Methodology
How every number and score on Sixth Element is produced. Current version: v1.0.
Principles
- Facts, calculations and interpretations are kept separate and labelled throughout the product.
- Deterministic first. Every metric and score is computed by code from stored data. AI is used only to phrase explanations and to translate plain-English screens into filters.
- No false precision. Scores are rounded to the nearest 5 and shown with a band (Strong, Solid, Mixed, Weak).
- Missing is missing. If an input is unavailable or fails a data-quality check, we show “—” and say why. We never fill gaps with estimates.
Periods and TTM
Trailing-twelve-month (TTM) figures sum the latest four consecutive fiscal quarters. If a quarter is missing, TTM is not shown and the latest fiscal year is used instead. Balance-sheet items come from the most recent period end. Quarterly growth compares a quarter with the same quarter a year earlier to avoid seasonality. When a filing is restated, the most recently filed version is used.
Per-share figures and share counts are adjusted for stock splits so history is comparable on today's share basis.
Scorecard
Each of the four pillars is a weighted average of components. Each component maps a metric onto 0–100 using a fixed piecewise-linear scale (listed below). If some components are missing, the remaining weights are rescaled; if less than half of a pillar's weight is available, that pillar shows “insufficient data”. The overall score needs at least three pillars.
| Pillar (weight in overall) | Component (weight) | Scale |
|---|---|---|
| Growth (25%) | Revenue growth (1Y) (35%) | −10% → 0, 0% → 30, 10% → 60, 20% → 80, 35%+ → 100 |
| Revenue CAGR (3Y) (35%) | Same scale as 1Y revenue growth | |
| EPS CAGR (3Y) (15%) | Same scale; excluded if EPS was negative | |
| FCF CAGR (3Y) (15%) | Same scale; excluded if FCF was negative | |
| Profitability (30%) | Gross margin (20%) | 0% → 0, 20% → 30, 40% → 55, 60% → 80, 75%+ → 100 |
| Operating margin (30%) | −10% → 0, 0% → 25, 10% → 50, 20% → 75, 35%+ → 100 | |
| FCF margin (30%) | −10% → 0, 0% → 25, 10% → 55, 20% → 80, 30%+ → 100 | |
| ROIC (20%) | 0% → 10, 5% → 35, 10% → 55, 20% → 80, 30%+ → 100 | |
| Financial health (25%) | Net debt / EBITDA (40%) | Net cash → 100, 1× → 80, 2× → 60, 3× → 40, 5× → 10 |
| Current ratio (20%) | 0.5 → 10, 1.0 → 45, 1.5 → 75, 2.0 → 90 | |
| Interest coverage (25%) | 1× → 10, 3× → 40, 8× → 75, 15×+ → 100 | |
| Positive free cash flow (15%) | Positive → 100, negative → 20 | |
| Valuation (20%) | P/E (30%) | 8× → 100, 15× → 85, 20× → 70, 30× → 45, 50× → 20 |
| P/FCF (25%) | Same scale as P/E | |
| EV/EBITDA (20%) | 6× → 100, 10× → 80, 15× → 60, 25× → 35, 40× → 10 | |
| P/E vs own history (25%) | At historical low → 100, median → 50, historical high → 0 |
The valuation pillar scores lower multiples higher. That is a description of the multiple, not a judgement that a stock is a bargain: low multiples often reflect real risks. For banks and insurers, leverage-based health metrics and EV/EBITDA are not meaningful and are excluded.
Why did the score change?
We recompute the scorecard as of the previous fiscal year end using only data available then, and attribute the difference to each component: change in sub-score × component weight × pillar weight. Drivers under one point are omitted.
Sixth Element Apps: Find Good Companies at Reasonable Prices (Quality + Value Finder)
The finder reuses the scorecard above rather than introducing new scores. Quality = profitability 40% + financial health 30% + growth 30% of the existing pillar scores (missing pillars re-weighted; profitability and at least one other pillar required), rounded to 5. Value = the valuation pillar score. Overall = the overall score. Default filters: quality ≥ 60, value ≥ 40, positive free cash flow, debt/EBITDA ≤ 3× (net-cash companies always pass), delisted companies excluded. “Why it appeared” notes are generated by fixed rules from the same metrics (e.g. operating margin ≥ 25% → strong profitability).
Valuation metrics
- P/E = market cap ÷ TTM net income; not shown when earnings are negative.
- P/S, EV/Sales, EV/EBITDA, P/FCF, FCF yield use TTM figures and the latest close. Enterprise value = market cap + total debt − cash and short-term investments.
- PEG (trailing) = P/E ÷ (3-year EPS CAGR × 100). It uses historical, not forecast, growth.
- Forward P/E requires licensed analyst estimates and is not available in this version.
- Historical ranges use monthly observations over five years, pairing each month's price with the latest fiscal-year results known at the time.
Data quality checks
Every ingestion runs automated checks: missing values, duplicate or restated filings, impossible values (e.g. gross profit above revenue, negative share counts), negative revenue, balance-sheet identity (assets = liabilities + equity within 0.5%), free-cash-flow arithmetic, sudden revenue changes over 60%, missing quarters, stale prices and stale fundamentals. Error-level failures hide the affected metrics and are shown on the company page.
AI explanations and their limits
- The model only receives a short list of labelled facts drawn from our database and must use those numbers verbatim.
- Every output is checked automatically: numbers must match the provided facts, statements must cite fact ids, and recommendation language (“buy”, “sell”, price targets, “cheap”, guarantees) is rejected. Anything that fails is replaced by our rules-based explanation.
- AI can still phrase things imperfectly. Treat explanations as a starting point and check the numbers next to them.
- If no AI model is configured or the monthly AI budget is reached, explanations are generated by rules from the same facts.
Methodology changes
Changes are versioned. Scores record the methodology version they were computed with.
| v1.0 | Initial scorecard: four pillars, piecewise-linear component scales, rounding to 5. |