How it works — the whole method, including the parts that don't flatter us

How we value a company

The core estimate is a discounted cash flow. The idea is older and simpler than it sounds: a business is worth the cash it will hand its owners from here on, with cash arriving years from now counted for less than cash arriving next year. Everything below is just that sentence, done carefully.

1. Start from a real year, not a guess. We read the base year — revenue, operating margin, depreciation, capital expenditure, working capital, net debt and the diluted share count — out of the company's own latest annual filing on SEC EDGAR. Each figure on the page links to the filing it came from.
2. Forecast, and fade. We project revenue forward for the horizon shown on the page, with growth fading toward a long-run rate rather than extrapolating the last good year. This is deliberately conservative, and it is the single biggest reason our number often lands below the market's: a company the market expects to keep compounding will always value lower under a fading lens.
3. Turn revenue into cash. For each year: revenue → operating profit at the margin we hold → tax it → add back depreciation → subtract capital expenditure → subtract the cash tied up in working capital as the business grows. What is left is free cash flow. The same waterfall is printed line by line on every company page.
4. Value the years after the forecast. Beyond the last forecast year, the final year's cash flow is assumed to grow at a modest long-run rate forever, and that stream is valued in one figure — the terminal value. In most models this is the majority of the answer, which is exactly why we show the sensitivity grid.
5. Discount it. Every future amount is brought back to today's money at the company's cost of capital (WACC): the cost of equity — the risk-free rate plus the company's beta times an equity risk premium — and the after-tax cost of debt, weighted by how much of each the company uses. The WACC slider on a company page recomputes the entire model at any rate you prefer; the curve behind it is real engine output, not a re-implementation.
6. Get to a price per share. Add up the discounted cash flows and the terminal value to get enterprise value, subtract net debt, and divide by the diluted share count.

The number in the headline is not the DCF alone. One model is a fragile thing to stake a page on, so the published fair value is the median of a blended range — the DCF alongside peer-multiple lenses (P/E, P/S, EV/EBITDA), book value and, where the payout justifies it, a dividend-discount read. Methods that produce implausible values for a given company are dropped rather than averaged in, and the width of what remains is shown to you and feeds our confidence level.

What “researched” and “auto-modeled” actually mean

Coverage comes in three depths, and every page states which one it is:

• Hand-researched. A model a human built and reviewed line by line, with sourced assumptions and, where it matters, segment-by-segment forecasts. There are few of these, deliberately.
• Researched from filings. The base year comes straight out of the company's latest annual filing; the forward growth, margin and capital assumptions are still algorithmic estimates. Most large US companies are here.
• Auto-modeled. Every input, including the base year, is an algorithmic estimate from live market data. No human has reviewed the assumptions. Where a company is only auto-modeled, the page says so in those words.

Where the numbers come from

All of it is public data. Nothing here is proprietary, which means every figure on this site is one you could go and check yourself — and should, if a number surprises you.

The six things we check

Every company gets the same six checks. Each one is scored 0–100, then combined using the weights below. If we can't compute one of them, it drops out and the rest are re-weighted — we never fill a gap with a guess.

Why we compare against similar companies

A rule like “a P/E under 25 is cheap” is really a rule about large American companies. Applied to a utility, a bank, and a small Indian company it produces numbers that look comparable but aren't. So instead of fixed thresholds we rank each figure against companies in the same industry and the same market.

That's why the score can say something you can go and check: “cheaper than 63% of technology companies in the US”. When we don't have enough comparable companies, we widen the comparison — to the industry worldwide, then to everything we track — and we tell you on the page which one we used. If even that isn't possible we fall back to general rules of thumb, and we say so.

What the grades mean

Because every check is a ranking against comparable companies, a typical company scores near 50. The grades are set from the real spread of scores, not picked out of the air.

How sure we are — and when we refuse to answer

A score with no confidence attached is half a number. Every score carries a confidence level based on how much data we actually had, how much our own valuation methods agree with each other, how liquid the stock is, and whether our tools even fit the business (a cash-flow model is the wrong lens for a bank, and it understates a company growing 30% a year). A fair-value range wider than its own midpoint is never marked high confidence, however complete the data.

When there isn't enough to be honest about, we publish no score at all. You'll see what we could check and what was missing instead. On thin data we show a range rather than a single number, and we won't award a top grade — we'd rather look less confident than be confidently wrong. The same holds when our own cash-flow model is known not to fit the business (a company growing 30% a year on flat-margin, fading-growth assumptions): confidence is capped at medium and the grade is held at B, however well the remaining lenses score it, and the card says so.

The rules we hold ourselves to

These are not aspirations. Each one is enforced in the code that builds the page, which is why you will regularly see this site decline to answer.

1. We publish a range when a range is the honest answer.The fair value on a company page is the median of several methods, and the range around it is shown. On thin data the score itself is published as a band instead of a point.
2. “Not enough data” is a valid output, and we use it.Below a coverage floor we publish no score at all — you get what we could check and what was missing. Refusing beats guessing, and it is the reason the numbers we do publish mean something.
3. An analyst price target is reported, never adopted.Consensus appears beside our number as a second opinion, with the analyst count and the date. It is never blended into our model, never treated as a fact about the future, and a target with no analyst count behind it is not published at all.
4. We refuse the cases our tools do not fit.Banks, insurers, REITs, regulated utilities and models that land implausibly far from the market price are routed away from the DCF, with the reason stated on the page.
5. Every number shows the figure behind it.Each assumption cites the filing it came from; each check shows the actual value and the threshold or ranking it was judged against. You should be able to disagree with us on the evidence rather than on faith.
6. We say when we are the outlier.Where our fair value diverges sharply from the market or the street, the page argues the point in writing instead of leaving a naked contradiction on screen.

