The opportunity hunt

Two rounds. Thirty-seven researchers, adversaries and critics. Sixty-four opportunities examined. This is everything they found, everything they killed, and what is worth doing next.

6 August 2026 · findings only — nothing here asks you to decide anything

Read this first, because it frames everything below Nothing cleared the bar. Not one opportunity in sixty-four survived the standard we set — that it must get past a hostile fact-check and at least two of three critics. That sounds like failure and it mostly isn't. What came out instead were four corrections to how you understand your own business, and they are worth more than any of the opportunities would have been. They are the first section.
1 · The four corrections — the most valuable thing in here 2 · Two free checks that kill a bad market in ten minutes 3 · Round one — what survived when everything was killed 4 · Round two — the four that got closest 5 · The evidence problem, and it is not going away 6 · What to hunt next 7 · The four questions only you can answer

1 · The four corrections

These came out of an agent whose only job was to read your own business records and ask whether you could actually do any of the things being proposed. Every researcher in both rounds got the same fact wrong, and it is the fact everything rests on.

① You do not have a delivery team. You have a recruiting machine.

Every researcher wrote some version of "he already runs an offshore delivery organisation, so he just points it at a new industry." Round one's own scorecard graded that as your real cost position. It is wrong, and it kills about twenty of the sixty-four outright.

What Heroes and Sidekicks actually is: you recruit, vet, place and manage a person who then sits inside the client's business. You invoice weekly and earn the spread — $16.74 billed against $11.96 cost, about $4.78 an hour. The internal payroll is five people and none of them do client work.

Why it matters: nearly every opportunity found in either round is a managed desk — people on your payroll, doing work you are accountable for, paid before the revenue arrives. You have never run one. Building it is a new company with its own cash requirement, not an existing asset being pointed somewhere new.

What you do hold is better than people realise, in a narrower way. A documented sourcing method you describe yourself as "a marketing/sales function, not HR — sell the opportunity to candidates the way you'd market to a customer", a funnel, a screening system and a full-time head of recruitment. That fits any business where recruiting and placing IS the product. It does not fit any business where running a desk is the product. Exactly one opportunity out of sixty-four was the first shape.

② Your cost base is priced for specialists. Almost everything found needs clerks.

Three researchers independently hit the same wall in three different industries and each treated it as a local pricing problem. It is one structural problem everywhere.

Offshore admin and support in the Philippines sells at roughly $7–$13 an hour all-in this year; specialist work sits at $18–$35. Your $11.96 is not a cheap number — it is a specialist number. You place media buyers, designers, video editors, social managers, and in that band $16.74 is genuinely competitive.

But insurance verification, form review, certified payroll, records checking — that is all clerical. Named competitors publish $9.50 an hour with healthcare training included. Your cost sits at or above their retail price before you add a cent of margin.

You can build a second, cheaper funnel and pay band. That is a project with a cash cost and a quality risk — not an advantage you carry into the room.

③ The $84,000 cost of winning a client is mostly your own hours, valued at a rate you are not currently earning.

Round one produced a number that reframed the whole business: winning one client costs $60,000–$108,000 in year one, so a client must produce about $336,000 of provable value just to break even at a 25% share. That number is why "get in through a network" ranked first.

Then round two took it apart. Between $40,000 and $80,000 of it is 160–320 of your own hours valued at $250. Two problems, both pointing the same way: you work about two hours a week on the business, so those hours are not being sold to anyone else; and $250 is a price you quote, not income you are turning away — there is no live engagement billing it anywhere in the record.

An unearned price used as an opportunity cost is not a cost. If you are not actually turning away work at $250 an hour, that figure falls by something like a factor of four — and the table saying you need $8.4M to $67.2M clients comes down with it. This changes more than any opportunity in the report.

④ The real constraint is cash out and payroll risk, not your time.

Against roughly $10,800 liquid, $86,000–$90,000 of card debt, a $40,000 tax installment and a $7,680 loan, the shapes sort themselves cleanly.

What it spendsReal cost to youVerdict
Your time — selling, research, marketingCheap. You have surplus.Viable
Cash on lists, ads, tools, travelPainful but survivable in low four figuresTest and see
Carrying salaried people before revenueClose to impossible at $10,800 liquidBlocked
Paying out 12–18 months later — equity, exit fees, contingencyNo working capital to bridge itBlocked

There is a fifth constraint underneath that one. The team assessment is unambiguous: the only two people who could run any of this are Dean and Mae, and both are holding up the client book that funds everything. So the only shapes that work are ones you can run yourself with offshore labour underneath.

