What changed
FACT (usaspending.gov): DHS/FEMA has obligated very large Public Assistance amounts to state pass-through agencies — $17,365,135,822.49 to the NEW YORK STATE DIVISION OF HOMELAND SECURITY & EMERGENCY SERVICES under disaster 4480DR, $13,280,988,999.99 to the same agency under 4085DR, $14,314,679,947.44 to the TEXAS DIVISION OF EMERGENCY MANAGEMENT under 4485DR, $35,301,159,434.96 under 4339DR (Puerto Rico), plus twenty-plus similar state-level awards ranging $1.5B–$5B. The described purpose is verbatim 'GRANT TO LOCAL GOVERNMENT FOR REPAIR OR REPLACEMENT OF DISASTER DAMAGED FACILITIES' — i.e. the state is not the end user of the money, it is a conduit. HYPOTHESIS (not in source text): these obligations are now aging into the closeout and audit phase rather than the application phase.
Why now
HYPOTHESIS, clearly labelled: the award IDs 4480DR/4485DR/4486DR/4488DR/4489DR/4494DR/4499DR are the March-2020 COVID-19 major disaster declarations, and 4671DR/4673DR/4834DR carry pandemic reimbursement language ('EMERGENCY PROTECTIVE MEASURES TAKEN DURING THE PANDEMIC' — FACT, present in the source detail for ASST_NON_4671DRPRP00000001_070). If that mapping is right, the obligation window is closed and the live workload is closeout, quarterly reporting, and defending documentation against audit. That is a materially different product than 'help me apply' — and the input's own framing agrees: this is the third instance of the same PA-documentation idea reaching this system, so the application-side is not where the unclaimed ground is. I cannot verify the deadline: the input explicitly states 'DEADLINE: Not stated in text.' Treat 'why now' as the weakest link in this brief.
Converging signals
Three things meet at one point, all FACT from usaspending.gov: (1) tens of billions obligated to state emergency-management agencies; (2) award descriptions that name LOCAL GOVERNMENT as the beneficiary, establishing a pass-through and therefore a subrecipient filer class; (3) the same pattern repeating across NY, TX, CA, FL, LA, NJ, NC, IL, MI, WA, OR, MA, PR, VI — 26 distinct state/territory awards in this evidence set alone, which is the replication surface. What is NOT in the evidence and must not be asserted: any complaint, job posting, RFP, or price point. The convergence is structural, not observed.
Customer pain
HYPOTHESIS (founder-domain knowledge, no source URL provided): a FEMA PA subrecipient must retain records for the federal retention period, produce force-account labor and equipment records tying every hour to a specific Project Worksheet, show procurement compliance under 2 CFR 200.317-327, and survive a DHS OIG or state single-audit years after the staff who did the work have left. The failure mode is not 'the form was hard' — it is 'we cannot find the timesheet, so FEMA deobligates $1.2M.' The pain is concentrated at closeout and audit, not at application. I have no citation for OIG deobligation rates; the input asserts it as inference and I am not upgrading it to fact.
Who pays
Two candidate buyers, and choosing correctly is the whole business. (a) The subrecipient itself — a county, school district, or public authority. Reachable but buys through municipal procurement, which is exactly the long-cycle sale the founder avoids. (b) The PA consulting firm — Tidal Basin, Hagerty, Witt O'Brien's, ICF, and dozens of regional shops — that is already retained by those subrecipients and already carrying the documentation burden manually, billing for it. This buyer is a private company, buys software on a credit card or a one-page MSA, has 20-200 concurrent subrecipient engagements, and feels the audit risk directly because their name is on the workpapers. RECOMMENDATION: sell to (b). Selling to (b) is not enterprise procurement; selling to (a) is.
Solved today
HYPOTHESIS: FEMA Grants Portal for the official submission, plus SharePoint/Dropbox folders, spreadsheets, and a consultant's manual binder assembly. Consultants bill hourly or as a percentage of the award for exactly this labor. Incumbent grant-management SaaS (Euna/eCivis, AmpliFund, GrantVantage) targets the state grantor and general grant lifecycle, not the PA-specific project-worksheet-to-audit-evidence linkage. Not verified — I have no source in this input naming any of these.
Why current solutions are bad
The record vault and the audit narrative are separate artifacts. Files live where whoever created them put them. Nothing enforces the linkage from a dollar on a Project Worksheet to the specific timesheet, invoice, bid tab, and photo that substantiates it. So closeout is a manual archaeology project, and audit exposure is unbounded and unmeasured.
Proposed product
Not another 'PA documentation SaaS' — a narrow audit-trail layer. Per Project Worksheet, the product maintains an immutable, hash-chained evidence ledger: every uploaded document is bound to a PW line item, a cost category (A-G), a 2 CFR 200 procurement test, and a retention clock. It continuously computes an 'audit readiness' score per PW and names the specific missing artifact. At closeout it emits the assembled package plus a defensibility memo showing chain of custody. The wedge is the ledger and the gap report, not document storage.
MVP version
Multi-tenant web app. Consultant creates a subrecipient workspace, imports a PW list (CSV export from Grants Portal — do NOT promise API integration until the portal's actual export surface is confirmed, and do not build an unauthorized scraper against a federal portal). Per PW: cost-category tagging, drag-drop evidence with SHA-256 content hashing and append-only audit log, a rules engine encoding a first-pass checklist of required artifacts by cost category, a per-PW gap report, and a zipped closeout package with an index. No portal submission in v1. Postgres + object storage + a Python backend. Real work but squarely inside the founder's demonstrated range.
