Next test — divorce-financial-analysis

Prepared 2026-09-18 UTC for claim tsk_20260918032000_ph68b, task research-rollout-2026-09-17:next-test. Sources: screening.md and screening.json (decision: revise, dated 2026-09-18), brief.md (proposed "Paid attorney pilot" experiment row), and research/divorce-keywords-2026-09-17/REPORT.md. Screening said revise, not stop, so this is a draft next-test plan, not a stop recommendation. No outreach, spending, or pricing has been authorized by this document — brief.md explicitly states "No outreach is authorized by this intake."

Riskiest assumption (from screening.md)

That a California family-law attorney will pay ~$999 for an automated,

source-linked support/RSU workup without requiring a credentialed sign-off.

This is riskier than the search-volume gap because it is untestable with more keyword data: screening.md already shows professional-service search demand is real (forensic accountant divorce: 1,600 US/210 CA per month) but that this demand has only been observed for *credentialed humans* (accounting firms, CDFAs), never for uncredentialed software output. Screening also flagged a second, entangled risk — every volume/CAC number collected so far is consumer-search-derived, not attorney-channel-derived — so the cheapest test must generate attorney-channel evidence directly rather than buy more consumer keyword data.

Cheapest next behavioral test

Reuses brief.md's already-drafted "Paid attorney pilot" experiment, scoped down to the smallest version that produces a yes/no signal on the riskiest assumption:

What is offered: one narrowly scoped, honestly described, manually

fulfilled support/RSU workup (no software product exists yet — a human produces the source-linked schedule, described accurately as manual work, not "AI software").

Target customer: a California family-law attorney at a solo or small

firm with a currently open matter involving support income and RSU/stock compensation — the venture's stated ICP (venture.json), not the consumer searchers measured in the keyword report.

Channel: warm, permissioned introductions only (e.g., an attorney

contact willing to refer or vouch, a professional referral, or a CDFA/forensic- accountant contact who already serves this ICP) — explicitly not cold outbound at volume and not a public post, per the charter's no-go list. Brief.md's interview experiment ("10 qualified attorney/preparer interviews ... through a permissioned channel") describes the same channel constraint and should be run first or alongside, since it is cheaper and lower-risk than a paid offer.

Price assumption: the brief's unapproved $999/matter hypothesis (or the

$499/$1,500 alternatives named in brief.md) — stated to the attorney as a proposal to react to, not a published or committed price. Actually charging money requires price approval, which this document does not grant.

Costs to track (from screening.json base case): delivery cost per

matter (~$200 assumed, unverified — actual manual-fulfillment hours must be timed), founder/preparer hours spent per matter, and any professional-review or liability cost if a credentialed reviewer is looped in to test whether sign-off changes willingness to pay.

Sample size: 10 qualified firms with open matters (brief.md's number),

offered one at a time starting with 1–2 to catch delivery-feasibility problems before scaling attempts.

STOP / ITERATE / BUILD thresholds

Using brief.md's own decision rule, unchanged:

STOP this wedge: zero accepted paid offers after registered exposure to

qualified firms (i.e., attorneys with real open matters actually saw and responded to the offer, not just "no reply"). Also stop if delivering even one workup requires mandatory professional sign-off that erases the base-case $42/matter margin (screening.json economics).

ITERATE: partial signal — e.g., interest without payment, or payment

only at a lower price point, or acceptance only when bundled with a credentialed reviewer. Revise price, scope, or the professional-review boundary and re-test with a small additional batch before any S1 (build) or paid-traffic step.

BUILD-supporting signal (not a build authorization by itself): at least

1 accepted paid offer near the tested price, favorable measured delivery cost, and evidence that a second matter follows from the same firm (tests the durability gap in screening.md — whether attorneys are a repeat buyer). Per SCREENING.md, this still only "supports the S1 gate subject to economics" — it does not authorize S2 build or a public price.

Exact owner decision needed from Yuval

Two authorizations this document cannot self-grant, both required before the test above can run:

Approve a specific test price (or price range to react to) from the

$499 / $999 / $1,500 hypotheses in brief.md — none of these is an approved selling price today.

Approve the recruitment path — name or authorize the specific warm,

permissioned introduction(s) to reach qualified California family-law attorneys with open matters, since brief.md states no outreach is authorized by intake and the charter forbids cold outbound at volume and forbids posting to any audience without Yuval doing the posting himself.

Until both are granted, the venture stays at "revise": the keyword evidence in hand does not resolve the credentialed-sign-off assumption, and no further keyword purchases would resolve it either.