One entry per screen: what it shows, how to read it without drawing the wrong conclusion, and — the section that matters — what it cannot tell you. Every limit below was read out of the code that produces the screen, not inferred from its name.
The weighted average cost of capital, its equity and debt weights, and each component: CAPM cost of equity, pre-tax and after-tax cost of debt, the risk-free rate, the market risk premium and the effective tax rate.
How to read it
Read the weights first. They are market-value equity over market-value equity plus debt, so a company whose price has moved a long way since its last filing will show weights that moved with it.
Cost of equity is risk-free + raw beta × market risk premium. It uses the RAW regression beta, not the Blume-adjusted one shown beside it on BETA.
If the number is an em dash rather than a figure, an input was unavailable. The screen never substitutes a plausible value to complete the arithmetic.
What it cannot tell you
The market risk premium is not a survey estimate or a published ERP. It is the annualised total return of the chosen benchmark minus the risk-free rate, computed over whichever of 10y, 5y or 1y of history loaded first — so the window can silently be one year, and a strong or weak benchmark decade is baked into it.
Debt is book value from XBRL balance-sheet tags, not the market value of the debt. In a stressed credit that overstates the debt weight; in a low-coupon legacy stack it understates it.
Cost of debt falls back in three steps: FINRA TRACE bond yields, then SEC interest expense over total debt, then the risk-free rate itself — the last labelled a lower-bound proxy. Check which one you got before comparing two companies.
The effective tax rate is only accepted between 0% and 60%. Outside that band it is dropped, and with it the after-tax cost of debt and the WACC itself.
It cannot tell you a forward-looking or target capital structure. Every input is trailing.
It cannot price an issuer whose filings it cannot match. A missing capital structure yields null weights, not 100% equity — that failure mode is guarded explicitly in the service.
An ordinary-least-squares regression of the stock's daily returns on a benchmark's daily returns: raw beta, Blume-adjusted beta, alpha, R-squared and the number of overlapping observations.
How to read it
Check the observation count and R-squared before the beta. A beta with a low R-squared is a slope through a cloud, and the slope is the least interesting thing about the cloud.
Adjusted beta is 0.67 × raw + 0.33 × 1.0 — the standard Blume shrink toward the market. It is a convention, not a measurement.
What it cannot tell you
Alpha is the raw DAILY regression intercept expressed as a percentage. It is not annualised, and a daily alpha of 0.05% is not a 0.05% edge over the year.
It refuses to run below 60 overlapping observations rather than reporting a fragile slope.
Returns come from Yahoo adjusted closes, but the Stooq fallback path carries RAW closes — so on a day Yahoo is unavailable the regression silently changes its dividend treatment.
Daily returns only. There is no weekly or monthly estimation option, and daily data is the noisiest choice for an illiquid name.
Beta describes the past window. It is not a forecast, and it is not stable across window lengths.
Valuation multiples for the stock beside a peer group discovered at request time — trailing P/E, EV/EBITDA, EV/Revenue, price/sales, price/book and the peer mean and median.
How to read it
Look at the peer list before the multiples. The basket is discovered, not curated: recommendations first, then sector and industry screens.
This screen's P/E is market capitalisation ÷ net income, and it is withheld when net income is not positive. A blank is a loss-making company, not a fetch failure.
What it cannot tell you
The peer basket is capped at 14 names and the sector screen excludes anything under a $2bn market capitalisation, so small-cap comparables are structurally absent.
Peers are matched across two different taxonomies — Yahoo's classification against Nasdaq's rows — and the service itself records that only coarse sector overlap is reliable.
Rows are dropped, not estimated, when a peer's data does not arrive inside its per-row timeout. The same request run twice can therefore produce a different median.
When no peer row survives, the subject company is included in its own peer statistics rather than the screen showing nothing — read the peer count.
Multiples mix vintages: Yahoo trailing-twelve-month fields are preferred and any gap is backfilled from SEC ANNUAL values, so one row can blend TTM and last-10-K figures.
It cannot tell you whether a discount is deserved. It measures spread, not quality.
Income statement, balance sheet, cash flow, a ratio block and product and geographic segment tables, built from SEC XBRL company facts.
