Restoration first, then everywhere else it happens. What it looks like, why it happens, what we can do about it, whether we can take a cut, and how long the money takes to arrive.
Written 5 August 2026 · for Nick · internal
A 13-agent adversarial verification ran against this report a few hours after it was written. Read the verdict here. It killed a lot of what is below, and the pattern is worse than "some numbers were soft" — several were invented by vendors and falsely credited to real trade associations, then repeated back by search engines until they looked like consensus.
Dead, do not quote from this page: the 22–28% supplement figure (third failure — the document it is pinned to contains no statistics at all) · "law firms lose 15–30% of collectible billable hours" (a billing-software vendor's bare assertion) · the construction 8–12% and 30% unbilled figures (no primary source exists) · the veterinary 5–10% and 17%-of-lab-tests figures (vendors quoting vendors) · the dental 10–15% figure (a billing company's marketing page) · the healthcare 1% and 3–5% leakage figures (asserted by someone selling the fix).
Still standing: the recovery-audit and contingency fee ranges · the ATRI detention total · the mechanism itself in restoration and insurance-claim work. And a new finding that matters more than any of them: a fee taken as a percentage of insurance proceeds is legally dangerous in Florida and Texas — section 6 of the verification.
Every business that does the work before it agrees the final price has this problem. That is the whole pattern. When a job is quoted, done and invoiced in that order, almost nothing leaks. When the work expands while it is happening — extra rooms to dry, an extra hour waiting at a dock, an extra test the vet ran — the paperwork that would turn that extra work into money has to be created by a tired person, at the end of a long day, on the thing they care least about.
So the money is not lost to fraud, incompetence or bad pricing. It is lost to the gap between doing the work and writing it down.
And the second answer, which matters more commercially: taking a percentage of money you find is not a new idea we would be inventing. It is a mature industry with public companies in it — it just has never been built for businesses this size. That is the opening, and it is also the thing to be careful about.
How to read the numbers
Every figure below carries a grade. Measured means it comes from a real, dated, named source. Vendor claim means somebody selling the solution said it — useful for direction, never quotable. Unverified means it is circulating in the industry and I could not stand it up. No figure here came from anyone's books, and none of it may be shown to a prospect until we pull their real numbers.
Five separate holes. They are usually described as one problem — "we under-bill" — which is why nobody fixes any of them.
The insurance adjuster writes the first estimate fast and light, usually before anyone has pulled up flooring. The real scope shows up during demolition and drying: more affected area, more equipment, more days. That extra work only becomes money if somebody writes a supplement, with photos and moisture logs attached, and sends it to the carrier. Do the work and skip the paperwork, and the company paid for it out of its own margin.
The industry figure everyone repeats is that companies which consistently document overhead-and-profit entitlement collect 22–28% more per job. Unverified I have now tried twice to source this and failed. It is attributed to a 2024 Xactware analysis that I cannot find. Two other numbers I found this week — "18% of gross revenue lost to poor supplement processes" and "initial carrier offers fell from 87% to 59% of net invoice" — are not in the articles they were attributed to. I checked both directly. Treat all three as rumour until a real client's books say otherwise.
What I would trust more, because it is specific and internally consistent: supplements submitted in the carrier's own estimating format with photo documentation recover roughly $7,000–$8,000 per claim versus $3,000–$4,000 for the same supplement sent as a plain PDF invoice, because carriers downgrade non-standard submissions. Vendor claim The mechanism is real even if the numbers are marketing: format and evidence change the payout on identical work.
Air movers and dehumidifiers bill by the day. The carrier pays for units it can see evidence for and cuts the rest. If the moisture readings flatline for four days with no documented drying strategy, the equipment time gets slashed — the work happened, the billing does not survive.
Five hundred to two and a half thousand dollars per job that the homeowner owes and routinely never pays, because chasing it is nobody's actual job. It is not a pricing problem or a collections problem. It is an ownership problem.
They dry the house, then somebody else rebuilds it. The rebuild is worth several times the dry-out. Nobody owns the moment in between, so it leaks constantly. I could not find a single measured industry benchmark for how often that handoff is lost — which tells you how unmanaged it is. The first client's own number here will be the most interesting figure in the whole audit.
Insurers pay on their own schedule and nobody is chasing on a calendar. Not strictly "never billed" — but money earned, invoiced and then quietly waited on for four months is the same hole with a longer fuse.
