Property claim estimate disputes: 146 line items priced lower, only 28 missing

When a property estimate comes back far under the policyholder's number, the usual explanation is that somebody did not look hard enough: a hurried scope walk, a subfloor nobody opened, a trade nobody called. The line-item data says otherwise. In the 21 large disputed claims where Finch Claims compared itemised estimates from both sides, 146 line items were priced lower by the insurer and only 28 were missing from its estimate: roughly five pricing differences for every omission. Finch's own methodology note says it plainly: "most of the gap sat in the price of repairs already on the list, not in what was omitted."

That changes what kind of argument an estimate dispute is. If the fight were about missing scope, it would be a question about who looked carefully. Because it is about the price attached to line items both sides already listed, it is a documentable question: what did this unit of work cost in this market on this date, and what evidence says so. The file that survives appraisal is the one carrying dated, local, line-item pricing evidence, not the one with the more sympathetic adjuster.

What follows is the mechanism that fixes the number, what the much-quoted 43% figure does and does not say, and where inside the estimate the money actually sits. Then the disclosure duties two large property states already impose, what Florida's examiners penalised when they looked, and the documentation standard that ends the argument before a panel starts it. For how the same money gets counted at portfolio level, see the companion post on how claims leakage is measured by line of business. The dispute itself sits at stages 5 and 6 of the claims process from first notice to settlement.

The distinction decides which proceeding you are in. Whether the loss is covered is a question about the policy and the facts, the subject of our post on occupancy fraud in property claims. What a covered loss costs to put back is a question about scope, quantity and unit price. A ValuePenguin analysis of NAIC closed-complaint data for 2024, reported by Claims Journal, found claims handling accounted for 65.2% of all insurance complaints (39,680), with delays at 22.2% of claim issues and unsatisfactory settlements or offers at 12.2%. That is ValuePenguin's analysis of NAIC data, not an NAIC publication, and about one claim-handling complaint in eight is about the number.

The mechanism that resolves it sits in the policy, and in some states in the statute. California's standard form fire policy, Insurance Code section 2071, carries the appraisal paragraph: on the written request of either side, each "shall select a competent and disinterested appraiser and notify the other of the appraiser selected within 20 days of the request"; the two appraisers have 15 days to agree an umpire, failing which a judge of a court of record where the property sits selects one; each side pays its own appraiser and the expenses of appraisal and umpire are split equally; and "an award in writing, so itemized, of any two when filed with this company shall determine the amount of actual cash value and loss." The award is itemised by statute, so the dispute gets settled line by line whether or not either file was built that way.

That informality is defined, not rhetorical, and it favours whoever arrived with evidence. Section 2071 spells out what informal means: no depositions, no interrogatories, no requests for admission, no formal rules of evidence and no court reporter. Two of the three panel members fix the number, and nothing in the process creates pricing evidence; it weighs what each side brought. Florida narrows the panel fight further. Under Fla. Stat. 627.70151, a proposed umpire can be disqualified only on four enumerated grounds, among them a familial relationship within the third degree, prior professional representation of a party on the same claim or property, and an employment relationship with a party within the preceding 5 years.

Finch Claims sells analysis tools to public adjusters and policyholders' counsel, and its own Q3 2026 State of the Claim report, covering data through 1 September 2026, says the findings "come from selected property claim files and are not a representative national sample." The 43% comes from 43 claims that had both an insurer valuation and a public adjuster valuation on file. It does not say insurer estimates run 43% low across the book. It says that in files where a public adjuster was engaged and the two sides disagreed, the settlement landed at 1.43x the insurer's first number. Claims Journal, on October 13, 2026, describes Finch as providing claims analysis tools for public adjusters; Finch describes itself as giving "the policyholder side a view of the claim that carriers have always had."

Two findings in the same Finch Claims Q3 2026 report cut against the lowball reading. The median public adjuster request in that 43-claim cut was 1.58x the insurer's figure while the settlement landed at 1.43x, so the negotiation or the panel came in below the ask. And in a different cut, 134 claims where the insurer had paid nothing for six months or more, the median increase from the earliest valuation to the final amount was 1.29x while 26% of those files settled below 1.00x the insurer's earliest valuation. Roughly a quarter of that disputed tail ended under the carrier's first number. The 43 claims behind the headline are also about one file in 200 of Finch's own 9,333-file database (derived), which sets the ceiling on what the figure can carry.

