How it's calculated

Everything behind a Techdamentals verdict, start to finish. No black box. Written for someone who has never valued a company; the Learn page explains any term used here.

1. Where the numbers come from

The financial data (the price, the cash the business brings in, its debt and cash, its share count, its sales) comes from established market-data providers. No single provider's word is taken for it. Before anything is valued, a few numbers are pulled two different ways and checked against each other:

The price, from two sources. The main provider's price has to agree with a second, independent source within 5%. If the two disagree, the analysis is marked unverified. Better to admit doubt than to sound confident on a bad number.
Sales, two ways. Sales per share times the share count has to come to the sales the company reported, within 15%. The price also has to sit inside the stock's own 52-week range. A company with more than one class of shares is valued on the total count its market value implies.
Special cases. Funds and ETFs are refused, because a basket of stocks can't be valued this way. A company that listed in the last six months is marked provisional. Anything that fails a check is shown with the failure, never hidden.

The starting year. One year of cash flow anchors a twenty-year projection, so a single odd year would move every method at once. When the latest year's cash flow comes in below 60% of the company's own three-year average (a year of unusually heavy spending, say), the three-year average is used instead, and the input is labelled so you can see that it was. The rule only works downward: a latest year well above the average is what growth looks like, not a distortion.

2. The four ways of valuing a company

Each one answers "what is a share worth?" from a different angle. Each has a kind of company it suits, and each says on its card when it couldn't run and why.

Cash flow, 20 years. Projects the spare cash the business brings in after paying its bills (free cash flow) twenty years out. The growth rate for the first five years comes from the published analyst forecast for next year, capped at 35% a year; it then fades to a 2.5% long-run rate by year eleven rather than stepping down onto it. Where no forecast exists, the company's own cash-flow history is used instead, and the page says which. Forecasts win over history because a past growth rate only describes the years it was measured over: far too high for a company scaling off a small base, far too low for one part-way through a heavy investment cycle. Each year's cash is then counted in today's dollars using the discount rate (see below), a value is added for every year after the twentieth (assuming the cash keeps growing 2.5% a year, forever), then the company's cash is added, its debt subtracted, and the total divided by the share count. The years-after value is usually more than half the total, which is normal for a business that doesn't stop in year 21, and is why it's shown separately.

Cash flow, 10 years. The same idea over ten years, with one average growth rate for the whole stretch and the same years-after value. Less dependent on far-future guesses.

Profit per share, 10 years. Projects profit per share instead of cash, for ten years, at a fixed 4% discount rate, and adds nothing for the years after, deliberately. Profit isn't cash you can take out of the business, so carrying it to infinity made the numbers worse, not better (it was tested, and produced figures several times any sane price). This is the most cautious of the four: what is a decade of profit alone worth?

Book value. What a share would cost at the company's usual price-to-book, which is the price compared with what it owns minus what it owes. Built for banks and asset-heavy businesses. Not running yet for any stock: the data source supplies only today's price-to-book, never the years of history this method needs. It's left out of the blend entirely rather than counted as zero.

Sales growth. Values the company on its sales, how fast they're growing, and how much of each sale it keeps as profit: a fair price is roughly the profit margin times the growth rate, in years of sales, capped at 15 times sales because no business holds a richer multiple than that for long. The one method that still works for a fast-growing company with little profit yet.

The discount rate. Money in the future is worth less than money today, so future cash is marked down before it's counted. The rate rises with the stock's beta, which is how much it swings compared with the market: a jumpier stock has to promise more to be worth the same. For US stocks it runs from about 5.8% to 8%, with beta held between 0.80 and 1.60 so an unusually calm or unusually wild stock can't produce a rate nobody would defend. The line those rates sit on was fitted so that, across dozens of well-covered companies, the fair values land close to where analysts put them on average; it is a calibration, not a quote of any market interest rate.

3. The verdict

The methods that could run are blended into one number, the fair value, with the cash-flow methods getting the most say (roughly 47%, 24%, 18% and 12%). A range of 10% either side is drawn around it. The verdict follows where today's price sits: below the range is undervalued, inside it is fair value, above it is overvalued. The page also shows how far apart the individual methods landed, because that spread, not the tidy band, is the real measure of how sure the number is. When only one method could run, you get one number and no range.

Wall Street's average price target is shown next to the fair value. Not because analysts are right, but because you deserve to see when the two disagree.

4. Where AI fits (and where it doesn't)

AI does judgment: choosing reasonable growth assumptions within caps, picking which method fits the company best, and writing the plain-English notes about the business. Plain, tested code does all the arithmetic: every projection, discount and blend is fixed and unit-tested, and runs the same way for every company. AI models are excellent writers and unreliable calculators, so one is never allowed to do the math.

5. The honesty rule

When the numbers don't agree with each other, the page says so: the verdict is labelled unverified, the headline percentage is left off, and the notes at the foot of the page spell out which numbers disagreed and which methods couldn't run. You can still save it to your watchlist; that's your call, and the note says what the saved range rests on. A confident number built on bad data is worse than no number. That's the founding rule of the product.

6. What this is not

Techdamentals is an educational research tool. It produces impersonal estimates that are identical for every user. It is not investment advice, not a recommendation, and not a substitute for your own research or a licensed professional. See the Terms of Service and Privacy Policy.