Free access / Consumer evidence
Free-signal archives and a usable-item denominator
Keep the whole observed sequence while defining a narrower usable count before computing cost per item.
Population before outcomes
Set the observation window and inclusion rules before reviewing favourable results. Keep every original item encountered under that access route, including commentary, duplicates, revisions, cancellations and incomplete messages. The population is not automatically every message a supplier has ever published. State the archive coverage and gaps. A useful small observational record can be limited without pretending to be a full history. Do not convert a selected gallery of winners into a complete denominator or silently infer missing losing calls from promotional accuracy claims.
Usable for what
Define usefulness against one task. An economic observation may be usable for research, while an incomplete trade instruction may fail an execution-study definition. Keep those definitions separate instead of averaging unlike objects into one score. A received message that lacks a required identifier can stay in the total population without joining the usable count. Establish the rule before observing outcomes, then preserve changes to the rule as new versions. Otherwise, a favourable move can tempt the reviewer to relabel a previously incomplete message as complete.
Zero is not a rating
When the usable count is zero, cost per usable item is undefined. That remains true if the total bill is zero. Do not substitute zero cost, infinite poor quality or a star rating. The total-cost record still tells the reader what the workflow would cost; the usability record tells the reader why division is unavailable. A positive usable count permits arithmetic but not a performance conclusion. A very cheap informational item can be misleading, and an expensive one can be useful for a different task without being investable.
Export a reproducible conclusion
Retain the original definitions, source window, item statuses and user-entered costs with the calculation. State whether costs were actual documented expenses or invented assumptions. A machine-readable export should repeat these limits, not add unsupported financial ranks. The original workbench stores evidence state separately from provider quality, and the cost worksheet labels its output as a scenario. Together they can preserve a purchasing investigation. They cannot certify a complete vendor archive, actual execution, exact winning counts or a future profitable strategy.
Completed fictional consumer case
Fictional case: a five-item observation packet contains two complete timely instructions, one late message, one cancellation and one general lesson. All five remain in the packet; only two qualify under the invented execution-study definition. A hypothetical $10 bill gives $5 per usable instruction. If the two complete messages are absent in another period, the count is zero and the ratio becomes undefined. Neither period contains outcome prices or account returns. The completed record demonstrates denominator discipline, not a claim that any real publisher achieved a particular accuracy or cost.
| Record | Evidence object | Bounded finding |
|---|---|---|
| All retained items | 5 invented records | No deletion after classification |
| Usable count | 2 | Declared instruction rule |
| Cost scenario | Invented $10 / 2 | $5, not a quality score |

Primary references and their limits
Public documents reviewed 13 October 2026. A link supports its stated mechanism or access description, not a completed account trial or this desk's invented case.