Methodology

FrontierLab is a research filter for growth stocks in frontier sectors (AI, semis, cloud, energy, space, AgriTech and more). A grade ranks a name on evidence-based quality signals so you know what to look at first — it is not a buy/sell call, a price forecast, or financial advice. Every threshold below is published, with its source, on the evidence page.

1. The quality signals behind a grade

• 1.1 Gross profitability (gross profit ÷ assets), pass ≥ 0.33. How much gross profit a company earns per dollar of assets — a simple, hard-to-game quality measure, and one of the most robust in the research (Novy-Marx, 2013). The level is the signal, so it's judged on an absolute scale, never sector-ranked.
• 1.2 Rule of 40, pass ≥ 40 (software). Yearly sales-growth % plus profit-margin % should clear 40 — growth that isn't just burning cash. Applied to software-like models only.
• 1.3 Revenue growth & trajectory, pass ≥ 20%/yr. Is top-line growth durable, and are gross margins improving over time?
• 1.4 Dilution. Whether the company is printing shares (shrinking each owner's slice) or buying them back — heavy issuers tend to underperform; net repurchasers tend to do better (Fama–French, 2008).
• 1.5 Cash runway & burn (early-stage), pass ≥ 18 months. Can a pre-profit company fund itself, and how efficiently does it burn (burn multiple)?
• 1.6 Distress gate. Names showing financial distress are disqualified from a top grade regardless of growth (Campbell–Hilscher–Szilagyi, 2008).
• 1.7 Gross margin — sector-calibrated absolute bands. A fertilizer maker (~20–30% gross) shouldn't fail a SaaS bar it could never clear, so the pass/caution band is set per sector (e.g. cloud 72/55, AI 60/40, energy 35/22, AgriTech 28/18). The full table is on the evidence page.
• 1.8 Valuation is context, not a forecast. PEG (≤1 undemanding, ≤2 fair, ≤3 rich) with a P/S fallback (≤5/≤15/≤25) flags how much greatness the price already assumes (Lakonishok–Shleifer–Vishny, 1994) — never a price target.

Early-stage and established companies are graded on different rubrics — a young, speculative name is ranked among its peers ("worth a closer look first"), never against a mature company's scale.

2. ETFs are graded on a separate rubric

Funds get a character read, never the single-stock grade: cost (expense ratio ≤ 20 bps = core), size & liquidity (AUM ≥ $1B, ≥ $10M traded/day), diversification (holdings ≥ 50; top-10 concentration shown when available), and structure/age (≥ 3y; actively-managed funds capped at "caution" — active outperformance does not persist, SPIVA). Leveraged/inverse funds are hard-excluded from a buy-and-hold grade.

3. What we deliberately do NOT do

• No "fair value" or price targets. The single-number "this stock is worth $X" models (called DCFs — discounted cash flow) and analyst price targets run systematically optimistic and are often wide of the mark — we don't publish them.
• No RSI-as-a-sell-signal. Momentum context is shown as context only, never a timing call.
• No fake precision. Every track-record figure is shown with its sample size (N) and window; small samples are labeled "too early to tell."
• Insider selling is not treated as bearish — it's usually routine diversification. Only cluster buying is highlighted as a positive.
• End-of-day data only. No intraday, no real-time, no stop-loss simulation — and we never present cached data as live.
• Grades are frozen at entry. When you log a paper trade, the grade that applied that day is recorded immutably, so the track record can't be rewritten after the fact.
• Peer context is display-only — how a name compares to its sector is shown for context but never feeds the grade itself.

4. Does it work?

We publish a forward-looking track record — measuring each grade tier's later return versus the S&P 500, recorded daily, with confidence intervals. It is directional evidence, not a promise, and it discloses its own survivorship limits: the record measures ordering among names still covered at each horizon's end within the curated universe (not a survivorship-clean whole-market backtest), names later dropped from coverage are counted as attrition rather than hidden, and extreme single-session moves are treated as split/data artifacts and excluded.

5. AI assistance, disclosed

News digests and on-demand research summaries are drafted by AI from the data above and clearly labeled; they explain, they don't predict. The grade math itself is deterministic and identical between the data pipeline and the dashboard.

Every threshold, with its source → · See the track record → · Open the screener →