The Filter Lab is in active development. Some panels show illustrative data; full functionality ships with paid tiers. Your feedback shapes the product.
Three independent lenses on thousands of public companies - structural dynamics, traditional fundamentals, and forensic accounting. When they converge, the independent measurements are telling a consistent story. When they disagree, you've found the question worth asking.
3
Independent Lenses
0.801
Two-Axis AUC
28
Years Tested
1.27M
Company-Months
AUC - how well the signal separates firms that later collapsed from those that didn’t: 0.5 is a coin flip, 1.0 is perfect. Across the top-3,000 study cohort — 962,834 name-months and 3,939 bankruptcies over 28 years — the two-axis signal scores 0.801. The full record is on the Validation page.
Structural dynamics
How the company's own variance is behaving over time.
Traditional fundamentals
The balance-sheet and earnings-quality signals analysts already trust.
Forensic accounting
Flags where the reported numbers deserve a closer look.
Converge→ a consistent story. Disagree→ the question worth asking.
From thousands of companies to the few worth understanding.
Everything that matters about a company, organized into a single comparable research view.
Evidence, not assertions
Anchored to the company's own words. Verified directly from EDGAR - no paraphrase, no invention.
For your AI tool
Structured intelligence - organized before it ever reaches ChatGPT, Claude, or your own workflow. Every figure sourced, nothing to rebuild.
The read, in plain English
A skeptical, descriptive narrative of what the numbers show. What's worth questioning - never what to do.
See it on a real company
finance.thefilterlab.app / analyzer
NVDANVIDIA Corp
Partial Convergence
Turbulence
Elevated
Traditional
Solid
Forensic
Flagged
Turbulence 0.116Altman Z 52.40Piotroski F 6/9Beneish M −1.16
Turbulence and the forensic screen both flag stress; the traditional lens (solid) doesn't confirm - the signal is real but not yet fully corroborated. The Beneish flag is a screen for a closer look, not a verdict.
Skeptic's read
A real read from the live app - NVIDIA, where the three lenses only partly agree. Every figure is pulled and sourced directly from EDGAR.
New here? How It Works walks the four steps ·
Validation shows the 28-year record (0.801 AUC across 1.27M company-months) ·
Help is one click away.
The Framework in the Real World
Real point-in-time readings from the engine — and what the market did next. Filter Lab doesn’t predict outcomes; each case shows a different way the three lenses can line up — divergence, disagreement, convergence — and the part of the tool that surfaces it.
CVNACarvana Co.
observed Dec 2023
The price was pricing a collapse. The books weren’t.
The measurements at the time
Turbulence
Extreme
score 2.00
Traditional
Solid
Altman 3.1
Forensic
Clean
Beneish −2.8
Relationship: Wide — violent price turbulence against strong, clean fundamentals.
Turbulence band Extreme
LowElevatedHighExtreme
What deserved questioning
Whether the extreme price turbulence reflected a genuine solvency problem — or a market panic the fundamentals didn’t corroborate. The lenses disagreed, and the disagreement was the whole point.
Subsequent outcome · next year
The stock rose roughly +308% over the following year (running to about +800% across two years before giving part back). Shown for context, not as a result the reading forecast.
Where the tool does this
This is a Screener job: filter for extreme turbulence sitting on top of strong, clean books, and divergences like this surface before the name is on anyone’s radar. The Analyzer then shows you why the lenses split.
Why this case matters
Turbulence is not the same as fragility. When the ride is violent but the books are strong, the framework isolates the one question worth researching — instead of lumping panic and decline together.
Relationship: Split — a strong balance sheet and a forensic flag at the same moment.
Turbulence band Extreme
LowElevatedHighExtreme
What deserved questioning
The balance-sheet lens read solid, but the forensic screen flagged the quality of the reported numbers — a reason to read the filings more closely, not an accusation. Which lens do you trust when they disagree?
Subsequent outcome · next year
The stock fell roughly −66% over the following year (about −79% across two). Shown for context, not as a result the reading forecast.
Where the tool does this
The Analyzer runs the forensic lens independently of the solvency lens, so a disagreement like this doesn’t get averaged away. Put the name on your Watchlist and a flag that appears later becomes something you’re told about, not something you have to catch.
Why this case matters
The lenses are independent on purpose. A single blended score would have washed the flag out; keeping them separate is what let the disagreement stay visible.
Relationship: Broad convergence — the three lenses agree.
Turbulence band Extreme
LowElevatedHighExtreme
What deserved questioning
Cash burn and funding structure, and whether the strain showing up across all three lenses was company-specific or an industry-wide condition. The forensic item is a Beneish-screen flag — a question for closer examination, not an accusation.
Subsequent outcome · next year
The stock fell roughly −27% over the following year and kept sliding — down about −84% across the stretch since. Shown for context, not as a result the reading forecast.
