Diagnostics & Auditability
Review the capstone workflow as evidence, not just the forecast as display
Chapter 7 asks whether the result is traceable, reproducible, and review-ready. These checks summarize whether the authoritative bundle, assumptions, outputs, and valuation inputs are still behaving like one controlled system.
Run Checks
Current diagnostic board
These checks are designed around the final capstone integration contract you defined.
Chapter 3 horizon drives the final bundle
PASSMaster horizon contains 12 forecast quarters from FY2026-Q4 to FY2029-Q3.
Base forecast rows stay identical to Chapter 3
PASSDashboard assumption overrides are active — the base case intentionally departs from the live Chapter-3 baseline and reflects the user's slider deltas layered on top of live data.
Official scenarios are limited to Base, Bear, and Bull
PASSLiquidity stress is excluded from the final dashboard contract.
Scenario sensitivity uses only the two live exogenous drivers
PASSWTI crude and CPI (3M annualized) are the only exposed final sensitivity drivers.
Dashboard valuation input stays isolated from operating outputs
PASSDiluted shares outstanding is used only in the target-price conversion layer, not in revenue, EBIT, FCFF, or scenario generation.
Target price layer is ready for review
PASSShares: 62.40m (live_fmp). Peer median FCFF yield: 3.28%.
Notebook and Flask use the same authoritative file
PASSBoth surfaces read capstone_project/outputs/authoritative/latest_output.json instead of recomputing separate models.
Dashboard assumption overrides are explicit and persisted
PASS1 non-default dashboard assumption override(s) are currently active.
Macro vintage within staleness tolerance
PASSOil cache fresh=True (≤30m), CPI fresh=True (≤35d), fundamentals fresh=True (≤36h), sensitivity fresh=True.
Sensitivity regression has sufficient data and significance
PASSoil→cogs_yoy (applied): n=36, r²=0.523, p=0.0000, fallback=False. oil→gross_margin (diag): n=40, r²=0.094, p=0.3128. oil→gross_profit_yoy (diag): n=36, r²=0.604, p=0.0000. Horse-race winner by R²: gross_profit_yoy. CPI→DIO(2Q lag): n=38, r²=0.809, p=0.0008, fallback=False.
PVM foundation is sourced from live data (FMP + SEC EDGAR)
PASSRevenue segments fetched at 2026-04-18T14:50:42 (55 rows); store counts latest quarter 2025-12-27 (72 rows). Driver table mtime: 2026-04-20T14:01:50+00:00.
Balance-sheet ties for every forecast period (A = L + E)
PASSAll 12 periods tie to the penny via the non-current-liability plug (max |gap| $0.0000m).
BS plug drift stays within tolerance (model consistency)
WATCHMax plug drift from anchor: $189.3m (17.46% of anchor). Thresholds: pass ≤15.0% watch ≤25.0%. The plug absorbs deferred-tax, short-term-investment, and other non-modelled BS movements; a tight drift means the model is consistent.
Cash flow reconciles (CFO + CFI + CFF = ΔCash)
PASSPeriods checked: 12. Max |gap|: $0.000m (tolerance $0.5m).
PP&E roll-forward reconciles (PPE_open + CapEx − Depreciation = PPE_close)
PASSRolled across 12 forecast periods from the latest-actual PP&E anchor.
Equity roll-forward reconciles (NI − Div − Buyback + SBC)
PASSRolled across 12 forecast periods.
Debt roll-forward reconciles (Debt_open + NetIssuance)
PASSRolled across 12 forecast periods.
Historical BS matches raw FMP data (upstream data-quality check)
PASSAll 14 line items match FMP raw BS within 2.0%. FMP filing date: 2025-12-27.
Manifest
Run metadata
Artifacts
Current file lineage
outputs/authoritative/latest_output.json
outputs/diagnostics/latest_diagnostics.json
run_metadata/state.json
assumptions/peer_fcff_comps.csv
notebook/final_capstone_model.ipynb
Source Lineage
Chapter ownership map
Governance Notes
Why the bundle stays reviewable
Chapter 3 is the operational forecast backbone for the final capstone bundle.
Chapter 6 feeds exogenous sensitivity and translated assumption paths, not the final absolute revenue and EBIT rows.
Dashboard and notebook are downstream views of the same authoritative output file.