2 · Two free checks that kill a bad market in ten minutes

Round two's most useful output was a method, not a lever.

One researcher downloaded the government's wage survey and found clerical jobs in behavioural health up 18.6% in three years — and called it a paperwork explosion. The fact-checker tested it against a different government dataset that counts jobs from unemployment-insurance records rather than surveying employers. The industry itself grew 19.0% over the same period. Admin per employee went slightly down. There was no explosion; an industry got a fifth bigger and hired proportionally.

Run that correction — clerical growth minus industry growth — everywhere:

IndustryBack office vs the business it serves
Labour unions+8.2% — growing faster
Outpatient behavioural health−0.3%
Veterinary−2.8%
Physical and occupational therapy clinics−5.8%
Offices of mental health practitioners−29.5%

In four of five industries the back office is shrinking relative to the business. You sell back-office labour. The one place administrative work is genuinely outgrowing the organisation is labour unions — the single buyer in the whole set who cannot accept offshore delivery.

The two checks, and both are free One: clerical growth minus industry growth, taking industry growth from the job-count files rather than the wage survey. Ten minutes per industry. It would have killed three of round two's own findings before anyone wrote them up.

Two: total industry revenue divided by number of businesses, to see how many firms genuinely sit in the size band you want. Thirty seconds. This is what corrected the restoration company count by a factor of three.
One thing that must never leave this page The individual percentages from the wage survey must not be quoted to anyone. The agency publishing it says explicitly not to compare it across years — the method, the collection and the job codes all change. "Routine clerical work is in sustained decline" is defensible. "Order clerks are down 33.7%" is not.

3 · Round one — what survived

Seven researchers, seven adversaries, three critics. Thirty-five opportunities inside the existing frame: independent home-services companies at $5M–$50M, paid a share of what we find.

Not one money-making lever survived. Better quotes, unbilled change orders, trucking waiting-time charges, job margins, chasing invoices, denied medical claims, missed calls, marketing reallocation — all killed, each for one of three reasons and usually more than one: the number behind it was not in the source it was credited to; a named competitor already sells it at a published price a fraction of ours; or the software these companies already own is shipping it as a feature this year.

What was left were positions, not levers

Get in through a network instead of one owner at a time. The largest owner network in home services has 610–800 members paying about $8,000 a year, clustered at $2M–$20M. One agreement reaches more qualifying companies than the entire target list. And the best-funded competitor in the market took exactly this route — over $125 million raised at a billion-dollar valuation, built on the two biggest distribution partners in home services, not on cold calls.

Own the benchmark data in whatever industry gets picked. The only asset in either round that gets more valuable as AI gets cheaper. Everything else is rented. Both industries currently in contention already have someone else holding it — in restoration it is the company that owns the estimating price database everyone prices against.

Your two restoration relationships. Nobody can buy a three-year working relationship inside a customer. It is the one thing no competitor, no software vendor and no acquirer can purchase.

Three numbers that changed the arithmetic

The single most encouraging fact in either round The biggest owner network in home services just expelled every private-equity-owned member — about a third of its membership and roughly half its revenue — deliberately, to keep serving independent owners. An organisation walked away from half its money to protect exactly the customer you picked. That is the strongest evidence anywhere that the independent mid-market is an organised, self-aware constituency rather than a convenient assumption.
And the most uncomfortable one The part of this work a machine genuinely cannot do — persistently chasing a human being at another company — is not a technology advantage. It is staffing. It sells offshore into these exact industries at $8 an hour against your own cost of $11.96. On that reading the durable part of the offer is the commodity part. Two critics disagreed; one did not. It stands as an untested claim.

Nobody has ever tested whether one of your people with today's tools actually finishes this work better than the client's own staff. It is assumed everywhere in the plan and measured nowhere. One client, one month, one timesheet settles it. Above 80% done by hand and this is a staffing business wearing an AI badge.

4 · Round two — the four that got closest

Frame lifted. Any industry, any size, any business shape. Six researchers using six different ways of searching rather than six topics — following government payroll data, hunting work that regulators force on companies, mining low-star software reviews, tracking where investors actually bought, finding what broke in the last eighteen months, and looking for work where a scarce licensed human sets the price.

Twenty-nine opportunities. The critics killed 28, 28 and 27 of them respectively. Applied honestly, the bar leaves the list empty. So these four are what got closest — three of them one afternoon of phone calls away from being settled either way.

① Manufacture the licensed human best fit for you

The idea: recruit career-changers, train them to a national certificate, and place them into jobs a statute requires a licensed person to do. The test case is California court reporting.

