4. Expanded Feature Catalog (What PCA Firms Need)

These are the features a small or mid-size firm writing lender PCAs would expect or ask for. They're each classified with the same honesty as the stated list. Being requested doesn't mean you should build it. Demand here is analyst judgment, not interview data: no firm has been interviewed yet. Replace this with real demand signals after the 20 interviews in 08.

Legend: TS = Table stakes, D = Differentiator, X = Distraction, L = Liability. Priority: P0 = v1, P1 = next, P2 = later, — = don't build.

4.1 Field capture

Feature	Category	Priority	Notes
Offline tablet capture (PWA): components, quantity, age, condition, photos, voice notes	D	P0	The structure is the product. Must survive a full site day without signal.
Unit sampling tracker (units inspected vs required, vacant/down units)	D	P0	Freddie has stricter unit-inspection rules than Fannie (10). A tracker prevents the most common miss.
Photo auto-labelling to component and location	TS	P1	Saves time, expected from any modern tool.
Drawing or site-plan markup	X	P2	Nice, rarely required. Layer does it well.
Drone or 3D capture integration	X	—	Expensive and not required by the formats.

4.2 Analysis and report

Feature	Category	Priority	Notes
Useful-life tables (GSE-published) with firm overrides	D	P0	Values outside the table are a classic reviewer flag.
Firm cost library (unit costs by component and region)	D	P0	Every firm guards its own. Import it from their last reports.
Immediate-repair list and replacement-reserve schedule	D	P0	The Excel work assessors hate most.
Fannie Mae Form 4099 output (full and streamlined)	D	P0	Start with the one format your pilot firms write most.
Freddie Mac Form 1105 output	D	P1	Second format.
Plain ASTM E2018 report (bank and CMBS lenders)	D	P1	Same data, different layout. Widens the market.
Pre-submission check against the lender format	D	P0	The painkiller: missing items, inconsistent totals, out-of-table values.
Form-version management	D	P0	Fannie redlines in 2022, 2024, 2025 and 2026. Update once, for every firm.
AI draft of narrative boilerplate from structured data	TS / L	P1	Label every AI paragraph for engineer review. Never let AI set ratings, useful life or costs.
AI "write the whole report"	L	—	One invented deficiency in a lender report ends trust.
Senior-engineer review workflow (comments, sign-off, version history)	D	P1	Firms already review every report. Doing it in the tool keeps them in the tool.
HUD capital needs assessment output	D	P2	Second report family. Rules not verified in this run.
Condo reserve study output	D / L	P2	Pivot option. Different buyer and standards.

4.3 Firm operations

Feature	Category	Priority	Notes
Job list with lender, due date, assessor, status	TS	P0	Simple. Not a scheduler.
Invoice export (QuickBooks, CSV)	TS	P1	Firms invoice net 30. Don't process payments.
Seats and roles (assessor, reviewer, principal)	TS	P0	Needed for pricing tiers.
Scheduling with client booking	X	—	Lender-ordered work.
Pre-inspection agreements	X / L	—	Set by lenders and master agreements.
Payment processing	X	—	No need to hold or move money.
Integration with due-diligence ordering platforms	D	P2	Only if a platform has an API and several customers use it. Not verified.

4.4 Firm-level insight

Feature	Category	Priority	Notes
Hours per report and bounce-back rate by assessor	D	P1	It's the money outcome, and the case-study number.
Anonymized cost benchmarks across firms (with consent)	D / L	P2	A data moat later. Firms may refuse to share costs. Ask first.
Portfolio capital-planning dashboards for owners	X	—	A different buyer, served by Kykloud and others.

4.5 Net effect

Of roughly 30 features, about 12 are real differentiators for this wedge, and about 9 belong in v1. Every residential staple (scheduling, agreements, payments) is a distraction here. That's the clearest sign that the as-described product and the recommended one are different businesses.