Financial Modeling Workflow
This skill provides a structured workflow for building financial models using the financial-modeler agent.
Workflow Overview
The financial modeling workflow follows 6 phases:
| Phase |
Name |
Description |
| 1 |
Scoping |
Define model purpose and requirements |
| 2 |
Architecture |
Design model structure |
| 3 |
Assumptions |
Document all model inputs |
| 4 |
Build |
Construct model with formulas |
| 5 |
Validation |
Test and verify model |
| 6 |
Documentation |
Prepare user guide and audit trail |
Phase 1: Scoping
MANDATORY: Define model purpose before building
Questions to Answer
| Question |
Purpose |
| What decision does this model support? |
Ensures relevance |
| What type of model is needed? |
Sets approach |
| What level of detail is required? |
Scopes complexity |
| Who will use the model? |
Tailors design |
| What scenarios are needed? |
Plans flexibility |
Model Type Selection
| Type |
Use Case |
Key Outputs |
| DCF |
Intrinsic valuation |
Enterprise Value, Equity Value |
| Trading Comps |
Relative valuation |
Implied multiples |
| LBO |
PE acquisition |
IRR, MOIC, sources/uses |
| M&A |
Strategic acquisition |
Accretion/dilution, synergies |
| Three-Statement |
Operating projections |
IS, BS, CF integrated |
Blocker Check
If ANY of these are unclear, STOP and ask:
- Model purpose
- Model type
- Key assumptions (WACC, growth, multiples)
- Scenario requirements
Phase 2: Architecture
MANDATORY: Design structure before building
Architecture Principles
| Principle |
Description |
| Input separation |
All inputs in dedicated section |
| No hardcoding |
No numbers in formulas |
| Consistent formulas |
Same formula across row |
| Error checks |
Validation on each sheet |
| Clear flow |
Logical left-to-right, top-to-bottom |
Standard Model Structure
| Section |
Contents |
| Cover |
Model name, version, date, author |
| Inputs |
All assumptions in one place |
| Historical |
Historical financial data |
| Projections |
Projected financials |
| Valuation |
DCF, comps, or transaction calcs |
| Sensitivity |
Sensitivity tables |
| Output |
Summary and key metrics |
| Checks |
Error checking |
Phase 3: Assumptions
MANDATORY: Document ALL assumptions with sources
Required Assumption Categories
| Category |
Examples |
| Operating |
Revenue growth, margins, CapEx |
| Valuation |
WACC, terminal growth, multiples |
| Transaction |
Entry price, financing, fees |
| Timing |
Projection period, exit year |
Assumption Documentation Standard
| Element |
Requirement |
| Assumption name |
Clear description |
| Value |
Base case value |
| Source |
Where value came from |
| Sensitivity |
Range for testing |
Blocker: WACC Components
MUST get user input on:
- Risk-free rate source
- Equity risk premium source
- Beta source and adjustment
- Capital structure assumption
Phase 4: Build
Dispatch to specialist with full context
Agent Dispatch
Task tool:
subagent_type: "financial-modeler"
model: "opus"
prompt: |
Build financial model per these specifications:
**Purpose**: [decision supported]
**Model Type**: [DCF/LBO/M&A/Operating]
**Architecture**:
[From Phase 2]
**Assumptions**:
[From Phase 3]
**Historical Data**:
[Attach historical financials]
**Required Output**:
- Model structure per architecture
- All assumptions in input section
- Sensitivity analysis
- Scenario analysis (base/upside/downside)
- Error checks
Required Output Elements
| Element |
Requirement |
| Model Summary |
Key outputs and conclusions |
| Assumptions |
Complete input section |
| Model Structure |
Per architecture design |
| Key Outputs |
Valuation, metrics, returns |
| Sensitivity |
2x2 tables for key drivers |
| Validation |
Error checks passing |
Phase 5: Validation
MANDATORY: Test model before delivery
Validation Tests
| Test |
Description |
| Balance Sheet |
Assets = Liabilities + Equity |
| Cash Flow |
CF reconciles to B/S cash |
| Circular Control |
Circulars iterate and converge |
| Formula Audit |
No hardcoded values in calcs |
| Reasonableness |
Results within expected range |
Sensitivity Testing
| Test |
Purpose |
| WACC sensitivity |
Valuation impact of discount rate |
| Growth sensitivity |
Valuation impact of growth |
| Margin sensitivity |
Valuation impact of profitability |
| Exit multiple sensitivity |
LBO return sensitivity |
Anti-Rationalization
| Rationalization |
Why It's WRONG |
Required Action |
| "Model foots, it's correct" |
Footing ≠ correct logic |
VALIDATE methodology |
| "Sensitivity shows expected range" |
Range needs to be appropriate |
VERIFY ranges sensible |
| "Used industry WACC" |
WACC must be calculated |
CALCULATE and document |
Phase 6: Documentation
MANDATORY: Document for future users
Documentation Checklist
| Element |
Status |
| Model summary |
Required |
| Assumption register |
Required |
| Methodology description |
Required |
| User instructions |
Required |
| Error check explanations |
Required |
| Version history |
Required |
Output Format
See shared-patterns/execution-report.md for base metrics.
Modeling-Specific Metrics:
- model_tabs: N
- assumptions_documented: N
- scenarios: N
- sensitivity_tests: N
- validation_tests: N passed
Pressure Resistance
See shared-patterns/pressure-resistance.md for universal pressures.
Modeling-Specific Pressures
| Pressure Type |
Request |
Agent Response |
| "Just give me a valuation" |
"Valuations require documented methodology. I'll build proper model." |
|
| "Use 10% WACC" |
"WACC must be calculated from components. I'll show the build-up." |
|
| "Skip sensitivity" |
"Sensitivity is required for decision support. I'll include key drivers." |
|
| "Copy last model" |
"Each model needs fresh design. Prior models are reference, not template." |
|
Anti-Rationalization Table
See shared-patterns/anti-rationalization.md for universal anti-rationalizations.
Modeling-Specific Anti-Rationalizations
| Rationalization |
Why It's WRONG |
Required Action |
| "Quick model doesn't need structure" |
ALL models need structure |
DESIGN architecture |
| "Terminal value is just a plug" |
TV often 60%+ of value |
CALCULATE properly |
| "Industry multiple is obvious" |
Multiples need source |
CITE and date source |
| "Circular doesn't matter" |
Circulars affect results |
HANDLE explicitly |
Execution Report
Upon completion, report:
| Metric |
Value |
| Duration |
Xm Ys |
| Model Tabs |
N |
| Assumptions |
N documented |
| Scenarios |
N |
| Sensitivity Tests |
N |
| Validation Tests |
N/N passed |
| Result |
COMPLETE/PARTIAL |
Quality Indicators
| Indicator |
Status |
| All inputs in input section |
YES/NO |
| No hardcoded values |
YES/NO |
| Error checks pass |
YES/NO |
| Sensitivity included |
YES/NO |
| Methodology documented |
YES/NO |
majiayu000