Methodology
Recourse adjudicates from the evidence bundle alone: the agreed terms, the contemporaneous interaction log, and the deliverables. The engine reads the log against the terms and allocates fault on three axes — delay, scope, and quality-versus-spec — under the published doctrine below. Where the terms are silent, published silence-in-terms defaults apply, so parties can know the rule before they ever dispute.
Outcomes are proportional, never winner-take-all: fault is expressed in percentages summing to 100, and the money instruction follows the allocation applied to the disputed amount, adjusted for value retained. Every ruling carries a structured reasoning trace citing the rules applied, a confidence score, and — where evidence is messy, values are high, or confidence is low — an escalation flag routing the case up the ladder:
| Tier | Who decides | When |
|---|---|---|
| 1 — Automated verdict | The Recourse judgment engine, in minutes | Clean bundles; the bulk of the queue |
| 2 — Enhanced review | Deeper engine pass, confirmed or overridden by a Recourse reviewer | Messy evidence, low confidence, amounts above the automatic ceiling |
| 3 — Human panel | Independent adjudicators contracted by Recourse — trained on this doctrine, conflict-screened per case, never employees of the platform or either party | High value, fraud allegations, novel questions |
Escalation never goes back to the platform. A platform judging its own transactions is the conflict this venue exists to remove: the judge would profit from the deals it rules on. Every tier is staffed by Recourse; the platform pays for the tier used but has no voice in the outcome, and the venue's fee is never a percentage of the disputed value.
Anonymised rulings feed a growing precedent corpus. Error rates and overturn rates will be published as the corpus grows.
Rule-pack v1 — Marketplace / Contractor Disputes
Silence-in-terms defaults
| When the terms are silent on… | The default is… |
|---|---|
| Deadline | A reasonable time for the vertical; typically 14 days from full briefing (SD-1) |
| Revisions | One reasonable round within the original spec; further rounds are scope changes (SD-2) |
| Acceptance window | No objection within 7 days of delivery — or putting the work to use — is acceptance (SD-3) |
| Early termination | Provider is owed the proportion of fee for work completed and usable (SD-4) |
| Ownership of rejected work | A refunded buyer acquires no right to use the work (SD-5) |
Worked rulings
Ruling I — The rejected logo
A buyer commissioned a logo package with a written style spec and one included revision round. The provider delivered one day late, to spec. The buyer rejected on taste — "not what I imagined" — and refused the offered revision round outright.
Allocation: buyer 75% · provider 25%. Split — 75% of the fee releases to the provider.
[MC-2] The deliverable met the written spec; taste is not breach. [MC-5] The buyer rejected without giving the offered cure a chance, which weighs against the buyer. [MC-3] One day's delay on a ten-day deadline is minor but real — a modest deduction against the provider. [SD-5] The refunded portion carries no right to use the marks.
Ruling II — The missed trade show
A site build carried a hard July 1 deadline tied to a trade-show launch. Mid-project the buyer added a careers page and gated downloads; the provider flagged the added time but went quiet for three weeks and delivered roughly 60% of the site three weeks late — after the show.
Allocation: provider 65% · buyer 35%. Split — roughly a third of the fee releases to the provider for the usable partial build.
[MC-3] The deadline was binding and its purpose was known; missing it is serious provider fault. [MC-7] Three weeks of silence past a hard deadline compounds it. [MC-4] The buyer's mid-project scope additions caused part of the slip and carry part of the fault. [MC-8, SD-4] The buyer retains a deployed, usable partial site and pays for the value retained.
Ruling III — The ghost
A copywriting provider confirmed kickoff, reported progress "going great" at week two, then stopped responding entirely. Nothing was ever delivered across six weeks and three follow-ups.
Allocation: provider 100% · buyer 0%. Full refund.
[MC-6] Non-delivery without communication is the most serious provider fault and yields a full refund. [MC-7] The misrepresented progress report and sustained unresponsiveness compound it. [MC-9] The log is decisive.
Accountability
Recourse is designed to be the accountable second generation of dispute resolution — after token-voting oracles publicly failed under money pressure. Accountability here is structural, not rhetorical:
① Published doctrine. The rules above are public before any dispute arises. ② Published reasoning. Every ruling carries a reasoning trace citing the rules applied. ③ Conflict screening. Recourse is never owned by the platforms it judges; the fee is never a percentage of the disputed value. ④ Liability-backed rulings. On the backed tier, rulings carry a capped, insurance-backed warranty — no support team or token oracle can offer that. ⑤ Published error rates as the corpus grows.