# How the FolioScore works

## Overview

The FolioScore is FolioFundamentals' proprietary 0–100 rating of a company's fundamentals. It is produced by an AI and machine-learning algorithm that was built by backtesting more than 30 years of market history across 70+ metrics and tens of thousands of companies — millions of data points — and keeping only the factors that actually predicted returns. The model reads those metrics from a company's financial statements and market data, groups them into six pillars, adjusts for the company's sector, and measures each reading against the company's own history. The score is recomputed every trading day. No analyst opinion goes into the number.

Every score maps to one of four verdicts at fixed cut-offs that are the same for every company:

- **Strong Buy** — most pillars are strong at the same time.
- **Buy** — the fundamentals lean favourable, with a pillar or two holding it back.
- **Hold** — strengths and weaknesses roughly offset each other.
- **Avoid** — the weak pillars outweigh the strong ones on today's numbers.

The cut-offs, like the weights, are part of the model and are not published.

## Why judge a company against itself?

Absolute numbers mislead. A P/E of 25x sounds expensive until you learn the same company traded at 35x for a decade. A 6x debt-to-earnings ratio is alarming for a software company and ordinary for a utility. The FolioScore looks at every reading two ways: against sector-aware standards, and against the company's own history — up to 30 years of it for valuation measures. That leaves one question: is this business stronger or cheaper than it has usually been?

Coverage spans US, Canadian and UK-listed companies. Inputs come from audited financial statements and market prices, and the same rules are applied to every company, every day.

## The six pillars

Each pillar answers one question about the business. Below is the kind of evidence each one weighs. The specific factors, how they are scored and how much each pillar counts are proprietary.

### Valuation — is the stock cheap for what you get?

- Earnings multiples, both trailing and forward-looking
- Free cash flow per dollar of share price
- Enterprise-value multiples that account for debt as well as equity
- Book-value based measures for banks and insurers

### Growth — is the business getting bigger?

- Revenue growth over the last year and over several years
- Earnings-per-share growth over several years
- Whether growth is speeding up or slowing down

### Quality — is it a good business, not just a growing one?

- Return on the capital invested in the business
- Whether gross and operating margins are widening or shrinking
- How much of reported profit turns into free cash flow
- Return on equity and assets for financial companies

### Financial Health — can it survive a bad year?

- Debt relative to earnings
- How comfortably earnings cover interest payments
- Short-term liquidity
- Capital and leverage ratios for banks; real-estate leverage bands for REITs

### Momentum — is the market agreeing or disagreeing?

- The share price against its own long-term trend
- Recent performance against the company's own sector, not the whole market

### Shareholder Returns — is cash coming back to owners, and can it last?

- Dividends and buybacks combined, as a yield on the share price
- Whether the payout is affordable from the cash the business generates
- Funds-from-operations based payout for REITs

## How the model judges each reading

- **Two lenses on every metric.** Each reading is measured against sector-aware standards and against the company's own multi-year average.
- **Sector-native factor sets.** Banks, insurers, REITs and regulated utilities file different statements and run different balance sheets. Each gets a factor set built for it instead of failing tests that never applied.
- **Missing data is never a zero.** When a pillar lacks the evidence to be judged, it is set aside and the remaining pillars carry the score.

The pillars do not all count the same, and the weighting is part of what makes the model work. Rather than publish the formula, every stock page shows the reading on each of the six pillars, so you can always see where a score comes from. The plain-English write-up on that page is AI-written; the number is not, and nothing the AI writes can move it.

## What the score deliberately leaves out

The FolioScore is a quantitative signal, not investment advice. It does not include:

- Analyst ratings and price targets — shown separately as "Street View", never in the number
- Macroeconomic or sector-wide headwinds
- Management quality, governance or competitive moat
- Litigation, regulatory risk or geopolitical exposure
- Anything that has not yet reached a filing or a price

## What we publish, and what we don't

We publish what goes into the score — the pillars, the kinds of metrics behind each, the sector adjustments and the history comparison — and, on every stock page, the reading on each pillar, what changed since the last score, and the analyst view kept separate for comparison. We also publish the model's point-in-time backtest, including the years it got wrong.

We do not publish the factor list, the thresholds, the pillar weights or the verdict cut-offs. Those are the model, and they are proprietary.

## Frequently asked questions

**How often is the score updated?** Every trading day, and again whenever a new financial statement is filed. When a score moves meaningfully or the verdict changes, the stock page shows what changed and which pillars drove it.

**Does the score work for companies with no earnings?** Yes. Metrics that need positive earnings are set aside for that company and the pillars that can be judged carry the score.

**Do analyst ratings affect the score?** No. They are shown next to the score as "Street View" for comparison, but never go into the number.

**Is a "Buy" verdict a recommendation to buy?** No. It means the fundamentals lean favourable on today's numbers. It is a starting point for your own research, not a recommendation.

**Can I see the exact formula?** No. The factor list, thresholds and weights are what years of machine-learning work on 30+ years of market data produced, and they are proprietary. What you can always see is the reading on each of the six pillars for any stock.

Full methodology: https://foliofundamentals.com/methodology