What changed
The CDFI Fund published the FY2026 BEA Program funding round NOFA on 2026-06-30 (Federal Register, FACT). Formula-based grants go to FDIC-insured banks/thrifts that increased loans, investments, and service activities in economically Distressed Communities or increased assistance to certified CDFIs during specified baseline/assessment periods (FACT from NOFA text).
Why now
A live NOFA with a fixed application window makes this seasonal and urgent: every applicant bank must complete the same geocoding + period-over-period computation before the deadline, and the FY2026 clock started June 30. Selling into an open round means the buyer's timeline is set by statute, not by the founder's pipeline (deadline date itself is in the NOFA, not the excerpt — verify immediately).
Converging signals
Three signals meet at one point: (1) the FY2026 NOFA (the rule/money), (2) a defined applicant class — FDIC-insured community banks and thrifts with distressed-community activity (the filer class), and (3) a known submission surface — the CDFI Fund's application system, likely AMIS (INFERENCE, portal not named in excerpt). Adjacent Treasury/CDFI Fund awards on USAspending (Capital Magnet Fund awards of $3.8M–$5M) confirm this agency's money pipeline is active and recurring.
Customer pain
To apply, a bank must map transaction-level loan/investment data against BEA-qualified distressed census tracts for two time periods, compute qualifying increases exactly per the award formula, and assemble a compliant application. Community banks rarely have in-house geospatial capability; the work is annual, unfamiliar, error-prone, and the formula punishes mistakes (misgeocoded or misclassified transactions understate the award or draw a compliance finding). HYPOTHESIS: the CDFI Fund's free CIMS mapping tool handles one-off lookups but is painful for full loan-tape batch processing — this is the single most important assumption to validate with 3-5 past applicants before building.
Who pays
The applicant bank itself (compliance/CRA officer or CFO at a community bank), or the consultant the bank hires today. Two channels: sell direct to banks per application, and white-label the engine to the small set of CDFI/BEA consultants who currently do this in spreadsheets. Beneficiary and buyer are the same entity here, which is clean.
Solved today
INFERENCE: a cottage industry of CDFI-space consultants prepares BEA applications for a fee (commonly cited at 5-15% of award in grant-consulting generally — unverified for BEA specifically); other banks do it in-house with Excel, the free Census geocoder, and the CDFI Fund's CIMS tool. Bank CRA departments also own geocoding software (CRA Wiz, Ncontracts) built for CRA/HMDA, not the BEA formula.
Why current solutions are bad
Consultant fees scale with award size, not effort; spreadsheet workflows can't cheaply answer the question banks most want answered BEFORE committing effort — 'what would our award actually be?' An award-estimation step (upload tape → estimated award in minutes) is the wedge: it converts a vague maybe into a concrete dollar figure, and the application generation is the upsell. Incumbent CRA geocoding suites don't compute the BEA formula (INFERENCE — verify feature lists), and the free path (CIMS) does discovery, not computation, packaging, or estimation.
Proposed product
Web tool + productized service: bank uploads a loan/investment tape (CSV); engine geocodes via the free Census Bureau batch geocoder, joins against the CDFI Fund's published BEA-eligible distressed-tract table, classifies transactions into BEA activity categories, computes baseline-vs-assessment-period qualifying increases per the NOFA formula, outputs (a) an award estimate, (b) an exceptions report of unmappable/ambiguous records, and (c) the data tables formatted for the CDFI Fund application. Founder reviews each output before delivery (productized service, not self-serve, in year one — protects quality and justifies price).
MVP version
Pipeline, not platform: Census batch geocoder + BEA eligible-geography dataset (published by the CDFI Fund) + a formula engine coded directly from the FY2026 NOFA + a one-page report template. No dashboard, no accounts — intake via secure upload, delivery as PDF/XLSX. Buildable solo in 3-4 weeks with AI-assisted development; founder has capital for data-security basics (SOC2 is NOT needed to start, but encrypted handling and an NDA template are).
