What is a stock quality score — and how is one built?

A stock quality score is a composite number that condenses how profitable, stable and financially resilient a company's underlying business is — typically built from metrics such as return on invested capital, margin stability, accrual levels and leverage, normalized against a peer group. It describes the business, not the share price, and it says nothing about whether the price already reflects that quality. This article walks through what goes into such a score, the research behind each ingredient, how the pieces are combined, and where the whole construction reaches its limits.

Why “quality” is the most measurable factor family

Factor investing sorts stocks along recurring characteristics — value, momentum, size, low volatility, quality. Of these families, quality is the one most directly anchored in audited numbers. Value needs a market price and a view on what “cheap” means. Momentum is entirely a price phenomenon. Quality, by contrast, is computed almost wholly from the financial statements a company files: income statement, balance sheet, cash-flow statement. Every input has a definition, a source document and an as-of date.

That measurability is why quality frameworks converge on a small, stable set of questions. Does the business earn genuinely high returns on the capital it employs? Are its margins wide, and do they stay wide across cycles? Do reported profits turn into cash? Is the balance sheet financed conservatively enough to survive a bad year? Different vendors weight these questions differently, but almost every serious quality score is some arrangement of exactly these four.

The classic ingredients

Return on invested capital (ROIC). The core profitability question: how much operating profit, after tax, does the business generate per euro of capital tied up in it? A company that turns €100 of invested capital into €20 of after-tax operating profit year after year is doing something structurally different from one that produces €5. ROIC is deliberately capital-structure-agnostic — unlike return on equity, it cannot be flattered simply by adding debt.

Margin level and margin stability. Gross margin measures the economics of the product itself, before overheads; operating margin adds the cost of running the organisation. Level matters, but stability often matters more: a business whose gross margin sits within a two-point band for a decade exhibits pricing power and cost control that a volatile margin series does not. Many scoring systems therefore measure the variability of margins or earnings over five to ten years, not just the latest print.

Accruals — do profits become cash? Reported earnings mix hard cash with accounting estimates. The gap between the two is called accruals, and it is one of the oldest red flags in the academic literature: Richard Sloan's 1996 study in The Accounting Review documented that firms whose earnings rely heavily on accruals rather than operating cash flow subsequently reported systematically weaker results than their low-accrual peers. A quality score that ignores cash conversion is measuring the accounting, not the business.

Leverage and balance-sheet strength. Debt amplifies everything — including the damage a bad year does. Quality frameworks reward conservative financing: low debt relative to equity or to cash flow, comfortable interest coverage, no reliance on continuous refinancing. The point is not that debt is bad; it is that resilience is a component of quality, and leverage is its most direct antagonist.

The research behind the recipe

None of this is folklore; each ingredient carries an academic citation. Robert Novy-Marx's 2013 paper “The Other Side of Value: The Gross Profitability Premium” (Journal of Financial Economics 108) showed that a strikingly simple measure — gross profits scaled by assets — predicted cross-sectional returns about as well as classic value metrics, and that profitable firms earned this premium despite trading at richer valuations. Joseph Piotroski's 2000 study in the Journal of Accounting Research assembled nine binary accounting checks — spanning profitability, leverage and liquidity, and operating efficiency — into the F-Score, and found that the highest-scoring value stocks went on to report markedly stronger fundamentals than the lowest-scoring ones. Asness, Frazzini and Pedersen formalised the theme in “Quality Minus Junk” (Review of Accounting Studies, 2019), defining quality as profitability, growth and safety, and documenting a persistent return spread between high- and low-quality portfolios across 24 markets.

Index providers translate the same research into production rules. MSCI's World Quality Index builds its quality signal from exactly three variables: return on equity, debt to equity, and earnings variability. Three numbers, published methodology, rebalanced on a schedule. The lesson for anyone evaluating a vendor's score: the credible ones can always be written down this plainly.

From metrics to a single number

Raw metrics are not comparable across companies — a 45% operating margin is extraordinary for a grocer and unremarkable for a software franchise. Turning metrics into a score therefore involves four steps, and every one of them is a design choice.