Why it is interesting: no software release produces a person a law requires. And it is a recruiting business — the one thing you actually own.

The failure is spectacular and measured. California needs 458 more court reporters against 1,101 employed. Between January 2023 and March 2026 the courts hired 381.8 and lost 366.3 — a net gain of 15.5 people in three years. And 3,007,651 of 4,214,365 family, probate and civil hearings — 71.3% — went ahead with no verbatim record at all.

What changed, with a date: from 1 January 2025 California stopped requiring its hardest exam. A holder of the national voice-writing certificate sits two written papers and skips the dictation and skills test entirely. Voice writing trains in a fraction of the time machine stenography takes.

What it costs to start: $5,000–$25,000 in cash — an advertising test in two California metros, one contracted working instructor for a cohort of five, about $200 a head in exam fees. Roughly a tenth of the $84,000 the existing plan carries. The reason is the direction of cash flow: students pay during training, not eighteen months after.

The honest killer is not cost, it is yield. There is already free tuition at one college, a funded charity giving training away, eighteen programmes, $25,000 signing bonuses, and Los Angeles Superior Court building its own pipeline — and the state still netted fifteen people in three years. That is not a marketing failure, it is a pass-rate failure. 95% accuracy at 225 words a minute is a hard human skill gate. If your cohort passes at one in ten, the economics collapse however cheap the applicants are.

And delete the placement fee. The research applied a private staffing convention — 15–20% of first-year salary — to a public court. No evidence exists that any California court has ever paid an outside recruiting firm anything. What courts pay is a finder's fee to their own employees for referrals, plus a signing bonus to the new hire. Tuition is the business until a court says otherwise.

The threat, in one sentence: California law allows electronic recording for minor cases but not for family law, probate or unlimited civil. The machine already exists and sells into courts in 80+ countries. What holds it out is one sentence of statute — and three million unrecorded hearings are the pressure aimed at that sentence.

② The employment-eligibility paperwork backlog only thing the sceptical buyer would sign

The job: a company with 800 hourly staff and heavy turnover holds thousands of employment-eligibility forms in a filing cabinet that nobody has ever checked. Every one with an empty box is now a fine. Find every defect, fix what can lawfully be fixed, destroy what the rules require destroyed, and hand over a file they can put in front of a federal agent.

Why now, and it is exactly one thing with a date. On 16 March 2026 the enforcement agency rewrote its inspection guidance and moved eleven common paperwork errors out of "technical, you get ten days to fix it" into "substantive, we fine you now" — missing signature date, missing employer name, missing first day of employment, electronic signature defects. Until that date a company could rationally do nothing. That option is gone.

Do not sell it on audit volume. The original research leaned on notices running "ten times the 2024 rate" — true and misleading, because 2024 was a trough of about 230. The agency ran about 6,450 audits in 2019, which every existing firm already served. The loss of the cure period is the new thing. Nothing else is.

Worth about half what was first claimed: a 1,200-employee company at 70% turnover holds roughly 2,500–3,200 forms. First-year clean-up at a realistic quote is $25,000–$45,000; ongoing is $12,600–$29,400 a year. The published competitor price is $19 a form plus a $75 setup fee.

Why this one alone survived the buyer, and it is the most transferable lesson in the whole run: he is buying an artefact he can inspect — a remediated binder, a defect report with counts, a process with a lawyer's name on it — not a number we computed. Fixed price for a deliverable. Never a share of a fine that did not happen.

The door nobody opened, and it is why this is still on the page. A named non-lawyer firm already advertises independent third-party reviews for a large grocery chain's contractors. That is the "big customer forces its suppliers" shape, and it converts an expensive cold sale into a requirement handed down with a name attached. Nobody has checked whether such a slot is open.

The objection that mostly holds: it is a fixed federal form with a published eleven-item defect list — empty boxes, not judgement calls — so once the forms are keyed it is a rules check over structured fields. One critic puts it at 70–85% machine-doable within 12–18 months. On that reading this is an eighteen-month cash window, not a durable business, and it should only be entered knowing that.

③ A free route to ten thousand people about to buy a company channel, not a product

The product this was attached to is dead — first-hundred-days support is what capital providers give new owners free, to win the deal. The channel underneath it survived outright.

Searchfunder is an open online community of more than 10,000 people — search funds, independent sponsors, lower-middle-market buyers. They self-identify, they are about to buy a company, and the median company they buy costs $16 million. Posting one honest question costs nothing. Given that this plan's own arithmetic says finding a client costs $84,000, a free door to ten thousand qualified buyers is worth more than most products.

