Methodology

How Valuation Studio estimates what a company is worth, and where those estimates can go wrong.

What a valuation is here

Every company gets a low, base and high estimate of value per share. The base value is the central estimate; the low and high values show how much it moves when the growth, discount and terminal-growth assumptions change. We compare the base value with the market price. The "upside" is base value ÷ price − 1.

Models we use

• Free-cash-flow DCF (the default): projects free cash flow, discounts it to today, subtracts debt, adds cash and divides by shares.
• Earnings-based DCF: the same idea using net income, used as a cross-check or when cash flow is unreliable.
• Dividend discount model: for companies whose dividend is a large share of earnings.
• Price-to-book model: for banks and insurers, where book equity drives value (justified P/B from return on equity).
• Normalised free-cash-flow DCF: uses a three-year average cash flow when the latest figure is depressed by heavy investment or one-offs. The latest figure is kept when it is in line with net income.
• EV/EBITDA: a fixed 12x multiple, used when cash flow is far below earnings even after averaging.
• Peer EV/Sales: the median revenue multiple of similar-growth companies in the same sector, for high-growth companies priced far above their cash-flow value.
• Rule-of-40 EV/Sales: a revenue multiple scaled by growth plus margin, for technology and communication companies whose cash-flow value is far below the price.
• Steady-state FCFF (cycle-adjusted): for memory and storage makers, whose cash flow peaks and crashes with prices; current cash flow is held at long-run growth.

Revenue multiples are used only when reported revenue growth is at least 8% a year, so an analyst forecast alone cannot justify one.

Growth, discount and terminal-growth assumptions start from sector defaults, blended with the company's own revenue history and analyst estimates where available. Every company page lists the other models and why each was not used.

Confidence: High, Medium, Low

The confidence label shows how reliable the estimate is, not how well the stock will do. In the back-test, High-confidence companies (steady, mature businesses) returned less on average than Low-confidence ones, but the forecast ranked companies best inside the High group. It shows how far to lean on a result. It rises when a second model agrees with the first, when recent trailing-twelve-month data is available, and when profit and cash flow have been steady for three years. It falls when the models disagree strongly, when only old annual data exists, or when the value is more than twice the price. It is always Low for cyclical businesses (energy, materials, airlines, carmakers, memory chips), for investment holding companies (valued on net assets less a typical 10% discount, with book value only a proxy for their stakes), for companies whose dividend jumped far above earlier years (the lasting level is used instead), and when free cash flow is more than twice net income. Hover over a confidence label in the screener to see the reasons. On a company page the verdict card also shows the range and median of the other independent models (cash-flow DCF variants and revenue multiples are left out), and warns when the recommended value is more than 30% above that median.

Best value ideas

The "Best value ideas" screener preset and the dashboard list rank companies by the average of their free-cash-flow yield (three-year average) and earnings yield, because the back-test below shows value signals rank companies best. It requires at least a 10% margin of safety and Medium or High confidence, and leaves out financials, whose cash flow does not fit these models. In the back-test, ranking by these yields within High and Medium confidence companies gave the strongest 3-year ranking so far, on a small sample, so treat it as encouraging, not proven.

Worked example: free-cash-flow DCF

A company with 100m of free cash flow, 8% growth for 5 years, 3% terminal growth and a 10% discount rate. Debt is 200m, cash 100m, and there are 50m shares. Amounts in millions.

1. Present value of the 5 forecast years: 473.4.
2. Terminal value = last FCF × (1 + terminal growth) ÷ (discount − terminal growth) = 2162.0; discounted to today: 1342.4.
3. Enterprise value = 1815.8. Add cash, subtract debt: equity value 1715.8.
4. Per share: 1715.8 ÷ 50 = 34.32.

Worked example: dividend discount model

Dividend 2.00 per share, 3% growth, 8% discount rate. Next dividend = 2.06; value = 2.06 ÷ (8% − 3%) = 41.20 per share.

Every company page shows the growth, discount and terminal assumptions next to its fair value, and you can change them there.

Accuracy and backtest

Test run on 5 Nov 2026: 1311 company-years from 732 companies. Each year was valued with the lower of the free-cash-flow and earnings DCFs, on a three-year average cash flow and blended growth, priced 90 days after year end, then compared with the 12-month return.

Setup checks. The screener's "setup" score uses the valuation plus a positive six-month price trend. We also tested quality (score 60 or more) and improving fundamentals compared with the year before; neither added to returns on top of the valuation, so they are not part of the score. Same test, same company-years:

How to read this. Undervalued means model value at least 20% above price; overvalued at least 20% below. The undervalued group returned more than the average, but the gap is modest and the sample is small. The overvalued label barely differs from the average, so it is not a sell signal: most companies fall in that group under default assumptions, so "undervalued" is deliberately rare. Limits: survivorship bias (only companies still listed), a mostly rising market, overlapping years, approximate filing dates, no transaction costs, and past results do not predict future ones.

Forecast back-test

Each past company-year was run through the live valuation models and forecast using only the data known then, and compared with what the stock returned. Simple benchmarks run on the same company-years show whether the forecast adds anything: rank correlation above the benchmarks means it ranks companies better than that simple rule.

What it shows. The ranking power comes from valuation: the live forecast ranks about as well as plain earnings yield and a little below free cash flow yield. Growth, quality and price momentum did not rank companies on their own, and blending them in made the ranking worse, so they are not part of the outlook. The outlook is a relative ranking aid built on a transparent projection, not a prediction of the return you will earn.

Main comparison. The live forecast against the simple rules. The free-cash-flow-yield result rests on fewer company-years than the others, so treat it as early evidence.

Analyst estimates were not available for past years, so they are left out. Benchmarks that only rank companies show no predicted return. Same limits as above apply.

Known limits

• Cyclical businesses (airlines, autos, commodities) can look very cheap at the top of a cycle, because current cash flow is extrapolated.
• One-off jumps in cash flow, or statements that still include a business that was spun off, can overstate the value.
• Strong growth companies often look expensive because the model caps growth in the long run.
• Data comes from third parties and can be delayed, restated or wrong. See the data health page for the latest refresh status.

A large upside or a High confidence label is a reason to look closer, not a recommendation to buy. Valuation Studio does not give investment advice.

Data sources

Prices: EODHD. Financial statements: EODHD (with SEC and other public filings where available). Prices are the last close and can be delayed.