Methodology
Every researched-stock report on this site runs the same five frameworks against the same two sources: the company's own filings with the SEC, and Equibles for the share price. This page documents exactly how — every formula, the exact tag or field each one reads, and the handful of places a judgment call belongs to the model rather than to the data. It's the technical reference behind the plain-English article; nothing here is a simplification of what actually ran.
On this page
Data source and the one-snapshot rule
Every financial-statement figure comes from the company's own filings, read as XBRL from the SEC's companyfacts service — the same 10-K and 10-Q data the company filed, not a vendor's restatement of it. The share price and market capitalisation come from Equibles. Nothing is manually entered or adjusted. A report resolves one price quote, one diluted share count, and one debt figure; every one of the five frameworks below reads those same numbers rather than re-fetching or re-deriving its own version.
Reading filings directly imposes two rules that a standardized feed would hide. First, only real statement forms are read — 10-K, 10-Q, 20-F, 40-F and their amendments. A proxy statement restates the same figures at a different scale, and taking the newest filing regardless of form once read a net income of "4,433" (millions, from a proxy) in place of the 4,433,000,000 the 10-K reported. Second, every figure must belong to the company's current fiscal year. A tag a company stopped using is never carried forward — one filer's last interest-expense tag dated from 2012, and reading it as current implied a 0.1% rate on billions of borrowings.
Most large filers never tag a discrete fourth quarter, so a trailing-twelve-month figure is assembled by rollforward — the last full year, plus this year to date, minus the same period last year — rather than by summing four quarterly tags that do not all exist.
A published article is a frozen snapshot. It is never edited after publication — a fresh look at a company publishes a new snapshot and the old one stays exactly as it was, so the record shows what the frameworks actually said at the time, not a rolling "current view." That immutability is enforced at the database level, not just by convention.
Reporting currency check
Some foreign issuers file financial statements in their home currency while their shares trade in US dollars. When that happens, any figure that would combine the two — EPS, the Acquirer's Multiple, fair value per share — is withheld rather than silently computed at no exchange rate. This report does not attempt currency conversion; it only refuses to combine numbers that aren't in the same unit.
1. Earnings Yield
Net income is the trailing twelve months of NetIncomeLoss, assembled by the rollforward described above. Compared, for context only, against a roughly 4–5% benchmark — the return available from a safe, guaranteed alternative like a bond — never as a buy or sell signal.
2. Return on Capital
Joel Greenblatt's formula, from The Little Book That Beats the Market:
EBIT is the sum of four quarters of reported operating income. Net Working Capital deliberately excludes idle cash and interest-bearing short-term debt, so the denominator isolates the tangible operating capital the business actually needs — not its financing or its cash pile. EBIT is OperatingIncomeLoss; where a company does not tag it — some large filers never have — it is derived as pre-tax income plus interest expense, and where neither resolves, no return on capital is shown at all.
Net Fixed Assets is net property, plant & equipment, taken from the filed net figure or from gross cost less accumulated depreciation. It is deliberately required: when it cannot be resolved the whole ratio is withheld rather than computed with a zero in the denominator, which would report a company as far more capital-efficient than it is.
When invested capital is negative and EBIT is positive — customers or suppliers effectively fund the business, via deferred revenue collected before the service is delivered — the ratio is mathematically undefined, and no percentage is printed. That is reported as the strongest possible outcome the framework can produce, never as missing or failed data.
3. Acquirer's Multiple
Tobias Carlisle's formula:
Operating earnings is reported operating income over the trailing twelve months, with no one-off add-backs. Carlisle's method allows adding back items like impairments and restructuring charges; this report does not, and states so rather than leaving it implicit. Measured across ten tickers, those add-backs came to exactly zero for eight of them and under 3% for the other two — small enough that reading the operating line the company filed is worth more than a reconstruction that can go wrong in larger ways.
Peer selection. Competitors are named by a language model from its own knowledge of the business — not from a data-provider peer list, which was found to cluster companies loosely under a shared sector tag regardless of whether they actually compete for the same customers. Every name the model gives is then validated against the SEC's own registrant data: it must appear in the SEC's NYSE or Nasdaq ticker file and be registered as an operating company, not a hallucinated ticker, a fund, or something delisted. A validated peer runs through the identical enterprise-value and operating-earnings pipeline as the subject company — never a shortcut version.
A candidate is dropped from the multiple's calculation — shown, with the reason stated — when: it reports in a different currency than it trades in, its debt could not be resolved, its operating earnings are under 1% of revenue (a rounding artifact rather than a real multiple), or its multiple falls outside a 0–50x range. A peer that is currently losing money is shown with no multiple rather than dropped outright — it is still a useful comparison point. The subject company is compared against the median of whichever peers have a usable, positive multiple.
4. FS-Score
A ten-check composite comparing this year against last, across three pillars:
- Current Profitability — Return on Assets > 0 (Net Income ÷ Total Assets); Free Cash Flow to Total Assets > 0, where free cash flow is operating cash flow less capital expenditure — the SEC publishes no free-cash-flow tag, because it is not a GAAP line; Accruals, Cash Flow from Operations ÷ Net Income > 1.
- Stability — Leverage did not increase (Long-Term Debt ÷ Total Assets); Liquidity increased (Current Ratio, Current Assets ÷ Current Liabilities); Net Dilution — annual weighted-average diluted shares did not increase year over year.