The correction: the constraint is attention, not access — that platform already carries a directory of 300+ service providers fishing the same pond. A free door with 300 people standing in it is still a free door, but it is a marketing problem, not a gift.

The test: join, read six months of it, post one question with nothing to sell — "What did you most need in your first six months, and what did you pay for it?" If the answers are "a controller" and "a bookkeeper", that is a staffing product you can sell next month with the machine you already run.

④ The same shape, pointed at commercial property appraisal untested — for round three

This is not a tested finding. The version round two tested — training residential appraisers — was killed by all three critics, because the mortgage agencies are permanently removing residential appraisal orders from the market, so it would be building supply into falling demand.

Then the harshest critic named a narrower version his own objection cannot reach, and said nobody has written it. Commercial appraisal is untouched by those programmes, the number of holders of the commercial credential is reported down more than 20% in five years, roughly 4,000 people are reported waiting for a supervisor, and a new simulation-based route into the credential exists specifically because supervisors cannot be found — adopted in California effective 1 January 2026.

No number here should be trusted yet. No tuition price, no verified candidate count. It is on the page because it is the second independent instance of the one shape that survived the hardest critic, and because nobody has spent an hour on it. Two emails settle whether it is even enterable.

5 · The evidence problem

Round one's fact-checkers found roughly twenty-five separate figures that do not exist in the sources they were credited to, three of them carrying entire arguments. Round two, with harder instructions, produced about nine — and two whole lanes came back clean, with every government figure reproducing to within one percent when the checker downloaded the same files. That is real improvement.

But a subtler error replaced it, and it is worth learning: numbers attached to the wrong size of company. A median deal of $110 million quoted to describe a $10 million buyer. A $2.5 billion acquisition used to characterise a market of $7 million clinics. Every market number now needs a second label beside its source: which size of company does this actually describe.

Why this keeps happening, and the only fix It is not carelessness. It is how trade-press statistics work in every industry examined: a real-sounding number, credited to a real organisation, published by a company that sells the cure, with no document behind it. The only defence that works is getting one real company's own twelve months of data. That is not a restoration workaround — it is the first cost of entering any market at all, and it should be budgeted as such.

6 · What to hunt next

Three pulls that would move something from interesting to actionable

Court reporting — two calls, one email, one search, this week. Ask two large Superior Court human-resources offices: "if somebody brought you a fully licensed reporter ready to start, would you pay an outside firm a placement fee, and have you ever?" Ask two working instructors: "of every ten who start, how many pass, and how long does it take?" — that pass rate is the number the whole business turns on and nobody anywhere has it. Email the national certifying body about what a private preparation provider needs from them. And before spending a single day on any of it, search for any California bill authorising electronic recording in family law, probate or unlimited civil. That one event ends this entirely.

Employment eligibility — one afternoon on the door nobody opened. Find one large employer that already requires independent third-party review of its contractors' paperwork. This is the single highest-value unexplored thread in the whole run. Then get two real written quotes for a three-thousand-form backlog as if you were the employer — if both come back above $25 a form there is room underneath them; at $10 there is not. Then get one friendly company to hand over its file for free and produce the two numbers nobody has: forms reviewed per person per hour, and what percentage carry a now-finable defect.

The buyers' channel — join, read, post one question. Costs nothing.

The ground round two could not cover

7 · The four questions only you can answer

Not decisions about the business — facts about you that nobody else holds. The first one changes more than any finding in the report.

  1. Are you currently turning away work at $250 an hour? Half the cost of winning a client is your own hours at that rate. You work about two hours a week on the business, and $250 is a price you quote rather than income you are refusing. If the honest answer is no, that cost falls by roughly four times and every threshold in the plan comes down with it.
  2. Will you put people on your own payroll, doing work you are accountable for, before a client has paid? You have never done this. If the answer is no, every managed-desk idea is closed permanently and the search should only look at recruiting-shaped and placement-shaped businesses from here.
  3. Will you build a second, cheaper recruiting funnel and pay band? Your $11.96 is priced for marketing specialists and sits above retail in every clerical market examined. Fixable — but it is a project, not a decision in passing.
  4. Do you want to recruit Americans? Every credential-pipeline opportunity — the strongest shape found in either round — requires it, and the machine has never done it at volume.
The honest bottom line Two rounds, sixty-four opportunities, thirty-seven agents, and nothing has cleared the bar. What the work produced instead is a much sharper picture of what you are actually holding, two free checks that will stop the next twenty bad ideas before anyone writes them up, and one shape — manufacture the person a statute requires — that survived the hardest critic twice, in two unrelated industries, by accident. That shape has never been searched deliberately. It should be next.