First, cleaning and normalization: extreme outliers are winsorized (capped at, say, the 1st and 99th percentile) so a single distorted data point cannot dominate, then each metric is expressed as a z-score or percentile rank within the comparison universe. Second, sector neutrality: metrics are ranked against sector peers rather than the whole market, because otherwise a quality score degenerates into a software-industry detector — entire sectors run structurally higher margins and lighter balance sheets than others. Third, weighting: the sub-scores are combined with fixed weights — equal weights, research-derived weights, or weights tuned per strategy. Fourth, aggregation and scaling to a readable range, commonly 0–100 or 0–10.

Two scores built from the same filings can therefore disagree, legitimately. One vendor winsorizes at the 5th percentile, another at the 1st; one ranks against a global sector, another against a regional one; one weights cash conversion at 30%, another at 10%. Disagreement between scores is not evidence of error — it is evidence that construction choices matter, which is precisely why a published methodology is worth more than an impressive-looking number.

A worked example on filed numbers

To make the pipeline concrete, here is one pass over a real large-cap, using only figures from its annual report. Microsoft's Form 10-K for the fiscal year ended 30 June 2025 (filed with the SEC) reports revenue of $281.7 billion, up 14.9% from the prior year's $245.1 billion. Gross profit of $193.9 billion puts the gross margin at 68.8%; operating income of $128.5 billion is an operating margin of 45.6%. Net income of $101.8 billion against year-end stockholders' equity of $343.5 billion works out to a return on equity of roughly 29.6%. Long-term debt stands at $40.2 billion — about 0.12 times equity, and still only around 0.26 if lease obligations are counted in. Free cash flow of approximately $71.6 billion — operating cash flow minus an unusually heavy AI-datacenter capital-expenditure programme — is about 25.4% of revenue.

Run through the pipeline, each of those figures becomes a sub-score relative to a sector peer group, and the weighted combination lands — under essentially any reasonable weighting — in the top decile of a global universe. Note what the exercise also reveals: the weakest sub-score is cash conversion, not because the business deteriorated but because capital expenditure roughly doubled in two years. A score-only view would flag the dip; only the underlying statements explain it. This example is an illustrative analysis of published figures, not an assessment of the security for any investor.

Where quality sits in Acutic's seven-factor model

Acutic's scoring model composes seven sub-scores — fundamental quality among them — into a 0–100 composite with weights that are fixed per strategy profile and disclosed. The fundamental-quality pillar reads revenue growth, earnings, profit margin, debt-to-equity, free cash flow, return on equity and market-cap stability from the underlying data provider, applies threshold-based sub-scoring, and normalizes the result — the same construction pattern described above, with every rule written down on the public methodology page. Because AI-assisted analysis runs on top of these scores, each output is labelled as such — the AI transparency page documents where models are involved and where they are not, and the performance page tracks how the published scores have behaved over time, in the open.

What a quality score cannot see

The limits are as important as the mechanics. A quality score is built from filings, so it inherits their cadence: quarterly at best, with a reporting lag — a business can deteriorate for months before the numbers say so. It measures the past and the present, not disruption: a franchise with pristine ten-year margins can still be two product cycles from irrelevance, and no accrual ratio will flag it. It is silent on price: quality and valuation are different axes, and a superb business bought at any price is a different proposition from the same business at half of it — which is why multi-factor systems keep a separate valuation score rather than folding price into quality. And it struggles with intangible-heavy accounting, where R&D expensing and goodwill treatment distort both the profitability numerator and the capital denominator.

Used within those limits, a quality score does one job well: it compresses hundreds of audited data points into a comparable, rankable signal of business resilience — a starting point for research, and never a substitute for it. If you want to see how score disagreement plays out across real vendors, the companion piece a field guide to AI stock scores reads five scoring systems methodologically, and portfolio rules you actually keep covers what to do with a score once you have one — the process layer where most of the value is actually created.

Further reading: the Acutic methodology page documents the full seven-factor construction, and the performance page shows how the scores have behaved since launch. Create free account to run the model on a portfolio.

Acutic provides investment research and educational analysis under MAR Art. 20 / § 85 WpHG. Acutic does not provide investment advice (Anlageberatung per § 1 Abs. 1a S. 2 Nr. 1a KWG / Art. 4(1)(4) MiFID II), portfolio management, or any other licensed investment service. No content in this article constitutes a personal recommendation.