The audit trail¶

Every result is an envelope with four parts:

{
  "method": "dcf.fcff",
  "value":   { "value_of_operations": "1421.48…", "equity_value": "1221.48…", "...": "..." },
  "explain": {
    "formula": "Value of operations = Σ FCFF_k × DF_k + TV_N × DF_TV; ...",
    "conventions": ["timing: end of period (flow k discounted at t = k)", "..."],
    "steps": [ { "label": "Period 1: discount factor", "formula": "1 / (1 + 0.10)", "value": "0.909…" } ]
  },
  "meta": { "engine_version": "…", "method_version": "3", "input_hash": "b963…", "arithmetic": "decimal…" }
}

• value holds the results, exact, as decimal strings.
• explain.steps lists each intermediate calculation with the actual numbers substituted. An analyst can recompute any step by hand; an agent cites them instead of recomputing.
• explain.conventions states every choice that changes the number (timing, terminal value discounting). Conventions are never implicit.
• meta makes the result reproducible: the same input_hash with the same versions always yields the same value.

In Python, result.explain_markdown() renders a methodology note for reports.