30-day build
Week 1: read the full NOFA and application materials; extract the exact formula, activity categories, period dates, and the real deadline. Pull the CDFI Fund's published lists of prior BEA applicants/awardees (public) — that IS the prospect list. Weeks 2-4: build the pipeline against a synthetic loan tape; call 5 prior applicants and 3 CDFI consultants to validate the CIMS-pain hypothesis and price tolerance; pre-sell 'free award estimate, pay only if you file' to 10 banks.
60-day build
Run paid engagements through the open FY2026 window: flat $4,500 per application package (or $1,500 estimate-only credited toward the full package). Offer one consultant a white-label deal (they keep client relationship, founder runs the engine at $2,000/application wholesale). Every engagement hardens the exceptions-handling, which is the real IP.
90-day revenue plan
Realistic season: 6-12 paid applications = $27k-$54k direct plus white-label volume, all inside the FY2026 window. Post-deadline, pivot the same engine to the off-season products: BEA readiness monitoring ('here's what your award would be if you filed today — and which Q3 loans would raise it'), Capital Magnet Fund and CDFI-certification data prep, and state-CRA geocoding (IL/NY/MA have state CRA regimes) to break the one-window seasonality.
Distribution path
Direct, deadline-driven outreach to a public, finite list: prior BEA applicants and awardees are published by the CDFI Fund, so the entire TAM is a named list of a few hundred institutions. Lead with the free award estimate (demonstrated value, founder's preferred sales motion — no relationship selling). Secondary: white-label to the 5-10 consultants already serving this niche.
Pricing hypothesis
$4,500 flat per application package; $1,500 award-estimate-only; $2,000 wholesale white-label. Flat-fee positioning directly undercuts percentage-of-award consulting. NOTE: avoid pure success-fee pricing — awards are pro-rated if the pool is oversubscribed (FACT from NOFA structure), so the founder should never price on an award amount neither party controls.
Technical difficulty
Low-moderate and squarely in the founder's lane: batch geocoding (free Census API), tract-table joins, a deterministic formula engine, and report generation. The hard 10% is exceptions handling (bad addresses, multi-parcel collateral, activity classification judgment calls) — which is exactly why banks will pay rather than DIY, and why productized-service beats self-serve at first.
Legal / regulatory risk
Low. Data preparation for a federal grant application requires no license (FACT per NOFA — no licensure barrier for data prep). The founder is not making eligibility representations to the government; the bank signs its own application. Handle loan-tape data under NDA with encryption — borrower addresses are sensitive but this is standard consultant-grade data handling banks already extend to CRA vendors. Do not market the award as money the bank 'has' — it is competitive and pro-ratable.
Platform dependency
None that can deplatform. Dependencies are the CDFI Fund's published eligibility data, the free Census geocoder (both public), and the stability of the BEA program itself — BEA has run for ~30 years but is appropriations-dependent (see kill arguments).
Founder fit
High. This is the founder's proven ELDT shape one notch removed: a federal money flow forces a defined class to assemble and submit structured data to a government system, and he monetizes the paperwork layer per transaction. Differences from ELDT to respect: the money is competitive (not a compliance obligation), the buyer is a bank (more conservative than truck-driving schools), and the round is annual (seasonal revenue). Systems-thinking + public-records + fast prototyping all apply directly.
Breakout potential
Moderate. BEA alone is a $30-60k/season solo income stream, not a company. The breakout path is the engine, not the program: 'geocode a financial institution's activity against any government eligibility geography and compute the incentive' generalizes to Capital Magnet Fund, NMTC, CDFI certification, state CRA exams, and the recurring CRA-modernization data burden — a genuine micro-SaaS with the BEA season as the paid pilot.
Final recommendation
PURSUE, scoped as a seasonal wedge, not a standalone company. The demand check costs one week and five phone calls to prior applicants (a public list); the MVP costs three more weeks. If ≥3 of 10 pre-sold banks convert on the free-estimate offer, run the FY2026 season for $25-50k and use it to fund the generalized eligibility-geocoding engine. If past applicants report the DIY path is genuinely fine, kill it and keep the geocoding pipeline for the next money signal in this family.
Next action
Today: fetch the full FY2026 NOFA from the Federal Register document (2026-13199) and the CDFI Fund application materials; extract the exact deadline, formula, and activity categories; download the prior-round BEA awardee list to build the prospect sheet; schedule 5 validation calls with past applicants.