Define outcome and cost measures for later evaluation

[nothingnesses]

This issue records a proposal. Implementation requires a selected work brief and separate execution approval.

Background and problem

D-077 asks whether each part justifies its cost:

• Tool.
• Workflow.
• Method.

Output byte limits do not measure:

• Correct delivery.
• Human waits.
• Model cost.

The library checkpoint completed the agreed real-project reassessment. It found useful defects, but manual coordination imposed substantial cost.

Proposal

Define a bounded measurement proposal for later outcome and cost evaluation. Separate the contribution of each part:

• Tool.
• Workflow.
• Method.

Include:

• Correctly met approved requirements.
• Active time.
• Waits.
• Pauses.
• Tokens.
• Model cost.
• Prompt versions.

Scope and exclusions

Exclude:

• A new library checkpoint.
• Retrospective enrolment of the completed checkpoint.
• Claims of controller value from manual operation.

This proposal introduces none of the following:

• Round logs.
• Personal provider restrictions.
• A repeated benchmark.
  Broader statistical work remains deferred until the pilot report.

Acceptance criteria

• Outcomes receive independent assessment against approved requirements.
• Missing cost data and coordinator effort remain explicit.
• Measures expose shortcuts that inflate activity without correct delivery.
• The proposal distinguishes observed defects from unknown field escapes.
• A later execution needs an approved protocol and scope.

Alternatives and open questions

Use a small descriptive case study before a larger benchmark. This follows Ground decisions in evidence without treating one project's results as general performance.

References

• Decision report, D-077.
• completed trial and limitations.
• measurement methods.
• D-077.