Figure · FIG:foundational/financial-ratio-analysis-workflow

A controlled financial-ratio-analysis workflow

A model proposed and deterministically rendered workflow from a stated analytical question to a bounded conclusion.

Updated Aug 7, 2026 Review due Nov 7, 2026
A defensible ratio analysis preserves a controlled chain from question through aligned evidence and calculation to a qualified conclusion.
Detailed visual description

The workflow begins with the question to be answered. It then controls entity, period, definition, basis, and provenance; declares the formula convention; computes and compares with retained precision; and ends with a conclusion limited to what the controlled evidence supports.

Review apparatus

Visual provenance and review

Technique
ai specified svg
Role
explanatory
Review state
revised
Model source
openai · openai:gpt-5.6-terra
Rights
CC0-1.0 · Project-authored deterministic rendering of a reviewed model-assisted specification

Review notes

  • The real OpenAI call correctly emphasized that calculation is one link in a source-to-claim chain and that a ratio is not a verdict.
  • The raw proposal used an unsupported nodes/edges/callouts grammar, so it failed the deterministic renderer gate.
  • The accepted specification translates the same five instructional moves into the registered flow-diagram grammar and removes no material boundary.
Retained model-call description

scripts/figures/llm_spec.py constrained proposal using the complete C:financial-ratio-analysis record; the retained response contains the learning question, rationale, proposed specification, and fact-check list.