How PhaseLens Works
The honest frame first: PhaseLens is an educational research tool built on deterministic rules over public data. It is not financial advice, it can be wrong, and every limitation we know about is listed at the bottom of this page — read that section, not just this one.
The Core Idea
Most AI stock tools give you a conclusion and hide the reasoning. PhaseLens does the opposite: every signal is produced by fixed, printed rules applied to public fundamental data. Two people analyzing the same ticker on the same day see the same result, and this page tells you exactly why.
1 · Lifecycle Phase
Each company is classified by where it sits in the business lifecycle, using revenue growth (rg), operating margin (om), and free-cash-flow yield (fcf):
- PRE-REVENUE / DISTRESSED — no revenue growth and no gross margin to speak of
- GROWTH — revenue growth above
15% - MATURE — growth of at least
3%with operating margin above10%and FCF yield above1.5%: the profitable compounder profile - DECLINE — growth under
3%with weak margins (<5%) or negative free cash flow - TRANSITION — growth of at least
3%that doesn't yet meet the Mature quality bar
2 · The Score and Signal
A rules-based tally awards points for fundamental strength — margins, growth, cash generation, balance-sheet health, valuation — and subtracts for weaknesses. The displayed score is capped at 100; when a stock exceeds the cap you'll see its raw tally (e.g. raw 112) so two 100s remain distinguishable. The signal is mechanical: BUY ≥ 70, HOLD 45–69, SELL < 45. Every point-award is listed as a named driver in the analysis — you can audit the entire score.
3 · Forensic Checks (the Buffett Five)
Independent of the score, five pass/fail forensic checks probe quality the way a value investor would: earnings backed by real cash flow, return on invested capital, balance-sheet strength, margin durability, and share-count discipline (dilution vs. buybacks). Three or more passes with clean share discipline earns HIGH confidence.
4 · The Value Verdict
A separate deterministic layer asks a different question — not "is this a good business" but "is the price a value case?" Labels: VALUE STOCK, GROWTH PLAY (quality business, not a value price), NEUTRAL, and VALUE TRAP (cheap for a reason). The verdict and the signal deliberately cross-check each other: a strong BUY-scored business is never simultaneously labeled a value trap.
5 · Fair Value — Three Methods, Honestly Labeled
- Earnings Power Value — DETERMINISTIC: what the business is worth if it never grows again
- FCF Perpetuity — DETERMINISTIC: current free cash flow valued as a flat stream
- DCF — labeled GROWTH-DEPENDENT: requires a growth assumption, so we cap it (≤10%), fix the terminal rate (2.5%), and give you the discount-rate slider (6–14%) so you can see how much the answer moves when assumptions move. That sensitivity is the lesson.
6 · Lenses (Themed Screens)
Five fixed screens over a curated universe of ~50 liquid US large-caps, each with its criteria printed on the card — including the exclusions. Highlights of the ruleset: leverage is tested by Net Debt / EBITDA first (D/E with a positive-equity guard as fallback, so buyback-driven negative equity can't masquerade as "low debt"); banks and insurers are excluded from screens whose metrics don't apply to lenders; gross margins above 98% are treated as data errors; growth screens use 3-year revenue CAGR to avoid peak-earnings traps; and any lens concentrated >35% in one sector says so in an on-card warning. The list is a fixed daily snapshot — same list for every user, fully reproducible.
7 · Today's Radar & Market Pulse
The Radar surfaces a handful of names from the day's most-active list that pass quality filters — selection is deterministic; AI only writes the context sentence after selection. The dashboard's Market Pulse shows breadth (share of your tracked tickers positive today, or an S&P 500 proxy, labeled accordingly, with "as of last close" outside market hours).
8 · Where AI Is (and Isn't) Used
Every number, score, signal, verdict, screen, and threshold on PhaseLens is rule-based — no AI involved. AI (a language model) is used only to write narrative prose: plain-English explanations and the bull/bear debate. AI text is descriptive garnish around deterministic results, never the decision-maker, and it is banned from making predictions.
Using the App — 60 Seconds Per Tab
- Research — type any US ticker, get the full analysis: phase, score with drivers, forensics, verdict, fair value, peers, structural insights
- Portfolio — add holdings (quantities stay in your browser — see Privacy); live P&L per position
- Watchlist — tickers you're monitoring; one tap to analyze
- Backtest — what the signal said at past decision points and what following it would have returned vs. buy-and-hold
- Lenses — today's five screens with printed criteria; load performance vs. S&P 500 per lens
- Dashboard — everything at a glance plus Market Pulse
Known Limitations — Read This Part
• Fundamentals are TTM snapshots; except where noted (3-yr CAGR), cyclical peaks can flatter metrics.
• Lens performance charts show today's constituents projected backward — survivorship-biased by construction, and labeled as such in the app. An honest track record only accrues forward from each daily snapshot.
• Data comes from a third-party provider and can be delayed, revised, or wrong; sanity guards catch some errors, not all.
• The universe is ~50 US large-caps for Lenses — deliberate (quality over coverage), but it means small caps and most international names aren't screened.
• Backtests are hypothetical, ignore taxes and slippage, and past performance — real or simulated — does not predict future results.
• None of this is financial advice. See the Terms.
Questions or found an error in this methodology? Email phaselensai@gmail.com — errors get fixed, not defended.