How to read it
Columns are fiscal years as the issuer tagged them, not calendar years. Comparing two companies with different year ends compares different periods.
Where a line was not tagged directly it is derived — gross profit as revenue minus cost of sales, EBITDA as operating income plus depreciation and amortisation, free cash flow as operating cash flow minus capital expenditure.
What it cannot tell you
Annual filings only, and at most the last eight years. There is no quarterly view here, so the newest column can be almost a year old.
The whole screen depends on the issuer's own XBRL tagging. Where an exact us-gaap concept is missing, a similarity match may substitute a neighbouring tag; the substitution is exposed with its score rather than hidden, and it is worth checking on unusual issuers.
Segment tables come from the latest annual inline-XBRL filing. An issuer that does not tag segments gets no segment rows — never an invented split.
There is an opt-in Yahoo fallback, off by default, and it is a distortion by design: one TTM column in which total assets are mapped to market capitalisation. If you see a column labelled TTM here, you are not looking at a balance sheet.
It cannot restate anything. Non-GAAP adjustments, pro-forma figures and management's preferred metrics are not here — only what was filed.
Historical earnings surprises with the reported actual against consensus, and the forward EPS and revenue consensus periods a provider currently publishes.
How to read it
Surprise percentage is the provider's own published figure, which it computes from unrounded values it does not publish. Recomputing it from the rounded actual and estimate on screen gives a different and wrong answer, so that is used only when no published figure exists, and the row says so when it is.
The period column is derived, not supplied: the provider gives the ANNOUNCEMENT date, so the period shown is the calendar quarter that closed before it. Where that derivation would produce two rows claiming the same quarter — an issuer whose fiscal year does not follow the calendar — the column is dropped entirely and the announcement date is shown instead.
A period with no reported actual is omitted entirely, so gaps in the history are periods that have not been reported rather than periods that missed.
What it cannot tell you
One provider. There is no SEC input here at all, so a company Yahoo does not cover has no estimates on this screen and no second opinion behind the ones it does show.
The price reaction column is approximate by construction: it snaps to the first trading day on or after the fiscal QUARTER END, not the announcement date. For a company reporting weeks after quarter end, that is the wrong day.
The forward table is capped at eight periods and drops the provider's current-quarter and current-year placeholder rows.
Beat, miss and match are read from the sign of the surprise alone. The label carries no view on guidance, quality of earnings or how the estimate was set.
Consensus dispersion is only shown when the provider published a low and a high; where it did not, the screen says so rather than showing a false-precision point estimate.
An issuer's listed equity classes, its priced bond records from TRACE, its preferred shares, and — kept deliberately separate — the debt it discloses in its SEC filings.
How to read it
The two debt blocks are not the same thing. TRACE rows are instruments that actually traded and carry a price and a yield. SEC rows are balance-sheet aggregates with no price and no yield, and the screen labels them as not tradeable.
Where a yield is not published it is computed locally from price, coupon and maturity — a calculated yield, not a disseminated one.
What it cannot tell you
Issuer matching against bond records is heuristic string matching, not an identifier join: the name is uppercased, corporate suffixes are stripped, and a row is accepted if any of the first three tokens appears in it. False positives are possible on common names.
Bonds with no published maturity are dropped the moment you apply a maturity filter.
The SEC debt fallback only runs when TRACE returned nothing AND no maturity filter is set, so filtering can make the fallback block vanish.
Preferred shares are detected by string test on the security name and capped at 20 rows.
Every section has an aggressive per-source timeout; a section that bounded out is empty, which looks identical to an issuer that has none. Read the attempt notes.
It cannot tell you the full capital structure of a private or foreign issuer, and it cannot value an instrument that has not traded.
Not yet documented
65 live screens do not have a manual entry yet. Rather than generate one from a template, they are listed here with the one-line purpose from the index. A page that reads like documentation while carrying none is worse than an admitted gap — and the section this manual exists for, what a screen cannot tell you, cannot be produced by a generator at all.
Who holds the stock and how it is positioned: 13F institutional holdings differenced against the prior quarter, Form 4 insider transactions with option exercises and tax withholding separated from open-market trades, FINRA consolidated short interest, and federal awards. Every count states how much of the source was read.