How Stock Valuation Studio Works

Last updated October 30, 2026·10 min read

Stock Valuation Studio computes thirty-one valuation models for a company and reduces them to one answer: a Fair Value, a verdict on the current price, a margin of safety, and the price at which the stock offers that margin. This page states the method exactly as the engine applies it: the routing that sends a fast-growing company to a forward discounted cash flow and a cyclical company to a five-year average base, the fixed order that chooses the headline figure behind that routing, the construction of the discounted cash flow model, the dividend rule, the margin and verdict conventions, the confidence grade every headline carries, and the testing the engine passes before any of it reaches this page. The engine runs in your browser on data drawn from the company's own filings, covered on the companion page, Where the Studio's Data Comes From.

One Headline, Chosen by a Fixed Order

Thirty-one models produce thirty-one opinions, and a reader needs one figure to act on. The Studio chooses it by a published order, the same for every company:

• First, the discounted cash flow value on the base in force, when the cash flow base can be stated.
• Second, the multiples value, when trailing earnings per share are positive. For a real estate investment trust, this step runs on funds from operations instead, through the model the Studio prints as Multiples (P/FFO), whenever funds from operations can be stated.
• Third, the forward-estimate model, when it prints a value.
• Fourth, the revenue-build discounted cash flow, when it prints a value.
• Fifth, the median of every model that prints a value. When exactly two models print, the headline is the higher-ranked of the two by this same order, and the label says which model it is. It is never the midpoint of the two.
• Last, and only when no model can compute at all, the Studio prints No Reliable Value rather than a number it cannot support.

Three disciplines hold the order honest. The steps trigger on whether the inputs are sound, never on how far a model's answer sits from the price. A model that yields no value falls through to the next step instead of blocking the answer. A result at or below zero, or one that is not a finite number, is never used as a value anywhere in the Studio: the model reports its state and its reason instead. Every headline carries its basis label, naming the model and, for a cash flow headline, the base in force. A reader who prefers a single model can pin any model that prints a value as the headline, and the verdict, the margin, and the buy price re-key to that model, with the automatic headline still shown beside it.

Ahead of the order, two routings apply, and a sector rule stands in front of both: a bank, an insurer, or a real estate investment trust keeps the sector treatment below and is never re-routed. A company whose revenue grows faster than 15 percent a year is priced first on a forward discounted cash flow: five years built from the consensus revenue path (growth capped at 15 percent), a margin traveling to a target read from the forward earnings, the trailing record, or the industry group, and reinvestment measured from the company's own sales-to-capital history with the group's as fallback. A company in a cyclical industry (semiconductors, energy, materials, autos) that is not growing at that pace is priced first on the discounted cash flow over the average free cash flow to firm of the last five years. A routed construction that cannot print falls through to the order, and the basis label names what printed: "Forward DCF" or "DCF (5-yr avg)" alongside the labels above. Every headline also carries a confidence grade, High, Medium, or Low, scored on the agreement of the headline-grade models around their own middle value, the share of a cash flow headline's value resting on the terminal year, and the representativeness of the trailing cash flow against its five-year average and the growth the current price implies. A Low grade names its driver.

The Discounted Cash Flow Model

The cash flow base is the company's free cash flow to firm per share over the trailing twelve months. Free cash flow to firm is reported free cash flow plus after-tax net interest: operating cash flow minus capital spending, plus interest paid less interest income, after tax. When the trailing figure cannot be stated, the base is the latest fiscal year. When the trailing base is not positive, the engine reaches for a normalized base instead: the median of the company's annual free cash flow to firm over the five filed years, used when that median is positive. The basis in force always prints with the figure.

The projection runs ten years, five years at the first-stage growth rate and five at the second. The first-stage rate is calibrated from the company's own record: the engine takes the median of four trailing per-share growth measures (free cash flow, revenue, earnings, and a bridge measure that connects the oldest filed year to the trailing twelve months) and, where an analyst forward estimate exists, takes the greater of that estimate and the trailing median, because a forward estimate may raise the rate to the forward-looking level but never cut a proven record. The result is bounded between zero and 15 percent. When neither a trailing measure nor an estimate exists, the first-stage rate is 8 percent. The second-stage rate is the midpoint fade between the first-stage rate and the perpetual rate of 3 percent, so growth glides toward its long-run level instead of falling off a cliff. Cash flows are discounted at the company's own weighted average cost of capital where that build-up can be stated from the filings, and at a stated default of 9 percent otherwise. The terminal value applies the 3 percent perpetual growth rate to the tenth year. Net cash is added, and the total divides by the share count.

