财务模型公式与逻辑审计
帮你核查财务模型公式、假设与报表勾稽,形成分级问题报告,并标注待业务确认事项。
完整提示词
I want a forensic audit of an Excel, Sheets or programmatic financial model with a focus on formula correctness, assumptions, circularity, hardcodes, scenario logic, units, timing and output reliability. Main objective: > Prove that the model does, mathematically and economically, what it claims to do. This is not: - a formatting audit - a preference for a particular spreadsheet style - automatic criticism of hardcoded assumptions - a replacement for business validation ## 1. MODEL PURPOSE ## 2. INPUTS ## 3. CALCULATIONS ## 4. OUTPUTS ## 5. DATA FLOW Map: ```text source ↓ input ↓ calculation ↓ statement ↓ decision output ``` ## 6. HARD-CODE Hardcode is acceptable in designated assumption cell. ## 7. HARD-CODE IN FORMULA Review. ## 8. BROKEN LINK ## 9. EXTERNAL LINK ## 10. CIRCULAR REFERENCE ## 11. INTENTIONAL CIRCULARITY ## 12. ITERATIVE CALCULATION ## 13. FORMULA COPY ## 14. RANGE ERROR ## 15. OFF-BY-ONE ## 16. SIGN ## 17. UNIT ## 18. CURRENCY ## 19. THOUSANDS/MILLIONS ## 20. PERCENT ## 21. DATE ## 22. MONTH/YEAR ## 23. LEAP YEAR ## 24. TIMING ## 25. BEGINNING/ENDING BALANCE ## 26. CASH FLOW ## 27. BALANCE CHECK ## 28. THREE-STATEMENT ## 29. RETAINED EARNINGS ## 30. DEBT ROLL-FORWARD ## 31. PP&E ROLL-FORWARD ## 32. WORKING CAPITAL ## 33. TAX ## 34. SCENARIO SWITCH ## 35. SENSITIVITY ## 36. DATA TABLE ## 37. LOOKUP ## 38. XLOOKUP/INDEX-MATCH ## 39. ERROR SUPPRESSION IFERROR can hide broken model. ## 40. BLANK VS ZERO ## 41. NEGATIVE ## 42. CAP ## 43. FLOOR ## 44. ASSUMPTION ## 45. SOURCE ## 46. MODEL CHECK ## 47. CONTROL TOTAL ## 48. BALANCE SHEET CHECK ## 49. CASH CHECK ## 50. MODEL STRESS ## 51. EXTREME INPUT ## 52. ZERO REVENUE ## 53. NEGATIVE GROWTH ## 54. HIGH INTEREST ## 55. SENSITIVITY MONOTONICITY Where expected. ## 56. OUTPUT ## 57. CHART Do not audit visual formatting unless it changes interpretation. ## 58. VERSION ## 59. CHANGE LOG ## 60. FALSE POSITIVE RULES Hardcoded assumption is not an error if intentional and documented. Circularity is not always wrong. ## 61. EVIDENCE TIERS ```text A - formula traced and recalculated independently B - deterministic model inconsistency C - strong structural anomaly D - suspected issue requiring business confirmation E - modeling hardening ``` Materiality is the severity scale of this analysis: **HIGH** changes the overall conclusion, a decision or the liquidity outlook; **MEDIUM** changes a key metric or trend but not the conclusion; **LOW** has a limited effect and is noted for completeness. Judge it against an explicit base (the model output used for the decision (value, NPV, cash balance, covenant test)) and state that base. A finding is CONFIRMED only with evidence tier A or B; tier C is LIKELY; tier D stays NOT VERIFIED, and tier E is a SCENARIO. Never present an estimate or a scenario as a fact. ## 62. FINDING FORMAT ```text ID: Materiality: Sheet/module: Cell/formula: Status: Evidence tier: Expected logic: Actual logic: Impact: Affected outputs: Evidence: Fix: Independent verification: ``` ## 63. MODEL CHECK MATRIX | Area | Control | Pass | Notes | |---|---|---|---| ## 64. SECOND PASS Change key inputs and verify: - outputs respond correctly - no broken hardcodes - statements balance - scenarios remain internally consistent ## 65. FINAL QUALITY GATE Confirm: - formulas - units - timing - assumptions - statements - debt - cash - scenarios - controls - outputs ## 66. OUTPUT `FINANCIAL_MODEL_AUDIT.md` # FINAL RULE A model is not correct because it has no Excel error. A model is correct only when: ```text formula + economic logic + timing + units + statement relationships ``` produce a consistent result.
分类与标签
财务理财、财务模型、公式审计、情景测试、报表勾稽、证据分级