Report v1
The Supplement Recovery Benchmark
This is one supplement desk's own closed-file data. It is not an industry survey, not a market study, and not a projection.
By Kyle Hamrick, Founder. The Estimate Company.
What this is, and what it is not
It describes the files this desk has written. A desk serving different contractors, in different markets, with a different mix of carriers, would see different numbers. Read it as a reference point from a real book of business, not as a benchmark of the trade.
Everything below comes from a single export of this desk's claim system, run 2026-08-21. 210 claims total. 88 carry a recorded recovery figure. Every table names both counts, because the difference between them is the most important methodological fact in this document.
Method, stated before any number
Dual-n. A claim only counts toward a median when it carries a recorded recovery figure. So every table shows a claim count and a measured count, and they are never the same. This matters more than it sounds: a carrier can look well represented on claim count and be statistically empty on measured recoveries. Allstate is the worked example below.
Suppression. Any group with fewer than 5 measured recoveries is suppressed outright (k-anonymity floor K=5). Publication requires at least 20 measured recoveries.
Segment boundary. Two populations appear in this report and they are not interchangeable:
| Population | What it is | n |
|---|---|---|
| Full-replacement-approved segment | Claims where the roof was already approved for full replacement | 65 |
| All measured recoveries | Every claim carrying a recovery figure, including partials and overturns | 88 |
The all-measured population runs higher than the segment population, because claims that start denied or repair-only recover more when they turn. Any figure below states which population it belongs to, in the same sentence. Mixing them is the single easiest way to make this data lie.
The typical/overturn split rides a ratio proxy labeled APPROXIMATE. The desk's system does not yet carry a claim-type field at intake, so the segmentation is inferred rather than recorded. A claim-type field is the fix and it is on the roadmap; when it lands, the proxy retires and this section changes. Stated here rather than in a footnote because it is a real limit on the next section.
No personal data. The export reads an allowlist of 10 fields and never reads the 17 that carry identifying information. Output is scanned against the source database before it is written.
The typical file
On claims where the roof is already approved for full replacement, our own closed files show a median recovery of about $4,000, with half landing between roughly $1,800 and $9,500.
n = 65, all measured recoveries in that segment.
The qualifier is part of the number. A median recovery figure without the already-approved-for-full-replacement condition attached describes a different population and overstates what a typical file returns.
Distribution, across all measured recoveries
This table covers the all-measured population (n=88), not the segment above. It is shown because the spread is the story the median cannot tell.
| Recovery band | Files |
|---|---|
| Under $1,000 | 8 |
| $1,000 to $3,000 | 25 |
| $3,000 to $6,000 | 17 |
| $6,000 to $10,000 | 10 |
| Over $10,000 | 28 |
Median across this population: $4,740 on n=88.
Read the two ends together. 33 of 88 files landed under $3,000 and 28 landed over $10,000. A contractor who prices supplement work off the median alone will misjudge it in both directions: the floor is lower than the median suggests, and the ceiling is much higher.
Carrier behavior: one row, and the reason there is only one
| Carrier | Claims | Measured recoveries | Median recovery |
|---|---|---|---|
| State Farm | 55 | 27 | $4,698 |
What is not in this table, and why. Allstate appears in 28 claims and USAA in 22, which both sound publishable. Neither is. Allstate carries 12 measured recoveries and USAA carries 12, both below the 20-measured publication floor. Nationwide, Travelers, and Erie are suppressed entirely under K=5.
A six-carrier table built from those cells would look more authoritative and would be worth less. Naming the exclusion is the point. Suppression here is a floor on what this desk will publish, not a statement about any carrier's conduct.
Turnaround
Held. The current export supports a median of 1 day from open to submitted on a sample of 9 files. Nine files is not a benchmark. Next-day turnaround is this desk's standing commitment and is stated as a commitment elsewhere; it is not published as a measured statistic on this sample.
Fee structure: permanently excluded
This report does not publish a distribution of realized fee percentages, and no future version will. The fee schedule is a published position, not a statistic to be derived from closed files.
What this does not measure
- One desk, one book of business. Geographic and carrier concentration are real and unmeasured.
- Survivor bias in what arrives. Files that reach a supplement desk are already files somebody thought were short. That is a selected population by construction.
- The ratio proxy behind the segment split, described in the method section.
- Recovery is not profit. These are recovery figures, not margin, and they carry no adjustment for the contractor's own cost of pursuing the file.
- No time series. This is a snapshot, not a trend. Nothing here supports a claim about direction.
Figures from a claim-system export dated 2026-08-21. Segment figures are the desk's ratified publication set. Distribution and carrier figures are from that export, under dual-n reporting, K=5 suppression, and a 20-measured publication floor.
Want your own file read against these numbers?
Send the carrier estimate, your photos and measurements, and the declarations page. We read it line by line against the roof and hand back what the scope actually supports.