Three scenarios recompute the same model with fixed offsets, identical for every company: the Bear case lowers growth by three points and raises the discount rate by one, the Bull case does the reverse, and the Base case, with no offsets, is the headline Fair Value the verdict keys on.

Multiples, Dividends, and the Rest of the Thirty-One

The multiples value averages the earnings-based legs that can be stated, the Graham formula leg and the earnings multiple leg among them, and it stands aside when trailing earnings are not positive, because a multiple of a loss is not information. The wider roster covers the ground a single family cannot: dividend models, an earnings power value, the Rule #1 family, an exit-multiple cash flow model, and funds-from-operations models for real estate investment trusts. Each model prints its own value or its own state with the reason, and the full roster with every figure sits one tap from the headline in the workup.

Sector shape matters, and the order handles it without special pleading. A bank or an insurer keeps no free cash flow line in the cash flow sense the model needs, so the cash flow step falls through and the headline lands on a model built for the sector. The method never forces a company through a model its filings cannot feed.

The Dividend Line at One Percent

The dividend model is the Gordon growth form: value equals the annual dividend per share, grown one year at the growth rate, divided by the required return less the growth rate. The shipped settings are a growth rate of 3 percent and a required return of 9 percent, and both are the reader's to change. Two-stage and H-model variants sit beside it in the roster. A company that pays no dividend gets the no-dividend state, plainly.

One guard was added in October 2026, after the model printed a precise-looking value for a company whose dividend is a rounding error on its price. When a payer's trailing yield is under 1 percent of the price, the dividend family reports Not Meaningful and prints its reason, Dividend yield is under 1% of the price, instead of a Gordon value. A dividend that small carries no information about what the company is worth, and a formula that divides by a small difference will manufacture any answer asked of it. A stood-aside dividend is excluded from every median and average it would otherwise feed. Payers at or above the line keep their dividend values to the cent.

Margin of Safety and the Verdict

The Studio states the gap between value and price in two forms, and the form depends on which side of the value the price stands on. When the price is at or below the value, the headline prints the margin of safety, measured on the value:

Margin of safety = (value − price) / value

At a value of $100 and a price of $85, the margin of safety is 15 percent, and the caption under the figure reads At the current price. When the price stands above the value, the headline changes both the label and the base: the figure prints as Fair value vs price, measured on the price:

Fair value vs price = (value − price) / price

The caption then states the result in words: Fair value is X% below the current price, with X the size of the gap. Measured on the price, the figure can never fall below minus 100 percent, however far the price runs above the value.

The buy price follows the margin convention in both cases: to demand a margin of m, pay no more than value × (1 − m). The Studio's preset required margin is 15 percent, and the buy price on every print moves with the reader's own setting.

The verdict reads off the margin of safety. Above zero, the stock is Undervalued. From zero down to minus 10 percent, it is Fairly Valued. Below minus 10 percent, it is Overvalued. When the headline model reports Not Meaningful, the verdict says so instead.

How We Test the Engine

The method on this page exists twice, in two independent implementations: the Studio's engine, and the valuation workbook published as the Automated Stock Valuation Suite. The two were built separately and share nothing but the method and the inputs. The release gate runs both over the same companies and the same inputs and compares every figure they print: 9,484 comparisons in the current gate.

Every comparison matches except ninety-six, and the ninety-six are accounted rather than unexplained. All of them sit in the dividend family, and all of them trace to the one percent line above: the Studio adopted the line in October 2026, and the workbook, a separate product with its own release cycle, still applies the earlier dividend treatment. The differences are the ruling working as designed, they are named figure by figure in the gate record, and the expected results were deliberately not rewritten to hide them. A gate that edits its own answers proves nothing.

The inputs pass their own audit. Before the current data feed shipped, every field the engine consumes was compared against the underlying filings for a panel of nine companies, differences were traced to the filed line that explains them, and the feed's selection rules were written down from the results. That record is the basis for the companion page on data sources.

Sources

• Company filings, public at SEC EDGAR: the inputs to every model on this page, selected as described in Where the Studio's Data Comes From.
• The Complete Guide to Stock Valuation Studio: the workup read field by field, including the model roster and the assumption settings.