What this is bad at

Every valuation method has cases it handles badly. Most sites let you find those the hard way. Here they are, stated by the people who built the thing:

• Buybacks and capital returns. Our model discounts the whole equity's cash flows and divides by TODAY's diluted share count. A company steadily retiring shares therefore looks worse under our lens than it will turn out to be. On the free worked example we now size that: at the pace the filings show, continued through the forecast, we print the most it could add to the per-share number — and the reason that figure is a ceiling rather than an estimate.
• Banks, insurers and other balance-sheet businesses. A free-cash-flow DCF is the wrong instrument for a business whose debt is raw material rather than leverage. We refuse to publish one and show multiples, book value and dividends instead. A site that prints you a DCF fair value for a bank is making it up.
• REITs and rate-regulated utilities. A REIT's value sits in property and FFO/AFFO, not free cash flow, and a regulated utility earns a return set on a rate base. Both are routed to the same multiples-and-book view, for the same reason.
• Companies that do not yet make money. A DCF needs a margin to fade toward. Pre-profit companies have none, so what comes out is a forecast of a forecast. We would rather show what we can see — cash, burn, growth, dilution — than dress an assumption up as a valuation.
• Cyclicals at the top or bottom of a cycle. We take the base year out of the latest annual filing. For a chip maker, a miner or a homebuilder, that year may be a peak or a trough, and fading growth from an unusual starting point carries the distortion all the way through. Check the financial-history table on the page before trusting the base year.
• The terminal value does most of the work. In any DCF with a long horizon, most of the value sits in the terminal value — a number produced by two assumptions, the discount rate and the long-run growth rate. That is why every page carries a WACC slider and a sensitivity grid instead of one confident figure: the honest output is a range, and you should look at it.
• Everything that does not reach a spreadsheet. Management quality, competitive position, regulation, litigation, technological obsolescence, fraud. None of it is in these numbers. A model is a floor for thinking, not a substitute for it.
• Non-US companies and small listings. Coverage outside the US is thinner: EDGAR is US-only, so foreign filers lose the filing-sourced base year and fall back to estimates, and small or thinly traded listings publish less and publish it later. Where that leaves us unable to be honest, we show the light view or no number at all.
• Free data, with the delays free data has. Quotes are delayed and labelled as delayed. Filings arrive when they arrive. A figure can be stale between an event and the filing that records it, and we would rather you knew that than believed the page was live.

Track record — what we can and cannot claim

There is no backtest on this site, because we have not run one yet. Any accuracy figure you see quoted for TickerWorth today would be invented, and we are not going to invent one.

That is an uncomfortable thing to publish on the page whose job is to earn your trust, and it is also the only sentence here that could be written honestly. The service is new; the valuations it has published do not yet have enough forward time behind them to say anything meaningful about how they performed. A chart claiming otherwise would be the easiest thing on this site to fake and the fastest way to deserve none of your confidence.

Here is exactly how it will be measured and shown, so you can hold us to the method before there is a result to argue about:

• The cohorts. Every valuation we publish is already stored with its date, its fair value and the price on that day. That set splits into the companies we called materially undervalued, the ones we called materially overvalued, and the rest.
• The measure. Forward total return of each cohort over 1, 3 and 5 years, against a plain index benchmark over the identical window. Equal-weighted, no survivorship filtering, no dropping the ones that went badly.
• The sample. Published with its size, its start date and its composition — and left visibly small while it is small. A result on 40 names over 18 months will say so rather than being rounded up into a claim.
• The misses. The worst calls get named individually, with what the model got wrong. A track record without its failures is marketing, not evidence.
• The cadence. Refreshed quarterly once there is a year of data, and never restated backwards. The public changelog already runs — every revision to the model, its data and what this site claims about a number, dated and append-only. Per-company revision history (when a fair value moves, what filing moved it and which assumption changed) is the part that still has to be built.

Until that exists, judge this site the way you would judge an argument rather than a record: the assumptions are all on the page, the arithmetic is shown, and the free worked example lets you audit the whole method on a real company without paying anything.

Who's behind it

TickerWorth is independently owned and self-funded. There is no research department behind these numbers, no outside capital, no advertisers and no financial-industry credential. What stands behind them is the model, the filings it reads, and code that refuses to publish when the two don't agree.

That is worth knowing in both directions. Nobody here is paid to hold an opinion, nothing on this site is sponsored, and no company can call to have its number changed. It also means coverage is narrower than a large provider's, a data problem is fixed at one person's pace, and no compliance department reads a page before you do — which is precisely why every assumption behind a number is printed beside it.

Found something wrong? Corrections are the most useful thing anyone sends this project — every number on a company page carries the filing it came from, so a disagreement can always be settled against the source rather than against us.

Who owns this, and how to reach a human →

What this is not

• Not a prediction. Nothing here says what a share price will do. It can't.
• Not advice or a recommendation. We don't tell anyone what to buy or sell.
• Not a rating. A grade describes how a company compares with others on the things we check — no more than that.
• Not complete. We use free public data. It has gaps, it can be stale, and it says nothing about management, competition or the things that don't reach a spreadsheet.

Every number on a company page shows the figure behind it and the threshold or ranking it was judged against, so you can disagree with us on the evidence rather than on faith.