Where the tool does this
Convergence only means something in context, which is a Peers job: line the read up against the rest of the cohort to see whether the whole group is strained or just this name — the difference between a sector you avoid and a company you do.
Why this case matters
When three independent lenses lean the same way, the framework is describing a consistent story across measurements — not predicting timing. Convergence sharpens the picture; peers tell you whether it’s the company or the sector.
Readings are the engine’s point-in-time output (no lookahead); outcomes are the actual realized returns, shown for context — not as evidence that similar readings will produce similar results. Filter Lab is a descriptive risk tool, not investment advice.
The Filter Lab Research Group LLC and The Filter Lab provide structural analysis tools for educational and informational purposes only. We are NOT a registered investment advisor, broker-dealer, or financial planner. Nothing on this site constitutes investment, legal, tax, or accounting advice.
About the methodology
The Filter Lab measures equity-return dispersion (turbulence) alongside accounting fundamentals for public companies. While the turbulence signal has been validated across every US-listed company over a 28-year period (see the Validation page), past patterns do not predict future outcomes for any individual company. Markets are complex adaptive systems subject to many factors this methodology does not measure.
No Recommendations
The Filter Lab does not recommend any particular security, sector, strategy, or transaction. References to specific tickers, sectors, or historical companies (Enron, WorldCom, Spirit Airlines, etc.) are illustrative of the framework only and do not constitute recommendations to buy, sell, or hold those or any related securities.
No Warranty
Analysis results are provided AS IS without warranty of any kind, express or implied. We do not warrant that results are accurate, complete, current, or fit for any particular purpose. Data is sourced from third parties (SEC EDGAR, Tiingo) and may contain errors, omissions, or delays.
User Assumes All Risk
You are solely responsible for your investment decisions. You should consult a licensed financial advisor, tax professional, and/or legal counsel before acting on any analysis from this site. Past performance is not indicative of future results. You may lose money - including all of your invested capital.
AI Disclosure
This site uses artificial intelligence (Claude by Anthropic) for educational analysis. AI-generated content may contain errors, fabrications, or outdated information. AI outputs are not a substitute for professional advice.
Limitation of Liability
To the maximum extent permitted by law, The Filter Lab Research Group LLC, its officers, employees, contractors, and affiliates shall not be liable for any direct, indirect, incidental, consequential, or punitive damages arising from your use of this site, including but not limited to investment losses, lost profits, or data inaccuracies.
Jurisdiction
The Filter Lab Research Group LLC is a Texas LLC. Use of this site constitutes acceptance of the laws of the State of Texas as governing law. By using The Filter Lab, you acknowledge you have read, understood, and agree to this disclaimer.
The Filter Lab collects: (1) ticker queries you submit; (2) aggregate usage analytics such as page views and session duration; (3) your email address and subscription status when you create a paid account - account creation and authentication are handled by Supabase, and payment card details are handled entirely by Stripe, so we never see or store your card number. Free-tier use does not require an account or personally identifiable information.
Data Sources
We retrieve company financial data from public sources: SEC EDGAR (filings) and Tiingo (market data). We do not permanently store this data; it is fetched fresh per analysis.
What We Do Not Do
We do NOT sell your data. We do NOT share your data with advertisers. We do NOT use your queries to train AI models without explicit consent.
Third-Party Services
Payment processing is handled by Stripe, which stores your card data (we never see or store card numbers). Authentication and account data are managed by Supabase. Hosting is provided by Vercel and Render. Aggregate, cookieless usage analytics are provided by Plausible. Each provider operates under its own terms, and we share only what is necessary to run the service. No data is sent to any third-party AI provider from this app.
Cookies
We use only essential cookies required to keep you signed in (session and authentication via Supabase). Our usage analytics (Plausible) are cookieless and do not track you across sites or collect personally identifying information. We use no advertising or cross-site tracking cookies, and we do not use Google Analytics or similar broad tracking tools.
Your Rights
You may request deletion of any account-associated data by emailing support@thefilterlab.app. We will respond within 30 days. EU/UK users have additional rights under GDPR including data portability and the right to be forgotten.
Children
This service is not directed to anyone under 18. We do not knowingly collect data from minors.
Changes
We may update this policy. Material changes will be communicated via email to account holders. Continued use after changes constitutes acceptance.
By accessing The Filter Lab (operated by The Filter Lab Research Group LLC), you agree to these Terms. If you do not agree, do not use the service.
Service Description
The Filter Lab provides structural financial analysis tools based on its equity-turbulence methodology. The service is provided for educational and informational purposes only. See our Disclaimer for limitations on the nature of this analysis.
Account & Acceptable Use
Free tier use does not require an account. Paid tiers require an account. You agree to: (a) provide accurate registration information; (b) keep credentials secure; (c) not attempt to reverse-engineer, scrape, or otherwise exploit the service; (d) not use the service for any unlawful purpose. We may suspend or terminate accounts for violations.
