The 22 inputs behind a PAS

A PAS is not computed from a formula you could apply to one company in isolation. It is computed from 22 raw ratios and signals, each of which is first compared against the same measurement for every other company in the same GICS sector and country.

That comparison is the z-score: how many standard deviations from its peer group's mean this company sits. It is what makes a 40% gross margin read as ordinary in software and exceptional in grocery, without anyone hard-coding a threshold per industry.

The inputs, by pillar

Profitability (25%)

GP/AGross profit per dollar of assets — efficiency of the asset base.
OP/BEOperating income per dollar of book equity.
ROIC 5yFive-year average return on all capital employed.
FCF margin 5yCash generation per sales dollar, smoothed over five years.

Value (25%)

EV/EBITOperating earnings multiple — lower is cheaper.
FCF/EVCash yield on enterprise value.
E/P − 10YStock earnings yield vs Treasuries.
B/MBook equity relative to market cap.
Sales yieldRevenue per dollar of market value.

Quality / Safety (25%)

F-ScorePiotroski nine-point health checklist.
Altman ZDistress risk score.
Beneish MManipulation risk flag (lower is safer).
EPS stabilitySmoothness of five-year earnings.
Interest coverageEBIT vs interest expense.

Momentum (15%)

12-1m returnYear-long trend excluding the last month.
6m returnHalf-year price performance.
% of 52w highHow close price is to its yearly peak.
Est revisionDirection of analyst EPS estimate changes.

Growth (10%)

Rev CAGR 5yAnnualized sales growth.
EPS CAGR 5yAnnualized earnings-per-share growth.
FCF CAGR 5yAnnualized free-cash-flow growth.
Reinv runwayCapacity to reinvest at attractive ROIC.

What happens to a missing input

That sounds obvious and is the single most consequential rule here. The tempting alternative — substituting the sector median, or a neutral zero — is a fabrication: it asserts a measurement nobody made. A company with no five-year ROIC history has unknown five-year ROIC, and a score that fills that in with the median is quietly claiming otherwise.

The same rule runs one level up. A whole pillar with too little data is dropped from the weighted average rather than scored at 50, so a recently-listed company with no momentum history is scored on what is actually known about it.

Winsorizing, and why

Before averaging, every z-score is clipped at plus or minus three standard deviations.

Without that clip, one company with a deranged ratio — a near-zero denominator, a restated year — would dominate its own pillar and distort the sector statistics that everyone else is measured against. Clipping keeps an extreme value extreme without letting it become the entire signal.

Reading it against a screen

The inputs above are deliberately close to the published factor literature rather than novel. That is the point: each traces to a paper, and each can be argued with on the evidence.

If you disagree with a weighting — and there is no objectively correct weighting — the pillar breakdown on any company page shows which pillar is carrying the score, so the disagreement can be about something specific.