- Recent Operational Improvements — Return on Assets increased; Free Cash Flow to Total Assets increased; Gross Margin increased; Asset Turnover increased (Revenue ÷ Total Assets).
Liquidity exemption. A current ratio below 1 still earns the point when three conditions all hold: the ratio is genuinely below 1; deferred revenue is at least half of current liabilities (the dominant reason for the shortfall, not merely present); and current assets still cover every other current liability once deferred revenue is set aside. That combination means the shortfall is a service-delivery obligation funded by cash the company has already collected, not a cash problem — a subscription business collecting a year of revenue upfront, for example.
Total score: 8–10 is labeled "Financial rockstar," 5–7 "Solid but mixed," 0–4 "Proceed with caution." All figures come from the two most recent annual income statements, balance sheets, and cash flow statements.
5. Intrinsic Value (DCF)
A ten-year discounted cash flow plus a terminal value, then a bridge from business value to fair value per share.
Year-1 free cash flow — unlevered
Free cash flow is operating cash flow minus capital expenditure, and under US GAAP that already sits net of interest paid — a levered figure. Discounting it and then subtracting debt again in the equity bridge below would charge the business for its debt twice. The interest added back must be gross: tags that report interest already net of interest income are refused outright rather than used, because netting understates what the business actually pays on its borrowings. When a company tags only a net figure, or stopped tagging interest in an earlier fiscal year, no valuation is published at all and the report names the year the tagging stopped. The effective tax rate is income tax expense ÷ pre-tax income over the trailing four quarters, clamped to a plausible 0–40% range and falling back to a standard 21% when pre-tax income is zero or negative. As a sanity check, the implied rate (gross interest ÷ resolved debt) must fall between 0.5% and 20%, or the add-back is withheld entirely rather than trusted.
The forecast
Year-1 FCF grows at a single rate for years 1–10, discounted at a rate set by a risk bucket — low (9%), mid (12%), or high (15%). Both the growth rate and the risk bucket are chosen by the model for that specific company, grounded in its actual historical free-cash-flow trend and business, and recorded alongside the model's own one-line reasoning for each. They are not fixed constants and are not chosen per number — they are a judgment call the model makes once per snapshot.
Terminal growth is fixed at 3% for every company this app values — never chosen by the model — deliberately below long-run economic growth so the terminal value can never assume a business outgrows the economy forever.
Business value to fair value per share
Debt, resolved once
One debt figure feeds every framework above: interest-bearing borrowings plus finance lease liabilities. Operating lease liabilities are excluded — that cost already sits inside operating earnings, so counting the liability again as debt would charge the business twice. Reading the filings settles the lease question by construction rather than by inference: no operating-lease tag is ever read as a borrowing, so there is no mislabeled line to detect and no judgment call to disclose.
Resolution is tiered, because no single tag carries total borrowings for every filer. A company that states a combined figure outright is taken at its word; otherwise the current and non-current portions are resolved separately and added. That layering is what the filings demand rather than a preference — one large filer tags only a combined debt-and-capital-lease line, and reading the narrower tag alone reported $1.2B against a real $27.9B; another files its borrowings solely as senior notes, which read as zero until those tags were recognised.
When a filing carries no borrowings tag at all, debt is taken as zero and the report says that the zero was assumed rather than reported — a genuinely debt-free balance sheet is common, but an understated debt figure understates enterprise value and makes a business look cheaper to acquire than it is, so the two cases are never allowed to look alike.
What the model is and isn't allowed to do
The model never invents a figure. Every number it writes about is computed first by the formulas above and handed to it as ground truth; its job is to explain what the figures mean, not to estimate them. The text it produces is checked afterward for raw unformatted numbers, metrics outside the ones actually computed, internal code values, data-provider field names, and recommendation or endorsement language — with one retry, and a section is withheld entirely rather than published if it still fails.
Two judgment calls genuinely are the model's own, and both are disclosed rather than hidden: the DCF's risk bucket and growth rate, and which real-world companies are named as Acquirer's Multiple competitors. Because both are non-deterministic, running the same ticker again on a different day can produce a different — usually similar, not always identical — growth rate, risk bucket, or peer set. The fair value shown on any given snapshot is the number the framework produced for that snapshot, from those particular judgment calls, not a single fixed truth about the company.
Limitations
- Every figure is frozen at the moment a snapshot was generated and is never refreshed — a snapshot dated months ago shows the price and financials that were true then, not today's.
- Figures are only as good as the tags a company chose. XBRL lets filers tag the same economic fact under different elements, abandon a tag between years, or omit one entirely — every case above where a figure is withheld rather than estimated exists because of that. Restatements and late filings flow through into these figures, and there is no second provider to check them against.
- Filings are read for US-listed registrants on NYSE and Nasdaq. A foreign private issuer reporting under IFRS may not tag the elements these frameworks require, in which case figures are withheld rather than approximated.
- The DCF is a single-scenario, ten-year model. It is exactly as sensitive to its growth and discount-rate assumptions as any DCF is — this page states plainly which of those are fixed for every company and which are chosen per company.
- Peer selection depends on a language model's knowledge of which businesses actually compete with each other, validated only for being real and currently tradeable — not checked against an independent, formal business classification.
- Educational only. Nothing on this page, or anywhere on Investor Explorer, is investment advice or a recommendation.
See also: Data Sources and Disclaimer.