How Accurate Is a Hail Map? We Checked Ours Against 785 Spotter Reports.

Two numbers, same storm, same afternoon. The hail map says 2.5 inches. The spotter report two miles down the road says 1.25 inches.

Most roofers assume one of them is wrong. Usually neither is. They're answering different questions, and knowing which question you're actually asking is worth real money when you're deciding where to send a crew on Monday.

We pulled the audit on our own radar feed on September 9 — 189,924 radar hail points, cross-checked against every Colorado spotter report of the 2026 season. Some of what came back makes our data look good. Some of it is a caveat we'd rather you hear from us than discover in a driveway.

What a radar hail map actually is

Nothing in a NEXRAD radar has ever touched a hailstone.

Radar hail products — MESH is the common one, and it's what sits under most of the swath maps in this industry — work backward. The radar measures reflectivity and how it's stacked through the vertical column of the storm, then runs that against the freezing level to estimate the largest stone that column was capable of producing. The output is gridded into cells roughly a kilometer across.

So a radar hail map is a physics estimate of potential, resolved to about a half-mile square. It is not a measurement at 412 Elm Street. Every honest description of one says "radar-estimated," and that word is doing real work.

A spotter report is the opposite trade: an actual human with an actual stone and a tape measure or a golf ball for comparison. Ground truth, but wildly incomplete — spotters report where people live, where people are awake, and where somebody bothered. A 2-inch core that lands on empty grazing land at 11 PM generates zero reports and did absolutely happen.

You need both, for different jobs.

Where radar is excellent: did hail fall here, that day

This is the part that held up.

Across the 2026 Colorado season there were 785 spotter hail reports of 1 inch or larger. For 778 of them — 99% — a radar hail signature of 0.75" or larger exists on the same day within about seven miles. Tighten the bar to a 1"-or-larger signature and it's 764, or 97%.

In the Texas metros we ingest, the same check: Austin 24 of 24, Houston 35 of 35, Dallas–Fort Worth 211 of 231 — 91%.

Read that carefully, because it's a narrow claim. It does not say the radar got the size right. It says that when a human on the ground found hail worth reporting, the radar knew something happened there that day, essentially every time. For the job of finding storms you'd otherwise never hear about — including the ones that hit at night, or over a subdivision where nobody called it in — that is the number that matters.

Where it gets slippery: the size at one address

Now the harder question. Take each spotter report, find the nearest radar cell within about three miles on the same day, and compare the two sizes. Against 726 Colorado reports:

The median ratio is 1.00. Cell for cell, the radar is not biased. Half the time it reads high, half the time low.

But the middle half of those comparisons spans a ratio of 0.75 to 1.50. That's a wide band. A 2-inch radar cell routinely sits over ground truth anywhere from 1.3 to 2.7 inches, and the tails are worse.

By size band, the pattern is textbook:

Radar nudges small hail upward and under-reads the true monsters — once you're past about 2.5 inches, the physics the algorithm relies on starts saturating. If you've ever wondered why a storm everyone remembers as brutal shows up merely orange on a swath map, that's why.

The trap: "the biggest stone near this house"

Here's the one we'd most like roofers to internalize, because it's not a flaw in anyone's data — it's arithmetic, and nearly every hail map in the industry has it.

When a map (ours included) tells you the hail size at a specific address, it usually isn't handing you one cell. It's scanning the cells around that point and reporting the maximum. That feels conservative and roofer-friendly. It isn't.

Take the maximum radar value within seven miles of each Colorado spotter report and you get an average of 2.76 inches, against a spotter average of 1.52 inches. That's about 1.25 inches of inflation — not because any individual cell lied, but because taking the max of a few dozen noisy estimates reliably surfaces the noisiest one.

We found this in our own severity scoring, we're calibrating it, and we're telling you because the alternative is a roofer walking into a driveway saying "our data shows two and a half inch hail hit your roof" about a street that got marginal 1.25-inch stones. That conversation costs you the homeowner and it costs you the adjuster.

What to actually do with this on Monday

Use the map for where and when. Use the inspection for what. Radar is your storm-finding and date-of-loss tool. It is not a damage report, and it never was.

Never quote a radar-estimated size to an adjuster. Date and location are what radar legitimately establishes. Size is established by what's on the roof and, where it exists, the official ground record. A carrier that catches you citing an inflated max-around-the-point number has a reason to discount everything else you brought.

Stop sorting your doors by stone size. This is the real one. Given a radar band that wide, stone size is a noisy signal at any single address — while roof age is a hard fact. A radar-estimated 1.75-inch core over a 19-year-old three-tab is a far better door than a 2.5-inch reading over a roof installed in 2019, and it isn't close. That's the whole argument in the 2003 vs 2018 problem, and this audit is why we weight scoring the way we do.

Cross-check the big ones against ground reports. Before you commit a week of crew time to a swath, look for the spotter reports underneath it. If a 3-inch swath has no human report anywhere near it, that's not automatically fake — it may have hit open country at night — but it does change how much you stake on it.

Why we're publishing our own caveat

Because the honest answer to "how accurate is your hail map" is accurate at one job and approximate at another, and every vendor in this space, us included, is tempted to let you believe it's precise at both.

It's also worth saying that human review is a real answer to this problem. HailTrace puts meteorologists on significant storms and publishes confirmed swaths, which is genuinely a different product from an algorithmic feed — we laid out that comparison honestly here, including where they beat us. What we're built to do is something else: take the storm footprint, cross it against roof age and insurability, and hand one roofer per city a scored list with contact information instead of a map to interpret.

The short version

Radar found 99% of the hail Colorado spotters confirmed in 2026, which means it is a very good instrument for knowing that something hit a neighborhood on a given day. Cell for cell its sizing is unbiased but loose, it under-calls the biggest stones, and the "largest stone near this address" number that maps like to show runs roughly 1.25 inches high. Treat the swath as a search, not a verdict.

See what's radar-estimated near you on the Colorado hail map or the Central Texas map, then run your own numbers on the cost-per-job calculator before you commit a season to anybody's data. If your city is still open, take a look at the map — one roofer per city gets it.