The trigger condition for that tail is worth stating, because the dataset is often read as an accusation. Public adjusting is a licensed, regulated profession with statutory fee caps, and in Finch's data the public adjuster usually arrives before the insurer's estimate exists: the median time from the peril to engagement was 11 days across 7,870 files, with 43.1% engaged within one week. These are mostly not files where someone read a carrier estimate and objected, but files where both sides built estimates in parallel from the start. It is the asymmetry we described in casualty in AI-assembled attorney demand packages and claim severity: the claimant side assembles its ask faster than the carrier side can examine it.

Finch's methodology note is the most useful sentence in the report for a claims organisation: "most of the gap sat in the price of repairs already on the list, not in what was omitted." That inverts the story most estimate-dispute writing tells, in which a hurried scope walk misses the subfloor. In these 21 files the subfloor was generally on both lists. The two estimates disagreed about what replacing it costs. That is 174 contested line items in total, 84% of them pricing differences and 16% omissions (derived). Finch's own summary panel states the ratio as 9 to 1; the counts it publishes divide to roughly 5 to 1, so we use the counts.

The dollar concentration points the same way. Heavy equipment ($1.22M), labour ($0.78M) and permits and fees ($0.61M) together accounted for 42% of the entire measured dollar difference across those 21 claims. Those are totals across the sample, not per-claim figures. Three categories, two of which are price inputs rather than scope items, carry more than four dollars in every ten of the gap.

By trade, the percentage gaps run deepest where the price is least observable from a photograph.

Finch publishes no per-category sample size for those percentages, and they describe files where a dispute already existed, so read them as the shape of the argument rather than as a market error rate. The shape is informative: plumbing and labour are priced from rate assumptions and crew hours rather than from measurable quantities, and they sit at the top. Permits and fees land in the dollar table for a related reason. Only 32.3% of the claims Finch could read carried policy limits covering code upgrades (978 of 3,030 files), typically capped at 10% of dwelling coverage, so the fee line is a pricing question and a coverage-limit question at once. Composition drives the headline too: the median gap was $25,621 across all claim types in Finch's broader comparison, rising to $45,160 on fire claims, about 70% of that sample. The $76,000 median belongs to the 21 large itemised claims only.

EagleView announced in June 2025 that a benchmark study by CompassData, using drone and terrestrial LiDAR on single-family residences in the Denver metro in April and May 2025, found its roof linear measurements 98.77% accurate, with an average difference of 0.2 linear feet (2.4 inches); roof area came in at 98.43% and slope at 98.49%. The announcement does not disclose the sample size, and the study was commissioned by the vendor and run by an independent surveyor. Read conservatively it still describes an evidence standard: a named methodology, a named third party, a date, a geography and a published error.

There is no equivalent for price. We looked for a published neutral dataset that measures insurer estimate accuracy and found none as of October 19, 2026. The estimating platforms are not the gap either. Xactimate is the dominant property estimating platform in US claims, both sides of a disputed file normally produce an Xactimate estimate, and Verisk's price lists are regional and periodically refreshed, which is the point: the platform produces the number, not the evidence that defends a specific unit price on a specific date in a specific market. The rest of the stack is complementary and aimed elsewhere. EagleView and Hover measure; CoreLogic and Verisk estimate; Tractable's computer vision assesses damage from photographs, which answers what is damaged rather than what repairing it costs here and now. Managed repair and estimate review networks pre-agree the pricing, and the neutral evidence says that works: J.D. Power's 2026 U.S. Property Claims Satisfaction Study of 5,093 homeowners who filed a claim found 41% used a direct repair program, where higher-severity repair cycles ran more than two weeks shorter, and 49% submitted photos for estimates or payments. The photograph proves the damage; it does not price the repair.

That asymmetry is what Hesper is built against: investigation-grade evidence behind every decision, not just the suspicious ones. The depth built for SIU work, sourced and timestamped and reconstructable, is the depth a settlement valuation needs when a three-person panel with no discovery and no court reporter is about to fix the number. Clean claims resolve straight through; suspicious claims get an investigation-grade workup.

