The Fair Value tab on a company page asks a narrower question than the tab beside it: not what the business is worth, but what this market has historically paid for it, and whether today is dear or cheap by that standard. Seven multiples — P/E, Yield, P/FCF, P/OCF, P/S, EV/EBITDA and P/B — each drawn over this company’s own history, so today’s reading is a percentile of what the market has actually paid for it rather than a comparison with a sector average.
The tab opens on the answer rather than on a lens — the median across everything that applies, the spread around it, the upside against today’s price and a verdict — with a rail down the side carrying every lens and its own reading. Each lens then has a page where the multiple is drawn over time and you can argue with it.
Fair Value is part of Ultimate. Every new account gets Ultimate for seven days with no credit card, and everything you change on the tab lives in the address, so a link you saved opens on the same reading afterwards.
A lens is one multiple drawn over time against the share price. Its fair value is the figure per share today multiplied by the median multiple over the window you chose — this company judged against its own past, never against a sector or a peer group.
The verdict is a different measurement, and the more honest of the two: it is where today’s multiple sits inside that window, as a percentile. The two disagree on purpose. A share a quarter below its fair value sounds merely cheap; the percentile tells you whether it has been cheaper on one day in ten of its own history or on half of them. The five bands run from below the 10th percentile, through 10–30, 30–70 and 70–90, to above the 90th.
Two lenses are read differently and the tab handles both for you. The dividend lens is read the other way up, because a high yield is cheap. The EBITDA lens prices the whole company on its enterprise value, debt included, and comes back to a share price through the balance sheet.
A verdict is never shown without the window it was judged over. The same company scores differently over one year and over ten, so the chip always carries both — and the window it names is the one the lens could actually judge, which is shorter than the one you asked for wherever the history runs out. Ask a company with four years of statements for ten and it answers with everything it has.
The blend is the median of the lenses that apply, at equal weights — not their average. A mean is dragged by whichever lens stands furthest from the rest, and on a seven-lens spread that is routinely the one that suits the company least. The spread itself is printed beside the median, because the distance between the cheapest and dearest reading is the thing a single figure hides.
That spread is the part worth reading twice. A single figure implies a precision the method does not have: the same company can be cheap on cash flow and dear on book value at the same moment, and both readings are true about different things. The line under the median says how many lenses are in it and what the cheapest and dearest of them imply — and it counts what is left after your exclusions rather than the lenses that exist, so it always describes the figure above it.
Untick a lens in the blend card and it leaves the median immediately — no request, no reload, and the figure, the spread and the upside all move together. Price-to-sales on a mature bank, or book value on a software company whose assets are people, is a reading you are entitled to set aside.
The exclusions live in the address, so the link you send opens on the blend you were looking at rather than on the default one. They are not remembered between companies: an exclusion is an assertion about one business, not a standing preference.
The verdict goes dark while a lens is excluded. The blend’s verdict is computed from all the applicable lenses on the server and cannot be recalculated here, so leaving it on the card beside a median you have just moved would show you one figure that followed your tick and one that did not. A withheld verdict is the honest version of that; the lenses in the rail keep their own.
Some lenses cannot be judged for some companies, and that is a normal outcome rather than missing data or an error. A company that pays no dividend has no yield to measure. A loss-maker has no positive earnings, so there is no P/E to take a median of. A balance sheet holding more cash than the market pays for the whole company has no enterprise value above zero for EBITDA to divide into.
Such a lens keeps its place in the rail and states its own reason, shows a dash instead of a verdict, and is left out of the blend rather than counted as a zero. It has no checkbox at all — a box you could tick that would change nothing is a promise the screen cannot keep. What you are seeing is the tab declining to price something it cannot price, which is more useful than a number it would have had to invent.
The window decides which slice of history the median is taken over: one year, three, five, ten, or everything stored. Five years is the default. A window the data cannot carry is offered and disabled with the reason beside it rather than quietly falling back, so you always know which history you are looking at.
The switch on the overview moves all the lenses at once, and that is deliberate. The blend is a median across lenses, so one lens judged over a different stretch of history would make it a median of figures that cannot be compared with each other. There is no per-lens window beside a row in the blend card for the same reason.
A single lens can still be moved on its own page, where it is one reading rather than one term of a median. Do that and the rail re-reads that lens from the page you are on, so one screen never shows you two different verdicts for one lens.
The median is what the market has paid; it is not necessarily what you think the business deserves. Every lens page has a target-multiple field, and every applicable row of the blend card carries the same field in miniature. Type a number and the implied fair value, the upside, the shaded gap and the whole drawn history redraw around it immediately — the response carries the figure per share precisely so that this costs no request.
Type it in the unit you would say out loud: 22 for a P/E, 2.85 for a yield of 2.85%. Empty the box and the target is gone rather than zero — there is no fair value behind a multiple of nothing. The target rides in the address, one lens at a time, so it travels with a link you send.
The verdict goes dark under a typed multiple, here too. A verdict is a percentile of this company’s own history; a number you chose is a different coordinate system, and a verdict computed against your own wish would be a compliment rather than a measurement. Reset the field and it comes back.
Open any lens from the rail and it is drawn whole, from one response and with nothing further to fetch:
The earnings lens has one extra control: earnings can be read as the vendor adjusts them or exactly as filed. They are two genuinely different numbers, so the basis is named inside the formula line rather than assumed.
Intrinsic Value asks what the business is worth on its own cash, earnings and dividends; Fair Value asks what the market has historically paid for them. They are two questions rather than two halves of one, so nothing anywhere averages the two tabs together — you read them side by side and notice where they disagree.
In practice they fail in opposite directions, which is why both exist. A discounted cash flow can put almost all of its answer into a terminal figure nobody can check; a multiple against its own history cannot, but it inherits whatever the market has always believed about the company — including being wrong about it for a decade. When the two agree you have learned something. When they disagree, the reason for the disagreement is usually the most interesting thing on either tab.
The four models are covered in their own guide, how to value a company.