How to value a stock: six models, one honest number
Any single valuation method can be talked into almost any answer. The fix isn't a cleverer model — it's triangulating several, then being honest about the spread.
7 min read · Updated 2026-07-08 · By Caverian
Why one model is never enough
Every valuation model is a lens, and every lens distorts something. A discounted-cash-flow model is exquisitely sensitive to the growth and discount rates you feed it; a price-to-earnings multiple quietly assumes this year's earnings are representative; a dividend model is useless for a company that doesn't pay one. Lean on a single number and you inherit that one method's blind spot as your whole view.
The discipline that separates careful investors from confident ones is triangulation: value the same business several independent ways and look at where the answers cluster — and, just as importantly, where they scatter.
The six lenses
Caverian runs every stock through the same six reproducible models so the consensus you see is comparable across companies rather than hand-picked:
- Discounted cash flow — the present value of the cash the business is expected to generate for its owners.
- Reverse DCF / earnings power — what growth the current price already assumes, so you can judge whether that's plausible.
- Relative multiples (P/E) — earnings valued against the company's own history and its peers.
- Enterprise multiples (EV/EBITDA) — the whole capital structure valued against operating cash, useful where debt or tax distort the P/E.
- Dividend discount — for genuine dividend payers, the value of the income stream itself.
- Asset / book-value anchor — a floor for asset-heavy or financial businesses where earnings are lumpy.
Turning six numbers into one
Six estimates are only useful if you combine them honestly. Caverian takes the median of the models that actually apply, not the average — the median shrugs off a single wild output instead of letting it drag the answer.
Models whose implied price is absurd relative to the market price (below about a quarter of it, or above roughly four times it) are treated as outliers: they're shown in the breakdown with the reason they were set aside, but they don't poison the consensus. One broken input should never quietly move your fair value.
Reading the result honestly
A fair-value estimate is a range with a confidence level, not a price target to trade against. The gap between price and the consensus is your potential margin of safety — the room you have to be wrong. When the models disagree wildly, that spread is itself the signal: the business is hard to value, so demand a bigger discount.
And when the underlying data is thin or stale, the honest answer is a dash, not a confident number. A fabricated figure is worse than no figure, because it invites a decision you'd never make with your eyes open.
Key takeaways
- One valuation model inherits one blind spot — triangulate several.
- Use the median of the models that apply, not the mean.
- Exclude absurd outliers from the consensus, but show why.
- Fair value is a range with a confidence, not a trading target.
- Thin data earns a dash, never an invented number.
Frequently asked questions
What is the single best way to value a stock?
There isn't one. Every method has a blind spot, so the most reliable approach is to value the business several independent ways — DCF, multiples, dividend and asset lenses — and study where the answers agree and disagree rather than trusting one.
Why use the median instead of the average of the models?
The median ignores a single extreme output, so one broken assumption can't drag the consensus. An average lets an outlier quietly distort the number you rely on.
Value any stock in the screener
Related guides
Sources: Yahoo Finance, Financial Modeling Prep, SEC EDGAR.