The Intrinsic Value tab on a company page asks what the business is worth on its own numbers rather than on what the market pays for it today — and it asks four times, by four different routes. The tab opens on all four answers at once: one horizontal bar per method, on one axis, with today’s price drawn as a single vertical line through every row. Each method then has a page of its own where every assumption behind it is a box you can change.
None of it says where the share price goes. A value is what a method implies the business is worth on the assumptions in front of you, and a share can trade above or below that for years without either of them being wrong.
If you have no company in mind yet, the Intrinsic Value library is the way in: four lists over a snapshot rebuilt every night — the companies trading under what their own numbers imply, the ones trading over it, the largest on the market, and the dividend payers with twenty-five unbroken years of rises behind them.
No method describes every company, so the tab says which ones describe this one before it says what they found. Above the chart is a count — 3 of 4 methods fit — and the lead method, the one that suits this company best. Every method also carries one of four words, beside its row and on the strip above the chart:
Those last reasons are the useful part. A bank or an insurer is set aside for both cash-flow methods, because debt is its raw material rather than its financing. A company that pays no dividend, or whose yield is under one per cent, is set aside for the dividend model. Graham’s formula is set aside where earnings are negative or where there is not enough earnings history to measure growth over.
There is deliberately no blended figure, and no average of the four. The Fair Value tab next door does blend, and it blends one kind of thing — seven valuation multiples measured against the company’s own history. These four are four different questions, and a mean of them would be a number with nothing behind it. What the tab states instead is which methods fit, and the reason beside each one that does not.
Every one of them is computed on the server from this company’s own filings, and every one of them has a page where its assumptions sit beside the figure they produce. How much of the arithmetic that page shows depends on how much of it there is: Graham is one line, and it is written out with your own numbers in it; the Advanced DCF is a five-year build, and its page carries the year-by-year projection and the enterprise-value bridge; the two growing series are a sum of fifty terms, drawn as the shape of those terms with the model’s own total beside them.
Where a model has a scenario ladder, its page shows a Bear · Base · Bull switch, with Custom beside it for the assumptions you have typed yourself. (The Advanced DCF under its Perpetuity terminal method has none, and shows no switch: three positions that would all draw the same number are not a choice.) The first three are anchored to this company’s own record rather than to a spread somebody chose — the slow and fast ends of its growth history, or the low end and the upper quarter of its price-to-free-cash-flow range. Where there is a record to anchor to they never collapse onto the base case, so bear or bull can sit a little past anything the company has actually shown. Where there is none they fall back to the model’s own assumptions: the growing series take the base rate give or take two percentage points, and the Advanced DCF — which has no such spread to fall back on — runs all three on its own implied exit multiple and says so under the switch, which is why bear, base and bull can read one figure on a company with no price-to-free-cash-flow history. None of the three is a forecast or a target, and moving between them costs nothing: all of them arrived with the answer the page opened on.
Open a method from the strip above the chart, or by clicking its row. The assumptions sit on the left, the live result on the right, and every box starts from something this company has actually delivered — a margin from the median of its last three annual reports, a growth rate from the longest run of its own history we hold, a beta measured on its own shares. The information marker beside a box says where its default came from.
Type a new number and the model is re-run a moment after you stop typing; the figure on the right dims while it is being replaced and never disappears. A value outside the range the model accepts is refused on the box itself, with the range stated, rather than after the fact. The ↺ beside a box you have changed puts that one assumption back; Reset all, on the tab’s opening chart, puts everything back.
Advanced DCF has one switch rather than a box: the terminal value can be a Perpetuity or an exit Multiple. Graham has one too — the bond-yield adjustment, on by default; with it off the multiplier is 1 and the yield beside it is not read at all.
A few settings describe you rather than the company, and those follow you from one company to the next: the discount rate and the number of years summed, on each of the two growing-series models separately, and Graham’s bond yield together with its adjustment. Everything else resets with the ticker, because it is an assertion about one business.
Each value is set against today’s price. Ten per cent or more above it reads Undervalued, ten per cent or more below it Overvalued, and anything between them Fairly valued. A value at or below zero is Not meaningful — that is a statement about the method rather than about the company, and it means this route does not describe this business. Beside the verdict is the margin of safety: how far the price sits below the value, as a share of that value. It is zero rather than negative where the price is at or above the value — there is no such thing as a negative cushion.
A method can also suit a company and still produce no figure. The commonest case is a growing series whose growth rate sits at or above its discount rate, where the sum never settles — the row says so and points at the two numbers to move. Other causes have their own sentence, written by the calculation itself, so the reason on screen is always the actual one rather than a guess.
The axis stops at three times the price. A method that implies more than that still draws its bar all the way to the edge, with a › and its own figure beside it — the frame is capped so that one wildly optimistic perpetuity cannot flatten every other row into a sliver beside the price line.
The Summary tab carries a short version of all this: the lead method’s base case, named beside the figure, with the verdict against today’s price.
Everything you change lives in the address. The method you have open, the case you are reading, Advanced DCF’s terminal method and every assumption you have typed are all in the URL, so copying it out of the browser’s address bar sends exactly the screen in front of you — your numbers, not the defaults.
Only what you actually changed is written there. A box left as the model found it is not in the link, which means an untouched assumption follows the company’s own default as it is restated over time, rather than being frozen at whatever it was on the day the link was sent.
A link also outranks the settings that follow you between companies: open somebody else’s link and you see their discount rate, not yours. Reset all clears the address back to the company’s own defaults.
Set price target under the result writes the figure on screen against the company on a watchlist, and the list watches the market price against it from then on. If the account has more than one list, the dialog asks which; otherwise it names the one it will use.
What is written is the valuation, not a buy price. Each watchlist applies its own margin of safety to whatever it holds, so discounting the figure yourself before saving it would take that margin twice and put the buy zone somewhere neither of you meant. Save the value the method gave and let the list do the discounting.
If the company is already on that list, the valuation on it is replaced rather than refused — the same button updates a row it once created. A dividend-model valuation also carries the yield today’s declared dividend would give at that value, which is the one figure a watchlist’s own target-yield trigger can use.