Most articles on this topic hand you a checklist: P/E under X, margins above Y, growth above Z. That framing is quietly broken. A ratio is not a score. It is a piece of evidence, and evidence only means something relative to a base rate and a prior.

So this is the same list of ratios you will find everywhere else, with a different question attached to each one. Not "is this number good?" but "given this number, how should my probabilities move?"

1. Revenue Growth: the Prior-Setter

Growth is where every tech thesis starts, so start with the base rate: sustained hypergrowth is rare. Companies growing thirty percent or more tend to decelerate faster than analysts model, and the market systematically extrapolates the recent past.

The probabilistic question isn't how fast the company is growing. It's what fraction of companies growing this fast were still growing anywhere near this fast three years later. If your thesis requires this company to sit in the surviving minority, the scenario table should say so out loud, and weight it like the minority outcome it is.

2. Gross Margin: the Ceiling on Every Scenario

Gross margin tells you what the business can become. An eighty percent gross margin software company has a path to thirty percent operating margins at maturity. A thirty-five percent gross margin hardware company does not, no matter how good the story sounds.

Treat gross margin as a constraint on the bull case rather than a virtue in itself. If the upside scenario implies terminal operating margins that exceed what the gross margin structurally permits, the scenario isn't optimistic. It's incoherent, and its probability should be close to zero.

3. Rule of 40: a Compression of Two Trade-Offs

Growth rate plus free cash flow margin at or above forty is a popular filter for software. Its real value is probabilistic: it forces growth and profitability into a single number, which stops you from awarding a company full credit for growth and full credit for future margins as though the two were independent. They aren't. Most companies buy one with the other.

A company at twenty-five percent growth with a twenty percent FCF margin and a company at forty-five percent growth with a zero percent margin score identically. They are not the same bet. The first has a narrow, tightly clustered distribution. The second has fat tails in both directions. Same score, different distribution, and the distribution is what you are actually buying.

4. Free Cash Flow Margin, Minus Stock-Based Compensation

Reported free cash flow flatters most tech companies, because stock-based compensation is added back. SBC is a real cost. It is simply paid in dilution rather than cash. A company reporting a twenty-five percent FCF margin while running fifteen points of SBC is, economically, a ten percent margin business.

The base-rate framing: markets eventually reprice companies whose profitability is mostly an accounting artifact. If the valuation scenarios are built on headline FCF, you are assigning probability to a company that does not exist.

5. Net Revenue Retention: the Leading Indicator Hiding in Plain Sight

For subscription businesses, net revenue retention above roughly one hundred and twenty percent means the existing customer base grows even with zero new sales. It is the best single predictor of durable growth, because it measures revealed customer behavior rather than management ambition.

Probabilistically, high NRR should raise the weight on the persistence scenario more than almost any other single metric, precisely because it is the metric least contaminated by narrative. Customers expanding their spend is evidence. A keynote is not.

The multiple only tells you what you are being asked to believe. The base rates tell you whether believing it is sane.

6. Valuation Multiples: the Market's Implied Probabilities

Price to sales, EV to gross profit, forward P/E. The checklist tradition treats these as things to judge, expensive or cheap. The probability tradition treats them as information. The multiple is the market's compressed forecast of the outcome distribution, and it is the most honest forecast available, because real money is behind it.

A stock at fifteen times sales is not "overvalued." It is a statement: the market is assigning high probability to sustained growth and margin expansion. The job is not to react to the multiple but to ask whether the probabilities inside it are miscalibrated. Sometimes fifteen times sales is cheap, when the true probability of the growth scenario is higher than the one the price implies. That gap, and only that gap, is the trade.

7. ROIC: the Maturity Test

Return on invested capital matters most at the transition point, when a growth story has to become a business. High ROIC means each retained dollar compounds. Low ROIC growth is a treadmill that looks like progress.

The probabilistic use is to discipline the terminal-value scenario. Terminal value is where most of a DCF's value hides, and where most of the fantasy hides with it. A company that has never demonstrated high returns on capital should not be handed a high-ROIC terminal state at sixty percent probability just because the model needs one.

The Actual Takeaway

No ratio is a verdict. Each one is an update. The checklist approach fails because it treats seven pieces of evidence as seven independent pass-fail tests, then behaves as though seven passes add up to certainty. They don't. They add up to a thesis whose probability you now have to state out loud, where it can be checked.

Write the scenarios. Attach the probabilities. Let the ratios move them. That is the difference between analyzing a tech stock and grading one.

If the framework is new to you, the method is set out in full.