01 · WHAT A RUN DOES
Five stages, then a verdict.
Every report, at every price, runs the same five stages in the same order against the material you upload and nothing else. No interviews are conducted, no opinions are collected, and no human vote is added to the result.
STAGE 01
Script analysis
The screenplay is read for structure, genre signature, character architecture, tonal register, and the commercial hooks and risks that follow from them.
STAGE 02
Persona panel
A sample of 200 personas is drawn from a fixed 400-persona vault, stratified by country and up-weighted toward the genre preferences relevant to the film, and screens the material. The vault is identical for every client. Excitement, will-see intent, and demographic spread come back as ranges.
STAGE 03
Critic projection
Six critical lenses project likely reception of the submitted material, not of a finished film, and are reported with a stated confidence label.
STAGE 04
Distributor matching
Candidate buyers are ranked by match probability against a classified database of released titles, with deal shape and fit reasoning attached to each.
STAGE 05
Monte Carlo waterfall
The full recoupment stack is simulated 5,000 times for every weighted scenario: each matched buyer, plus the residual and no-deal paths. The total for a project is that number times its scenarios.
Five stages, one verdict, every driver named.
DECLINEPASSWEAK CONSIDERCONSIDERRECOMMENDSTRONG RECOMMEND
02 · VERSION STAMPS
What S6-MATCH-8 designates.
Every report carries an engine stamp on its face. It identifies exactly which version of the engine produced that verdict, so a report read a year from now can be placed in time rather than assumed current.
THE STAMP COMPOSES FROM FOUR PARTS
PIPELINE
The five stages above and the order they run in.
VERDICT RULESET
How stage outputs resolve into a band on the ladder.
P&A SCHEMA
The tier-aware prints-and-advertising model used in the waterfall.
ERA HONESTY
Comparable releases are read against their own release era, not flattened into a single market.
When any of those change, the stamp changes. Old reports keep their old stamp. A verdict is always readable against the engine that produced it.
03 · CALIBRATION PRACTICE
Backtested against films that already happened.
The engine is calibrated by running it against released films and comparing the verdict it would have issued to what the film actually did. Those fixtures are published in the calibration ledger, in full, including the ones the engine got wrong.
Entries are permanent. A published fixture is never deleted, never quietly revised, and never reframed after the fact. When a miss occurs it is annotated with what the engine got wrong and what changed in the methodology as a result. The DECLINE end of the ladder is fully specified in the ruleset and not yet demonstrated on a public fixture; when it is, it will be published at the same size as the hits.
04 · CONFIDENTIALITY
Your material is used for your report. Nothing else.
No person at FrameLore reads the screenplay; the engine does. Running the analysis means the text is processed by the AI model providers Oracle is built on, under commercial terms that prohibit training on it. Beyond that, your material is used to produce your report and nothing else: disclosed to no other party, and never used to produce another client's report.
On the platform side, project data is isolated at the database layer with row-level security, so one client's material is structurally unreachable from another client's session. Stored files are served only through expiring signed URLs, never from a public path. Reports are delivered only to the account that ran them, and custody of the finished report is yours.
The full handling record, including what every provider holds and how long anything is kept, is on the privacy and terms page.
05 · WHAT ORACLE IS NOT
Stated plainly.
Not investment advice. Oracle reports are analytical opinions produced by a calibrated model. They are not an offer, a solicitation, or a recommendation to buy, sell, or finance any security or project.
Not a guarantee. A band is a band. Ranges are stated as ranges precisely because outcomes are uncertain, and a favorable verdict describes calibrated odds, never a promised result.
Not for sale at any price. Pricing determines depth, custody, and service level. It does not, and structurally cannot, determine the verdict. There is one engine and one answer for a given piece of material.