Performance Estimate confidence scale

Status and interpretation

Live Index Performance Estimate — not reported box office. This version 1.0.0 protocol defines what Live Index will accept before publishing an estimate; it is not a claim that observations have already been collected or that the protocol has been validated against actual settlements. No estimate rows are released with this document. A method can establish a reproducible decision rule without establishing any particular show’s attendance, gross, sellout status or promoter profit.

Confidence is evidence completeness

High, medium and low describe the evidence behind a proposed estimate, not a statistical probability that an exact gross is correct. No percentage confidence is assigned without a calibrated error model. The weakest indispensable input sets the ceiling: strong price evidence cannot compensate for an unknown concert configuration, and a precise timestamp cannot prove how many unavailable seats have sold. The scale is an editorial protocol awaiting reconciliation data.

Review classes

• High: all required inputs are dated and primary; sellable configuration and inventory states can be reconciled; price scope is explicit; no material unresolved contradiction remains. The band and assumptions still need editorial review.
• Medium: all required source categories exist, but a bounded inventory or pricing ambiguity widens the band. Record the ambiguity, the evidence delimiting it and why it does not prevent the proposed comparison.
• Low: the mandatory source categories are present but substantial unresolved limitations remain. This is a review state, not permission to release a row. Withhold when the uncertainty cannot be bounded.
• Ineligible: a mandatory source category, timestamp or event identity is missing. Do not label an unsourced row low confidence as a way around the release gate.

What cannot improve the class

Artist popularity, a large social following, press enthusiasm or a busy resale page does not increase confidence in a show estimate. Repeating the same ticketing observation is not independent corroboration. A promoter’s sellout announcement supports the attributed announcement, not every unstated count or price. Separate evidence quality from demand: a poorly observed blockbuster can be less estimable than a modest show with a complete manifest.

Review record

Store the class, reviewer’s reasoning, each required source, the observed date and method version alongside the proposed band. Confidence can change after new evidence, but the prior observation retains its original classification. On reconciliation, compare errors by class only when there are enough comparable reports; do not imply that the proposed classes already predict measured accuracy. Readers should be able to trace why a row was held without reconstructing a private scoring formula.

Release gate

Publication requires a dated primary ticket-inventory observation, a venue configuration source, primary price-band evidence, promoter and ticketing evidence, a confidence classification and a record-specific limitations statement. A later reading is a new observation, not an overwrite of the previous state. The pilot remains withheld from search and the sitemap until ten observations individually pass these requirements. Downloads are offered only for real released rows, never for an empty demonstration or synthetic sample. Inventory protocol, capacity protocol and price-band protocol describe different inputs; none substitutes for the others.