30-day build
Do not write product code. Do customer discovery on the buyer choice, because the entire brief rests on it. Pull the subrecipient lists that sit behind three of these awards (FEMA's OpenFEMA public-assistance-funded-projects dataset is the obvious starting point) and identify the 20 largest consultants named in state PA procurements. Get 15 conversations with PA consultants and 5 with subrecipient finance directors. The falsifiable question: 'Show me your last closeout binder and tell me what it cost you to assemble.' Kill the idea in 30 days if consultants say the manual binder is a billable revenue line they have no interest in automating away — that is the sharpest kill argument and it is entirely plausible.
60-day build
Only if discovery survives. Build the ledger + gap-report core against one real consultant's live engagement, in exchange for their documentation corpus as design input. Encode the cost-category artifact checklist with a PA-experienced contractor reviewing it — the founder has capital, and a wrong checklist is a product that gives false assurance, which is worse than no product.
90-day revenue plan
Convert the design partner to paid, then sell two more consultants in the same state. Realistically first revenue is a single four-figure monthly contract, not a wave. I am explicitly NOT forecasting revenue inside 90 days from municipalities; that channel is slower than the founder's stated tolerance.
Distribution path
Consultant-first, and it is unproven. There is no evidence in this input of a working channel — no job postings, no complaint threads, no RFPs. Plausible routes: state emergency-management association conferences, the National Emergency Management Association and state chapters, direct outbound to consultants named in state PA contracts (public records — the founder's stated strength). The founder's fire-service background is real credibility in this room. Score distribution low anyway: 'I can get meetings' is not a channel.
Pricing hypothesis
Per-subrecipient-entity subscription to the consultant, roughly $400-800/month per active engagement, with a closeout package fee. Do NOT price per filing here — unlike the ELDT product there is no discrete, high-frequency filing event; the value is continuous custody. The percentage-of-award consulting fee is the reference price to undercut, but the consultant is the buyer, so the pitch is margin expansion, not fee replacement.
Technical difficulty
Moderate. The engineering is unremarkable — the hard part is the domain rules encoding, which the founder must buy rather than infer. Risk of the founder building a checklist that is subtly wrong and shipping false confidence to a party facing federal audit.
Legal / regulatory risk
Real and worth naming. The product asserts audit-defensibility to a party whose exposure is federal deobligation. That is a professional-liability posture: contract must disclaim, and the product should never claim the package satisfies FEMA or an OIG auditor — it organizes evidence, it does not certify compliance. No licensure is required to build it. No platform owner can deplatform it.
Platform dependency
Low if v1 stays at CSV import and package export. It becomes real the moment the product tries to authenticate into FEMA Grants Portal on a customer's behalf — that pattern worked for ELDT because the Training Provider Registry supported it; assuming Grants Portal does is unverified and load-bearing. Verify before promising it.
Founder fit
Strong on shape, weaker on specifics than the ELDT precedent. Shape matches: public money flows, a compelled documentation burden, a filer class, a portal. Fire-service and public-records background is genuine domain adjacency. But ELDT worked because it was a per-transaction submission with a discrete event and a per-upload fee. This is a subscription selling continuous custody to a buyer who currently bills for doing it by hand. That is a different sale, and the founder's stated preference is 'sells through demonstrated value, not relationship sales' — consultant sales are relationship sales.
Breakout potential
Genuinely high if the consultant channel works, because the same ledger serves any 2 CFR 200 federal pass-through — not just FEMA PA. HUD CDBG-DR, EPA SRF, IIJA formula money. 26 states in this evidence set is a replication surface, and the product is identical across all of them because the governing regulation is federal. That is the real prize, and it is why this deserves 30 days of discovery rather than an immediate kill.
Final recommendation
CONDITIONAL PURSUE, gated on a 30-day kill test — do not build yet. The public-money shape is real and the evidence is hard: $46B+ in FEMA PA obligations across 26 state pass-throughs is a documented fact from usaspending.gov, and the subrecipient filer class follows directly from the award descriptions naming LOCAL GOVERNMENT as beneficiary. The system's own heuristic (government-portal mandate opportunities are the founder's best fit, confidence 0.80) applies. But two things separate this from the ELDT precedent and both are unproven: there is no discrete per-transaction filing event to price against, and the most reachable buyer is the consultant whose billable hours the product destroys. The scoring guidance says a funded mandate justifies demand_evidence 8-10 on its own, and I have scored it there — but that guidance addresses whether the obligation exists, not whether the consultant will write the check. Spend 30 days and a few thousand dollars on interviews. If PA consultants will pay to expand margin, this is a genuinely large business with a 26-state replication path and an expansion into every 2 CFR 200 pass-through. If they will not, kill it — because the fallback buyer is municipal procurement, and that is a business the founder has correctly decided he does not want.
Next action
Download FEMA's OpenFEMA 'Public Assistance Funded Projects Details' dataset, filter to NY disasters 4480DR and 4085DR, and extract the actual subrecipient names and per-project obligated amounts. That converts the inferred filer class into a real list. In parallel, pull state PA technical-assistance contracts from NY, TX, and FL procurement records to name the consulting firms actually holding these engagements. Then book 15 consultant calls with one question: what does assembling a closeout binder cost you, and would you rather bill it or automate it.