These five explain the pattern in every industry in section 6, not just restoration. That is the finding that makes this a business rather than a restoration service.
"Your reports can only show you what you billed. They physically cannot show you what you did and didn't bill. That is what we go and find."
Four things, in this order, and the order matters.
First, count it. Go back over the last twelve months of completed jobs and find the gap between what was done and what was charged. This is the free audit, and it is the entire sales pitch: it produces a number from their records, not ours. It is also the only honest way to size the opportunity, since every published figure in section 1 fell over when I pushed on it.
Second, stop the leak on live jobs. A daily check on every open job that says what evidence is missing while the job is still open and it can still be taken. This is unglamorous and it is the highest-confidence thing on the page. It is also the precondition for everything else — an undocumented supplement is a denied supplement.
Third, recover what is still recoverable. Draft the supplement, attach the evidence, submit it in the format that gets paid rather than the format that is convenient. This has real competition already: several software vendors, offshore staffing shops at about $8 an hour, and the estimating software company itself now sells an AI version. We do not win this on tooling. We win it by being the only party who does all four steps and is accountable for the resulting number.
Fourth, own the handoff and the deductible — the two holes that have no vendor at all, because they are not software problems. They are "nobody's job" problems, and we are selling somebody whose job it is.
You asked whether we can take a cut of free money. The answer is yes, and the reason to be confident is that this is not a theory — it is an existing industry with several distinct mature versions of it. I found four, all charging exactly the way you described.
| The business | What it recovers | What it charges | Grade |
|---|---|---|---|
| Recovery audit (PRGX, apexanalytix, Cotiviti) | Overpayments hiding in big companies' accounts payable — duplicate payments, missed discounts, unapplied credits | 20–30% of what is recovered, industry-wide; some as wide as 10–40%. No recovery, no fee. | Measured public filings |
| Medical billing / revenue cycle | Claims that were never coded, never submitted, or denied and never re-worked | 4–9% of everything collected, rising to 10–12% for complex specialties | Measured |
| R&D tax credit firms | Tax credits the company was entitled to and never claimed | 15–40% of the credit recovered; 25–35% typical | Measured |
| Unclaimed property finders | Money sitting with the state that the owner does not know exists | Capped by statute — e.g. 5% for the first two years, 20% after | Measured state law |
1 · Our proposed share is normal, not greedy. A 20–30% share of found money is precisely the recovery-audit standard. We are not inventing a number that will make an owner flinch — we are quoting the going rate of an established profession.
2 · The model is proven at both ends of the market — enterprise (recovery audit) and small practice (medical billing at 4–9% of collections). It works. It is bankable. Public companies do it.
3 · And here is the gap, which is the whole opportunity: recovery audit exists for companies big enough to justify a team of auditors, and medical billing exists because healthcare is regulated into standard formats. Nothing serves the $5–50M owner-operated trades business. The recovery per company is too small to send consultants at, and too messy for software alone. That gap is the business, and AI is the only reason it is now closable.
On "free money we just take some" — one caution worth holding. It is free money to them, and that framing sells. It is not free to us: somebody has to go through twelve months of jobs, and the first pass is real labour before a dollar arrives. That is the float problem the plan already names. The recovery-audit industry solves it by auditing enormous payment files where one query finds thousands of errors at once. We solve it by making the first pass cheap enough to run — which is an engineering problem with a known answer, not an unknown.
Three different clocks, and they get confused constantly.
| Clock | How long | What it means for us |
|---|---|---|
| Finding it (our audit) | Days to a couple of weeks per client for a first pass over a year of jobs | This is our cost, paid up front, before any revenue. The 90-day commitment fee exists to cover exactly this. |
| Getting it approved (the carrier) | 2–3 weeks for a properly formatted, documented supplement · 6–8 weeks for a scrappy PDF one Vendor claim | Quality of submission roughly triples the speed. This is a lever we control, and it is worth saying out loud in the pitch. |
| Getting the cash (the payment) | Insurers commonly pay in 60–120 days; the best specialist billing firms claim an average of 32 Vendor claim | Add it up: money we find in week two is cash in their bank somewhere between month three and month six. |
Roughly 90 to 180 days from "we found it" to "it landed." If we get paid only when they get paid, our first meaningful revenue from a client is a quarter to half a year after we start work. That is survivable with a day-one commitment fee and fatal without one — and it argues for billing our share on approval rather than on collection, which is a real negotiating point and not currently decided.
