the pre-event condition and nothing else. It should be recorded as pre-event, not read as if it showed the roof today.
Step three: order the list
The ordering logic is a story about the gap between an event and a response. A roof that took a documented event and shows no permit and no visible work afterwards sits in a different band from one that was torn off the following month.
- Documented event, roof older than the event, no post-event roof permit, no visible work. This is the band worth opening first.
- Documented event, no post-event permit, but imagery predates the storm. Same story, weaker evidence, and the open check is imagery or a drive-by.
- Documented event, post-event roof permit on file. Likely already addressed. Worth noting, not worth leading with.
- Documented event, roof installed after the event date. Screen it out and say why.
- Property only inside the buffer, with no nearby report and no supporting evidence. Keep it visible at the bottom rather than deleting it, so the reader can see the edge of the rule.
Note that this is a research order, not a damage estimate. Band one means open this one first. It does not mean this roof is damaged.
The same ordering works across a mixed portfolio. A single storm track runs over the warehouse district and the subdivision behind it, and the logic does not change: event over the parcel, asset state at the time, response afterwards. Only the record quality changes, since residential roof cover and permit detail are often thinner in the assessor file than commercial improvement records.
What the screen cannot establish
It cannot establish that hail struck a specific roof, what size stone actually fell there, whether the membrane or the shingles were bruised or fractured, whether granule loss or a soft spot is present, or whether any damage that exists is functional rather than cosmetic.
It cannot establish the insurance position either. Whether a policy was in force, whether a claim was filed, whether a deadline has passed, whether the loss is covered, and what a deductible does to the economics are all outside every source used here.
It also cannot establish that a permit-free roof was never repaired. Minor work often goes unpermitted, and plenty of jurisdictions publish no searchable index at all. A blank permit column is a missing record, not a finding.
Turn each band into a next action
The screen has done its job when every row carries a specific next step rather than a score. Band one gets an inspection request or an owner conversation. The imagery-predates-storm band gets a fresher capture or a field look before anyone calls.
Screened this way, a storm season becomes a short ordered list with its reasoning printed next to it. Your team can argue with the buffer, challenge a band, and have the run repeated with a different rule, which is a great deal more useful than a map of a metro that took hail.