The constraint moved. It is no longer what can be built — it is what actually gets delivered.

A network effect means the product gets more valuable to each user as more people use it. Direct (each user adds value to other users), indirect or two-sided (more of one side attracts the other), and data (more usage makes the product smarter).

• Marginal cost of one more digital user is near zero. One more building is millions.
• Digital networks have no geographic ceiling. A Phoenix network is worthless in Dallas.
• Real estate's real advantages are scale and cost of capital. Good, but they don't compound.
• What looks like a network effect is usually agglomeration — value from clustering — and it accrues to land values, not to any single owner.

Why the producer side is better ground

• A producer network can tip with 200 nodes instead of 200,000.
• Track record matters enormously and is currently trapped inside private relationships.
• Every project is a one-off, so nothing learned on job one helps job two. That waste is the opportunity.

NAHB, June 2026. Of the total, $46,795 is embedded in the finished lot and $84,939 falls on the builder during construction. Code changes over the past decade are the largest single line item.

• A new conventional homebuilder — effectively closed. Regulation is a fixed cost, and fixed costs favor incumbents.
• A factory producer under the HUD Code — survivable. That's a federal standard preempting local building codes.
• Infrastructure serving producers — not a barrier. Here the regulation is the demand.

Became law July 11, 2026, after passing the Senate 85–5 and the House 358–32. Incentives rather than preemption, so local zoning control survives. But several national layers opened.

Incentives

• Sec. 107 — HUD guidelines and best-practice frameworks for state and local zoning.
• Sec. 208 — $200M annual competitive grants for localities showing measurable supply increases. Sunsets after seven years.
• Sec. 213 — Ties some CDBG funding (Community Development Block Grant, cities' main federal development money) to housing production.

New national layers

• Sec. 209 — Grants for localities to adopt pre-reviewed ADU, duplex, and townhouse designs. Five-year window, 10% reserved rural.
• Sec. 301 — Eliminates the permanent chassis requirement; HUD becomes primary authority on energy standards.
• Sec. 302 — FHA review of modular financing barriers, plus a study on a standardized modular code. A study, not a code.
• Sec. 104 — Every CDBG recipient must publish a searchable database of undeveloped parcels it owns.

Unchanged from Rev. 2 except the threat column, which is new. Rankings reflect strategic attractiveness; the threat column reflects how contested each lane already is.

Tech never solved fragmentation. It wrapped it and sold the wrapper.

Available to a new entrant

• The wrapper. Stripe didn't unify banks and card networks; it absorbed thousands of inconsistent backends and exposed a few lines of code. Plaid did the bank-by-bank integration once so 10,000 developers didn't have to.
• The neutral standard. USB, TCP/IP, EMV. Rivals adopt a shared spec because incompatibility costs everyone more than cooperation. Usually consortium-built; the realistic version here is a builder-owned data cooperative.

Not available

• City-by-city brute force. Uber worked at a cost of billions, with a consumer constituency to mobilize against city hall. Nobody rallies for a subdivision at the permit counter.
• Rewriting the rules. Tesla spent fifteen years litigating dealer franchise laws state by state and still isn't finished.

Rev. 2 named two adjacent players and moved on. That was the weakest part of the document. The lane is more contested than it implied.

• A network among three firms is not a network. It's a vendor relationship with three buyers who can each replace or squeeze you.
• Clayton is vertically integrated across manufacturing, retail, and financing, and Berkshire-backed. Every service you'd sell, they run internally for half the market.
• The industry runs at roughly half its own historical norm — the long-run average exceeded 200,000 units a year.

The realistic business is not a manufacturer network but a placement and delivery network across the fragmented downstream: independent retailers, land owners, transporters, installers, site-prep contractors, lenders, appraisers, insurers.

Their bottleneck isn't building homes. It's finding legal, serviceable places to put them. Berkshire earns twice on each home — the sale and the loan — so any added qualified placement is upside.

Workable structures

• Referral fee per delivered home
• Land-home packaging partnership
• Co-marketing through independent retailers
• Data licensing of the deliverability ruleset

First move

Don't call Clayton headquarters. Work with three to five independent retailers in one metro, prove you convert dead leads into delivered homes, and let the manufacturer conversation happen later from evidence.

Rev. 2 recommended answering what can legally be built on this parcel. Sheet 07 shows that question is being answered by funded competitors, by municipalities directly, and free by design-build firms as lead generation.

Rev. 3 narrows to the question nobody has commoditized: will this specific project actually get built, at what real cost, and what will stop it?

The four layers below zoning

• Private covenants. Both HB 2720 and HB 2928 explicitly preserve restrictive covenants between private parties. An HOA can deny an ADU regardless of what state law and city zoning allow. This data is in recorded CC&Rs, not in any zoning platform.
• Utility and site capacity. Sewer tap availability, panel capacity, service lateral distance, grade and drainage. Determines real cost and is invisible to code analysis.
• Delivered cost. Phoenix ADU construction runs roughly $120,000 to $350,000, with prefab from $85,000 to $200,000 and garage conversions from $60,000 to $140,000. The spread is the whole decision, and it's site-specific.
• Actual outcomes. What was submitted, what was approved, what was rejected, how long it took. The gap between what the code says and what the counter does.

Who pays — corrected from Rev. 2

Not homeowners. Design-build firms already give homeowners free feasibility to win the contract, so that customer has a zero-price alternative. The paying customers are those evaluating many sites or carrying risk:

• Land investors and small builders screening parcels in volume
• Factory-built retailers whose leads die at the placement step
• Lenders and insurers pricing time-to-approval risk
• Municipalities needing production evidence under Sec. 208 and 213

Bottom-up, single metro, with every input named. Three inputs decide the outcome and none has been measured. They are listed first so they can be attacked.

Resulting range, single metro

All figures are assumptions to be replaced by measurement after twenty hand-built reports.

Capital required

• $8,000–$15,000 total to reach first revenue: entity, insurance, parcel and GIS data access, county records access, simple site.
• No outside capital needed or recommended at Stage 1. Raising against zero customers would price the round badly and remove the option to stay independent.

Milestones

Stated in advance so the decision isn't made emotionally later.

• Fewer than 10 paid reports in 90 days at any price above $250. No market at a workable price.
• Hours per report still above 3.0 at report 100. The data isn't compounding; it's a consulting practice.
• Covenant data proves inaccessible or unreliable at scale. The defensible layer doesn't exist.
• A funded incumbent ships covenant and utility coverage before Month 12. Zoneomics, Symbium, and Clariti all have the distribution to do it.
• Buyers say their design-build firm already does this free. If that answer comes back more than half the time, the paying customer isn't there.

Rated against a business feasibility standard, this document currently scores poorly — not because of how it is written, but because Sheet 14 is nearly empty. Document quality here tracks evidence, not prose. The seven actions below are the cheapest available path from reasoning to proof.

They are ordered by cost to disprove the thesis, not by comfort. Action 1 is the fastest, but Action 4 is the one most likely to end the project — which is why it should be run early rather than saved for later.

What each tier buys

• Actions 1, 2, and 4 — roughly one week. Moves the document from a strategy memo to an evidenced hypothesis. Two of the three could end the project outright, which is the point.
• Actions 3, 5, 6, and 7 — roughly one month. Replaces every assumption in Sheets 11 and 12 with measurement, and empties most of Sheet 14. That is the version worth showing to a partner or a lender.
• Nothing beyond that should be written until the above is done. Additional analysis on an unvalidated premise adds length, not confidence.