This is the part of your question I found most interesting, because the answer is almost all of them — but the size and the ease vary enormously, and a few are much better targets than restoration.
| Industry | What goes unbilled | Size of the hole | Grade |
|---|---|---|---|
| Law firms | Time worked and never recorded, or written off at pre-bill review | 15–30% of collectible billable hours lost. Average write-off 14% at mid-sized firms. Firms billing within 14 days write off 6%; at 45+ days it is 18%. | Measured |
| Specialty trade contractors | Change orders — extra work approved verbally and never documented | 77% have written off change-order work they could not collect. Estimates of 8–12% of completed work unbilled. Internal processing averages 22 days, GC approval another 26. | Measured survey |
| Trucking / freight | Detention and accessorial time — hours a driver sat waiting | $15.1 billion a year industry-wide, and less than half of detention invoices get paid, because nobody documented the time. One carrier example: $36,000/yr missed on 20 loads a month. | Measured ATRI |
| Veterinary practices | Lab tests, injections, recheck consults — small items, huge volume | 5–10% of all charges missed on average, some audits near 20%. 17% of lab tests go unbilled. Automation moves capture from ~90% to 96.5%. | Measured |
| Hospitals & health systems | Care delivered and never charged | 1–3% of net patient revenue per year (HFMA), some estimates 3–5%. On a $500M system that is ~$5M of care given away. | Measured HFMA |
| Dental practices | Same shape, smaller scale | First-time audits find gaps on 10–15% of encounter days. | Measured |
| HVAC / plumbing / field service | Parts off the truck, time worked, trip charges — and unfiled manufacturer warranty claims | No credible industry-wide figure exists. Illustrative only: one missed $150 call a week is $7,800/yr. Warranty claims are the interesting half — technicians skip them because filing is tedious. | Vendor claim |
| Restoration | Supplements, equipment days, deductibles, the rebuild handoff | Every published figure I tested fell over. See section 1. The mechanism is solid; the sizing is not yet real. | Unverified |
Law firms, trades contractors, trucking, vets and hospitals all have measured, sourced, survey-backed numbers. Restoration has vendor blogs quoting each other. That does not mean the hole is smaller — the mechanism is clearly there. It means we would be walking into the one vertical where we cannot yet prove the number, when four others hand it to us in print. Worth sitting with before the vertical decision gets locked.
The three that look better than restoration on this specific lever:
And the honest counterweight to all of that: the plan does not pick a vertical on this lever alone. Restoration was chosen partly for job size, urgency, marketing fit and your own history. This section is one input into that decision, not the decision. What it does change is the order of the homework: verify the restoration number, or accept that the pilot doubles as the verification.
In several states — Florida, Texas and Oregon are named explicitly — helping a homeowner with an insurance claim can legally count as public adjusting, which requires a licence. The laws were recently widened to cover anyone assisting with a claim, not just people calling themselves adjusters. Courts have ruled against contractors over it, and the consequences can be criminal as well as civil.
The line, as the industry's own association describes it: a contractor may document the damage, the scope of work and the reasonable cost of their own invoice, and explain their own pricing to an adjuster. A contractor may not negotiate the claim on the homeowner's behalf, argue coverage, or dispute denials. Contracts should not contain language authorising us to handle their claim, and we should never advertise that we will.
This does not kill the lever — it shapes it. We build the evidence and the correctly-formatted submission; the client's own people submit and negotiate it. That is a legitimate, defensible position and several vendors already operate there. But it needs a lawyer's eye before the first pilot, not after, and it is genuinely new information for the plan. Measured — state regulators and the Restoration Industry Association.
The 22–28% figure has now failed verification twice, and two supporting numbers that surfaced this week turned out not to exist in the articles they were credited to. I am not going to keep quoting it. The plan already lists this as the single highest-priority unknown; this report confirms it rather than resolving it.
The route out is not more searching — it is one real client's realisation rate. One company's last twelve months of jobs settles it permanently, in our own data, in a form we can show. That is one afternoon with one willing owner.
They sit at the top of the decisions page, above the original three gates, so you see them first. Tick what you agree with, then hit "Copy all answers" at the bottom and paste it back to me. Nothing here is decided until you say so.