Methods and accountability
What this release knows — and does not infer.
Rail-1 uses only values entered by a user in the browser. It does not depend on a tribunal-decision corpus or send tool inputs to a server.
Rail-1 data boundary
Date, rent, comparator and condition entries are processed by code delivered with the static page. Entries are not placed in a URL or sent to Rent Outcomes. The lead endpoint accepts only an email, explicit email consent and a honeypot field; it never receives comparator data. Stripe checkout and webhook endpoints remain built but are unavailable while the payments feature flag is off.
The Date & Regime Card has no address field. The Evidence Structurer permits an address because address, source and date form its deduplication key. Its list is stored in localStorage with a 30-day expiry and is removed on the next site load after expiry. The immediate-delete control removes all keys created by Rent Outcomes in that browser.
Product events use a localStorage queue containing only an event name and non-identifying product properties. The queue sends to PostHog only in builds that provide a public PostHog project key. Comparator fields and email fields are excluded from that queue.
Published decisions and MNAR
MNAR limitation: published decisions are not a random sample of rent notices, applications or tenancies.
Publication, withdrawal, settlement, route filtering and whether a field is stated can each be associated with the underlying matter or outcome. Missingness is therefore treated as potentially Missing Not At Random (MNAR), not as neutral absence.
Rail-1 presents no corpus-derived percentage. The built rail-2 preview is noindex and outside the sitemap while evidence review is incomplete; it shows a denominator for each field, a missingness matrix, the data cut and a maturity note. A legal rule is displayed separately from an empirical observation.
Condition adjustments
In published decisions, deductions were applied individually. The Condition Checklist consequently records observable facts and photo references but does not aggregate, average or transfer monetary deductions between properties or cases.
Article 14 information for public-decision data
This notice describes build-time rail-2 processing of personal data obtained from public tribunal decisions rather than directly from the person concerned. The corpus is not published in rail-1; the evidence-review surfaces remain noindex and outside the sitemap.
Public output will omit party names and limit address precision. A person may request access, correction, erasure or restriction, object to processing, or raise a concern with the Information Commissioner’s Office. An objection triggers review and, where appropriate, suppression while the issue is assessed. Article 14 timing and any applicable public-information exception will be documented before rail-2 publication.
Independence and affiliation
Rent Outcomes is not affiliated with HMCTS. The public layer uses descriptive language and symmetric terminology. It does not submit an application, conduct correspondence with the tribunal, represent a party or decide what a user should do.
Revenue model
The Comparator Ledger is free while in beta in exchange for an email address and explicit consent to occasional emails about new decisions and product updates. The PDF and CSV are still generated entirely in the browser.
The later-stage product and corpus-derived features remain outside rail-1.
Version
Methodology version BT-5 · rail-2 remains noindex pending evidence review · data boundary updated 11 August 2026 · form baselines recorded separately on the forms hub.