Model Selection and Cross-Validation: What P/E, DCF, and SOTP Each Handle

Measuring a bank or a REIT with a P/E often produces absurd conclusions—because their value drivers are entirely different. The engine has a "model selection matrix" that picks a suitable primary valuation model by business type, then cross-validates from several angles.

Different Businesses, Different Rulers

Using the wrong model is one of the most common hard flaws in amateur valuation. The engine first determines "what kind of business this is," then decides the primary model and the required cross-checks.

• Software / compounding growth: primarily EPS-IRR (forward P/E plus return).
• Capital-light / cyclical: primarily DCF / free cash flow.
• Banks and other financials: often P/B-ROE (price-to-book and return on equity).
• REITs (real estate investment trusts): FFO/AFFO, NAV / cap rate, rather than P/E.
• Resources / commodities: reserve NAV, normalized earnings.

Cross-Validation: Do Not Trust a Single Number

The engine requires several independent lenses to confirm "expensive / fair / cheap"—intrinsic value, relative valuation, reverse-implied expectations, moat quality, scenarios and risk, and so on. A high-confidence conclusion requires multiple independent lenses to agree. A single metric saying "cheap" is nowhere near enough.

“One angle may deceive you; only when four independent angles point to the same conclusion is it worth overweighting.”— Institutional Research Principle

How to Adjudicate Model Conflicts

When P/E-IRR says "fair" but DCF says "expensive," the engine does not simply average them; it uses a "model-conflict ledger" to probe the source of the disagreement: is a capex cycle distorting current cash flow? Is it a terminal-multiple disagreement? Or is there a genuine valuation veto signal? Put the conflict on the table and find the cause, rather than splitting the difference.

How It Shows Up in the Report

The report states which primary model was used, which cross-checks were done, and whether any model conflict exists. This explains why the key metrics emphasized in the report differ across different types of company.