Learn

How the calculation works

Intrinsic value estimates are demonstrated with Alphabet as a worked example. Every company page runs the same logic on its own numbers; the methodology page covers the backtest and limits.

How the recommended model is chosen

Financials (banks, insurers): dividend model when the payout is at least 35% of earnings, otherwise earnings-based DCF.
Dividend payers whose dividends drive value: dividend discount model, unless the yield is far below what the model implies.
Everyone else: free-cash-flow DCF, falling back to the earnings-based DCF when cash flow is unusable or looks inflated.
Depressed cash flow (heavy investment): a three-year average cash flow is used, unless the latest cash flow is in line with net income (0.8–1.1×), in which case it is a real run rate and is kept. If an EV/EBITDA value at 12× lands closer to the price, that replaces the average-based value.
High-growth companies (analyst growth of 15% or more) priced far above their cash-flow value: peer EV/Sales, using the median multiple of similar-growth peers in the same sector. Revenue multiples are skipped when the company's own reported revenue growth is below 8% a year, so a forecast alone cannot justify one.
Technology and communication companies whose cash-flow models land far below the price: EV/Sales scaled by growth plus margin (Rule of 40), under the same reported-growth condition.
Memory and storage makers (Micron, Sandisk, Western Digital, Seagate): steady-state FCFF at the long-run growth rate, because peak-cycle cash flow should not be extrapolated. Always shown as low confidence.
Every company page lists the other models and why each was not used.

Worked example

Alphabet Inc. (Class A) (GOOGL)

Using TTM through 30 Sep 2026 from EODHD quarterly statements (data downloaded 05 Nov 2026 18:28).

What the app recommends: EV/EBITDA (12x)

This is the model the company page would use, chosen by the rules above. Confidence: Low. The current price is $282.57.

Margin of safety vs base case

Bear: $267.65, margin of safety -5.6%
Base: $319.18, margin of safety 11.5%
Bull: $370.72, margin of safety 23.8%

Bear and bull are the model's low and high cases. For DCF models growth moves by the company's own revenue-growth volatility (1.5 to 6 points, 2 without enough history), terminal growth ?1 point and the discount rate ?1 point; for multiple-based models the multiple is lowered and raised. They are sensitivity cases, not forecasts. Margin of safety is (intrinsic value ? price) / intrinsic value.

Other models as cross-checks

These are not the recommendation. Large gaps between them are normal and show why one model is chosen and confidence is shown.

$35.97 Multi-stage DCF Low when heavy investment holds down reported free cash flow.

$31.87 FCFF proxy No growth stage, so it undervalues companies that are still growing.

$21.91 FCFE proxy Treats free cash flow as shareholder cash and ignores net debt.

$319.18 EV/EBITDA at 12? Ignores capital spending, so it can flatter capital-heavy businesses.

$599.73 EPS-based Gordon proxy Assumes all earnings are paid out, so it overstates value.

Assumptions for the cross-checks: high growth 7.00%, discount rate 10.40%, terminal growth 2.90%. The reported free cash flow is a proxy, not independently reconstructed FCFF or FCFE.

The models

Multi-stage DCF

EV = Σ FCFₜ / (1 + r)ᵗ + TV / (1 + r)ⁿ

Five years of high growth, then slower terminal growth. Cash is added and debt subtracted, then divided by shares.

Used for: most companies with positive free cash flow

FCFF / FCFE proxies

V = FCF × (1 + g) / (r − g)

A single stage with no growth boost. The FCFF proxy adjusts for net debt; the FCFE proxy does not subtract debt a second time. Both use reported free cash flow, so they differ only by net debt.

Used for: cross-checks and the steady-state model

EV / EBITDA

V = (EBITDA × 12 − debt + cash) / shares

Values operating earnings at a fixed 12× multiple. Capital spending, for example on data centres and AI, depresses reported free cash flow, while EBITDA ignores its timing. A fixed multiple is a judgement, not a measurement, and can flatter capital-heavy businesses. It is not a peer-derived comparable-company valuation.

Used for: companies whose cash flow is depressed by investment

EPS-based Gordon proxy

V = EPS × (1 + g) / (r − g)

A rough reference that treats earnings as if they were all paid out. It can overstate value for companies that retain earnings.

Used for: reference only, never the recommendation

Revenue multiples

V = (Revenue × EV/Sales − debt + cash) / shares

Peer EV/Sales applies the median multiple of similar-growth, same-sector companies. Rule of 40 scales the multiple by growth plus EBITDA margin. Both need reported revenue growth of at least 8% a year and are rejected when the result is far from the price.

Used for: high-growth and technology companies

Earnings and steady-state models

V = f(net income after reinvestment)

Earnings-based DCF is used when free cash flow is unusable or inflated. Steady-state FCFF holds current cash flow at long-run growth for memory and storage makers, whose cash flow peaks and crashes with prices.

Used for: unreliable cash flow, memory cycle

Residual income

Needs shareholders' book value, which is not in the saved TTM snapshot, so it is described here but not calculated.

Used for: not used

Growth and discount inputs

Growth blends the company's own revenue growth with the sector default, and with analyst estimates when available. Discount and terminal rates come from the sector assumption; a per-company beta adjustment was tested and made results worse, so it is not used. Each company page shows a sensitivity table across growth and discount-rate combinations.

Confidence and track record

A value under 40% or over 250% of the price is flagged low confidence; the homepage hides those from the Undervalued and Overvalued lists by default.
Low confidence is also set for cyclical sectors and companies, and when free cash flow is more than 2.0× net income (it is probably inflated).
The published backtest is on the methodology page. The edge is modest and the overvalued label is the weaker signal.
Caveats: survivorship bias, a mostly rising market, overlapping years, and the sector-default growth rate rather than the blended one.
A data health page (staff only) lists companies with stale, missing or inconsistent data.