Florida: the estimate is a deliverable with a clock on it

Fla. Stat. 627.70131 treats the estimate as a deliverable, not working paper. Subsection (3)(e): "The insurer must send the policyholder a copy of any detailed estimate of the amount of the loss within 7 days after the estimate is generated by an insurer's adjuster." Subsection (7)(a) requires the insurer to pay or deny within 60 days of notice of an initial, reopened or supplemental claim, and to "provide a reasonable explanation in writing of the difference" whenever the payment comes in below its own detailed estimate. Subsection (6)(b) requires a partial payment to carry a continuing-evaluation notice in 12-point bold uppercase type.

California: three options when the claimant's estimate is higher

California comes at it from the document side. 10 CCR 2695.9(d) requires insurers to "supply the claimant with a copy of each document upon which the settlement is based," and where the claimant produces a higher written estimate it gives the insurer three options and no fourth: pay the difference; on the claimant's request, "promptly provide the claimant with the name of at least one repair individual or entity that will make the repairs for the amount of the written estimate"; or reasonably adjust the claimant's estimate and provide the adjusted copy. Option two converts a pricing position into a testable claim about the local market.

Texas: the leverage is in the case law and the fee cap

Texas has no equivalent statutory estimate-delivery deadline that we could locate, and its leverage sits in the appraisal case law and the fee rules instead. Public adjuster compensation is capped state by state rather than nationally: Fla. Stat. 626.854(11)(b) caps it at 20% of claim payments, or 10% for claims arising from an event that is the subject of a gubernatorial state-of-emergency declaration, while Tex. Ins. Code 4102.104 caps total commission at 10% of the settlement and bars a percentage-of-settlement commission entirely if the insurer pays or commits in writing to pay policy limits within 72 hours of the loss being reported. The NAIC's Public Adjuster Licensing Model Act treats a fee cap as an optional section, because most states do not require one.

Read those duties together and a claims organisation is already operating under a documentation standard it did not choose: the estimate goes out on a clock, a payment below it needs a written explanation, and in California every document the settlement rests on is disclosable. Whether the prices inside that estimate can be defended is a separate question, and the one the file usually cannot answer. That is the argument we made about claim decisions generally in the evidence layer behind a claims decision: the rules ask what you did and when, and the file has to answer a year later.

The Florida Office of Insurance Regulation announced on September 2, 2025 that it had examined 10 companies on Hurricane Ian and Idalia claims handling and penalised eight of them $2,075,000 in total: American Coastal, American Mobile Insurance Exchange and Clear Blue at $400,000 each, Monarch National $325,000, Tower Hill Prime $250,000, TypTap $150,000, Centauri Specialty $100,000 and Sutton National $50,000. A bulletin on November 3, 2025 added Kin Interinsurance Network and Slide Insurance at $250,000 each, taking the total across the 10 examined companies to $2,575,000.

The violation list is a list of process failures: improperly appointed adjusters, failure to acknowledge claim communications timely, missing disclosure statements on damage estimates, failure to provide the Homeowners Claims Bill of Rights, failure to pay required interest, and failure to resolve claims within the 90-day statutory period then in force for those hurricane claims, now 60 days. One examined company had an error rate above 60% on Hurricane Ian claims and above 80% on Idalia for missing disclosure statements. Commissioner Michael Yaworsky's statement in the November bulletin made the point in one sentence: "Claims management practices must always be efficient and fair, especially after Hurricanes."

The regulatory signal is narrower than carriers usually brace for, and it has a consequence. An examiner opens the file to see whether the estimate was delivered, disclosed and explained on time, so the estimate is regulated as a process output, not as a priced conclusion. Whether its unit prices were defensible is not a finding an exam makes, which means the pricing-evidence gap carries no regulatory penalty and the whole cost of it lands in the settlement. The forum shifts with the statute, and lawsuit share is not estimate-dispute share: per the state's July 2026 Property Insurance Stability Report, reported by Insurance Journal, Florida went from 79% of nationwide homeowners claims lawsuits in 2020 on 9% of claims filed to 41% at the end of 2025 on 5% of claims filed. The disagreement about the number did not go anywhere.

