feat: Fox Factory worked example — reconcile + forecast + divestiture on real SEC data

[JeffBrines]

What

A real public-company worked example (examples/foxfactory/) that runs the openfpa engine against Fox Factory Holding Corp. (NASDAQ: FOXF), pulled live from SEC EDGAR. Built as a credibility proof — Ridgeline is synthetic; this is real, audited, messy numbers.

Three phases:

• Phase A — reconciliation. Driven with Fox's actual reported drivers, the engine reproduces reported revenue, gross profit, Adjusted EBITDA, D&A, capex, and the working-capital cash mechanic — to the dollar for FY2024 and FY2025. The $557M FY2025 goodwill impairment and discrete tax benefits are shown as an explicit, documented bridge (the lean engine models the operating business, not one-time non-cash charges).
• Phase B — forecast. Segment-level (PVG / AAG / SSG → consolidated) FY2026–FY2027, anchored to the reported Q1 FY2026 print (+3.8% YoY, reproducible from committed data).
• Phase C — capital allocation. A labeled Marucci-divestiture sensitivity across sale timing and proceeds (default $300M, or EV/EBITDA exit multiples), showing FCF and leverage impact.

Engine changes (general, backward-compatible)

• pyfpa/models/cashflow.py: expense D&A in the P&L (EBIT = EBITDA − D&A) and add it back in operating cash flow — cash-neutral, correct double-entry. Previously D&A was added back without being expensed, which inflated OCF. Invisible for Ridgeline (D&A=0); the golden-snapshot and Ridgeline suites pass unchanged.
• pyfpa/config/schemas.py: optional da_monthly / capex_monthly (default 0.0).
• New pyfpa/analysis/: segments.py (net sales + Adjusted EBITDA per ASU 2023-07 — faithful to FOXF disclosure, not gross profit), divestiture.py (immutable carve-out scenario), reconcile.py (variance helper). All exported and unit-tested.

Self-extension (the hero thesis)

Fox reports segment Adjusted EBITDA, not gross profit, so fpa-learn-business generates a bespoke segment-rollup skill to fit (examples/foxfactory/skills/generated/). Committed business profile grounds it in the filings.

Provenance

pull_edgar.py fetches every figure from EDGAR's XBRL API + the segment-footnote HTML — zero hand-keyed financials. data/SOURCES.md traces each number to a filing/accession.

Test plan

• python3 -m pytest → 92 passing (incl. test_foxf_reconciliation.py regression guarding the to-the-dollar tie, and test_cashflow_fcf.py proving D&A is cash-neutral).
• python3 examples/foxfactory/pull_edgar.py && python3 examples/foxfactory/run_foxf.py → regenerates output/ (reconciliation, forecast briefing, divestiture, Excel).

Notes

• Reviewed by an independent agent pass: no correctness/accounting defects; every output number verified against the committed data.
• Spec + plan: docs/superpowers/specs/2026-06-07-foxfactory-worked-example-design.md, docs/superpowers/plans/2026-06-07-foxfactory-worked-example.md.
• The load_actuals/load_segments loaders from the plan were folded into foxf_model.py (direct CSV access) rather than added to the public API — kept the engine surface lean. Flag if you'd prefer them exported.

🤖 Generated with Claude Code