What changed
FACT (cited Federal Register rule, 2026-06-17): FDA classified radiological machine-learning-based quantitative imaging software with a Predetermined Change Control Plan into class II with special controls, creating a 510(k)-style pathway where de novo was previously required. HYPOTHESIS: a parallel opioid-oxygenation-monitor classification exists (signal referenced but not provided in demand_evidence).
Why now
The classification order is roughly three weeks old. The first cohort of startups targeting this device type is drafting submissions now, before consultancies productize fixed-price offerings or FDA publishes richer guidance/templates. The arbitrage window is the gap between the order's publication and the market's productized response — plausibly 3-9 months (INFERENCE).
Converging signals
One verified signal: the FDA classification rule (FORCED BUYER). The signals array is empty and no PAIN or HIRING/SPEND evidence was retrieved, so convergence here is a single regulatory event plus an inferred expertise gap, not a multi-signal convergence. This materially weakens the case versus the hypothesis as stated.
Customer pain
HYPOTHESIS: seed-stage imaging-AI founders must author a compliant PCCP, special-controls conformance matrix, and 510(k) predicate analysis — accumulated know-how held by $400/hr regulatory consultants priced for corporates. No complaint, forum, or job-posting evidence was provided to confirm founders are actually stuck or quoting these prices; the $400/hr and $100k figures are unverified.
Who pays
Seed-stage radiology-AI / quantitative-imaging startups pre-submission (funded, can pay by card or small PO). Secondary: fractional RA consultants who would buy templates to serve more clients. TAM is narrow — plausibly 50-300 companies globally in this device category at any time (INFERENCE, needs Crunchbase/LinkedIn count per the evidence-needed list).
Solved today
Regulatory-affairs consultancies (RQM+, Hardian Health, MCRA and hundreds of independents) and QMS/RA software (Greenlight Guru, Enzyme, Rimsys) with services attached. Some startups hire an in-house RA lead (~$150-200k/yr). FACT is only that consultancies exist; pricing claims are hypotheses.
Why current solutions are bad
HYPOTHESIS: consultants are engagement-priced ($50-150k) and slow; software incumbents cover QMS broadly but have not yet shipped templates specific to a weeks-old classification order. If FDA guidance already includes fill-in templates, or consultants now bundle PCCP cheaply, this is falsified — explicitly untested.
Proposed product
A documentation product: (1) PCCP template mapped clause-by-clause to the codified special controls in the classification order; (2) special-controls conformance matrix; (3) predicate-analysis workbook built from cleared 510(k) summaries pulled via openFDA; sold per-submission at $5-15k flat with a credentialed-RA expert-review upsell (revenue share with a contracted RAC-certified reviewer).
MVP version
Not software first — a 15-page teaser: annotated table of contents of the PCCP template plus one fully-drafted section derived from the published order, a landing page, and 20 direct emails to identified pre-submission imaging-AI startups. This is the convergence hypothesis's own testable prediction and costs under $500 and one week.
30-day build
Run the teaser test (target: 3 discovery calls or 1 paid pilot). Pull all cleared summaries for adjacent quantitative-imaging predicates from openFDA to confirm templating is feasible. Recruit one RAC-certified fractional consultant as named reviewer on revenue share — without this the product is likely unsellable given the founder has no RA credentials. Count the addressable startup population.
60-day build
If validated: author the full package against the codified special controls with the RA partner's review; sell 1-2 founding-customer pilots at $5k with heavy feedback rights; publish one authoritative teardown of the classification order as content-led distribution in RAPS forums and imaging-AI founder communities.
90-day revenue plan
2-4 package sales at $5-10k = $10-40k, plus expert-review upsells. Realistic first-revenue window is 60-120 days, inside the 180-day target, IF the week-one validation passes. If it fails, total sunk cost is under $1k and two weeks.
Distribution path
Direct outreach to a small, enumerable buyer list (fundable via Crunchbase/LinkedIn/FDA pre-sub chatter), RAPS forums, medtech-founder Slack/communities, and SEO on the classification order's exact language — a niche where one good artifact can rank. No ad spend needed. Founder sells via demonstrated value (the teardown/teaser), which matches his stated selling style.
Pricing hypothesis
$5-15k flat per submission package; $3-8k expert-review upsell (shared with credentialed reviewer); later $500-1k/yr update subscription as FDA guidance evolves. Card/small-PO friendly by design.
Technical difficulty
Low on software (document generation from public FDA data), HIGH on domain: the hard part is regulatory judgment, not code. The founder's AI-assisted drafting speed helps but cannot substitute for RA credibility with this buyer.
Legal / regulatory risk
Material. Selling regulatory-submission templates that lead to a rejected or deficient 510(k) invites blame and reputational destruction in a small community; needs strong disclaimers, E&O insurance, and a credentialed reviewer. Not FDA-regulated itself, but adjacent to the founder's stated avoid-list (heavily regulated medical).
Platform dependency
Low. Public Federal Register + openFDA data; no marketplace gatekeeper. Main dependency is FDA itself publishing free templates/guidance, which would gut the product.
Founder fit
Weak-to-moderate, and the government-portal lesson (confidence 0.80) does NOT transfer despite surface similarity: ELDT was a repetitive, mechanical, per-filing portal transaction monetized per upload; a PCCP/510(k) is a one-time bespoke expert artifact where the buyer is paying for credentialed judgment. The founder has no medtech or RA background, no RAC, and his avoid-list includes regulated medical products. His public-records/mandate-reading strength applies only to the research phase, not the sale.
Breakout potential
Moderate if validated: the PCCP mechanism extends to other ML device categories FDA classifies next (cardiology AI, pathology AI, the opioid-oxygenation monitors), making this a repeatable 'new classification order → template package' playbook. But each expansion re-requires domain review capacity.
Final recommendation
CONDITIONAL TEST, LEAN PASS. Do not build. Run only the one-week, sub-$500 falsification test embedded in the hypothesis (teaser + 20 outreach emails + openFDA template-feasibility check). Proceed past that ONLY if it produces a paid pilot commitment AND a RAC-credentialed partner signs on for revenue-share review; if either fails, kill and log the lesson that classification-order opportunities need a credentialed front. This is a genuinely clever pattern instantiation, but it is expert-services in product clothing, outside the founder's credibility zone, and one FDA guidance document away from worthless.
Next action
Query openFDA 510(k) database for cleared quantitative-imaging predicates with PCCPs to verify templating feasibility (half a day, free), then send the 20-email teaser test described in the hypothesis before writing any template content.