In Barbara Technologies Corp. v. State Farm Lloyds, 589 S.W.3d 806 (Tex. 2019), the loss occurred on March 31, 2013. State Farm's inspection put the damage at $3,153.57 against a $5,000 deductible and the claim was rejected on that basis. Appraisal was invoked on January 9, 2015; the award issued on August 18, 2015 at $195,345.63; and State Farm paid $178,845.25, the award less depreciation and the deductible, on August 25, 2015, nearly 29 months after the loss. The court held that invoking appraisal and paying the award neither establishes nor insulates the insurer from liability under the Texas Prompt Payment of Claims Act.

Ortiz v. State Farm Lloyds was decided the same day. State Farm valued hail and wind damage below the deductible; after being compelled to appraisal, as the defence firm Thompson Coe records in its summary of both 2019 opinions, the appraisers awarded "almost ten times more than State Farm's original damage assessment," and payment of the award barred the contract and bad-faith claims but not the prompt-payment claim. Both are legal landmarks rather than typical gaps. What makes them useful is the pattern they share: a first number set below the deductible, so the claim closed without the file ever having to defend a price.

The time is the part that generalises. Finch reports a 318-day median from intake to final payment across its 9,333 selected files, banded from 160 days for claims under $50,000 to 538 days above $1M, with fire claims at a 379-day median (n=2,000). J.D. Power's 2026 study of 5,093 general homeowners claimants put the average time from first notice to final payment at 40.7 days, down 3.4 days year over year, with repair completion averaging 29.6 days. Those are two different populations, not a like-for-like comparison. Read as populations, the spread is the honest reason a carrier should care about Finch's book rather than dismiss it: a file that enters that population runs roughly eight times the cycle time of a normal one (derived).

Neither case was decided by a better adjuster. Both were decided by what the file could prove about price, years after the people who set the number had moved on.

Written as a checklist it looks manageable. Applied to an estimate running hundreds of line items it is not, which is why it does not happen.

1. The scope, with the measurements it rests on and their source and date.
2. The quantity for each line item, derived from those measurements, not asserted.
3. The unit price, with the price list, region and date it came from.
4. The local market basis: what this unit of work was priced at in this county in this period, and on whose evidence.
5. The permit, code and fee requirements behind the fee lines, with the jurisdiction's own document and the policy's code-upgrade limit.
6. A revision record: every change to a line item, who made it, when and why, plus the written explanation Florida requires for any payment below the insurer's own estimate.

The reason it stays a checklist is volume. A typical claim file in Finch Claims' Q3 2026 set carries 57 emails across 14 threads with 15 documents attached, and the heaviest 5% hold more than 500 emails and 241 or more files. The policy side is no tidier: after deduplication Finch counted 2,400 distinct policy forms across those 9,333 claims from 3,103 raw templates, with 67.4% appearing in a single claim only. Nobody carrying a CAT caseload builds a sourced pricing record for every contested line item in a disputed estimate, and no sensible workload model asks for one.

This is the work Hesper's agents do. Hesper's property claims agents quantify damage from photos and documents, review and adjudicate supplements, cross-reference costs against regional pricing databases and OEM material costs, and recommend reserves, with every value carrying its source and timestamp. Hesper does not publish a price list and does not replace an estimating platform; it builds the record that defends the number the platform produced. By Hesper internal benchmark a manual investigation runs 14+ days per case, one investigator carries 200+ cases, and only about 25% of flagged claims get a real workup. The platform runs 15+ investigation phases in parallel, in minutes, not weeks, taking coverage of flagged claims to 100%. Adjusters review evidence-backed files instead of building them, and reserve and settlement authority stays with the adjuster.

The cheapest version of this is upstream. Sample closed files, pull the contested line items and ask one question of each: did this price carry a source and a date. That is a different exercise from the usual audit, and it is the one that finds the exposure before a panel does. It is the same discipline as our claims leakage audit checklist. The same evidence file then carries across stages: the coverage position at stage 3, the valuation at stage 5 and the settlement response at stage 6 all cite one sourced record, so the price defended in a negotiation is the price documented at estimate.

The policyholder side built a line-item analytics layer and published its methodology, its sample sizes and its disclaimers. There is no equivalent discipline on most carrier files. A carrier, TPA or MGA can match that documentation standard at the line-item level, or it can keep settling the difference in an informal proceeding with no discovery, no rules of evidence and no court reporter.