Subscription & Refunds
Paid plans are billed through Stripe. Daily is $10 per month or $96 per year; Investor is $29 per month or $279 per year. You may cancel anytime; cancellation takes effect at the end of the current billing period, and you keep access until then. We offer a 7-day money-back guarantee: if you request a refund within 7 days of your first payment, we will refund it in full. The Filter Lab does not offer free trials.
Intellectual Property
The Filter Lab methodology, source code, and brand are property of The Filter Lab Research Group LLC and Ryan W. Malone. You may not copy, redistribute, or commercialize the service or its outputs without permission.
Disclaimer of Warranties
The service is provided AS IS and AS AVAILABLE. We make no warranties of any kind, express or implied, including merchantability, fitness for a particular purpose, accuracy, or non-infringement. See our Disclaimer page for details on the educational nature of analysis outputs.
Limitation of Liability
To the maximum extent permitted by law, The Filter Lab Research Group LLC's total liability for any claim arising from your use of the service is limited to the amount you paid in the 12 months preceding the claim. We are not liable for indirect, incidental, consequential, or punitive damages, including investment losses.
Governing Law
These Terms are governed by the laws of the State of Texas. Disputes shall be resolved in the state or federal courts of Texas. You waive any right to participate in class actions to the extent permitted by law.
Changes
We may update these Terms. Material changes will be communicated via email to account holders or via in-product notice. Continued use after changes constitutes acceptance.
The daily read on the whole universe: the turbulence regime, the macro backdrop as a trend, and which sectors are warming today - your starting point before the Screener. It is free with an account, and a new account gets 14 days of Investor in full, no card.
Market Analysis
Today’s market structure
Every site shows you today’s news. This shows today’s structure — across the 2,718 companies Filter Lab follows, as of July 20, 2026. Descriptive: it characterizes conditions now, it doesn’t forecast them.
Showing an illustrative snapshot — live data is momentarily unavailable, so the figures below may be stale. They refresh automatically when the feed returns.
1Market Analysis→2Screener→3Analyzer→4Watchlist
We filter down — from the whole universe to the few names worth your attention — so you make better decisions.
2,718 companies, not 10,000+ — by design.
Every name here has the SEC filings and trading history to actually analyze. Fewer names, less noise — the first filter is already run.
Today’s market state
40%
in High or Extreme turbulence
vs ~28% on a typical day
44%
of those still warming
not stressed, not settling
2,718
companies covered
every one fully analyzable
Watch
macro regime
VIX calm · credit benign · inflation sticky
The read
What it adds up to
▲
Technology — rising fastest
62% of its high‑band names warming, led by semiconductors at 71%.
▶
Health Care — high, not intensifying
Most high‑turbulence names, yet warming and cooling cancel.
✓
Credit — benign
Spreads at the 13th percentile; stress below normal.
⚠
Inflation — above target
CPI 3.46%, core PCE 3.14% — the two Watch reads.
The longer read
The top‑down backdrop is quiet. Implied volatility sits in its calm band, the 10‑year/2‑year curve has steepened back to +0.39 points after a long flat stretch, and credit shows no strain — high‑yield spreads near the 13th percentile of the past year and the financial‑stress index below its normal range. The labor market is softening at the edges (unemployment drifting to 4.2%, claims flat near 208k) but nowhere near the Sahm recession trigger. The one thing still flashing is prices: CPI at 3.46% and core PCE at 3.14%, both easing but stubbornly above the ~2% target, which keeps policy restrictive and the dollar firm.
Against that calm surface, the universe’s internals are busier than the headline suggests. 40% of the 2,695 banded names sit in the top two turbulence bands, and 44% of those are still warming — a market that isn’t stressed, but isn’t settling either. The nuance that matters: turbulence and its direction are not the same thing. Health Care carries the single largest block of high‑turbulence names (372), but its warming and cooling roughly offset — a large, standing pool of dispersion that isn’t building, the signature of idiosyncratic, name‑by‑name stress rather than a coordinated sector move.
The energy in the tape is in Technology. It has fewer high‑band names than Health Care (230), but 62% of them are warming — the highest of any major sector — and the concentration tightens as you drill in: semiconductors are 71% warming, software close behind. Industrials and Materials are firming too (52% and 55%), which alongside chips reads as a capital‑spending and AI‑infrastructure theme, not a defensive rotation. Financials and Energy, by contrast, are stable‑to‑quiet.
Put together: a benign macro regime with rising, concentrated micro‑turbulence in the parts of the market tied to the AI build‑out. None of this points a direction — a warming band widens the range of what a name can do, it doesn’t say which way. But as a place to aim your attention today, the structure is unambiguous, and it is where the screener will take you next.
🤖 Written from the canonical macro object + universe breadth + sector movement, refreshed nightly and cached — descriptive only. Company‑ and event‑level “why” arrives with the news layer.
The regime
What today’s bands imply
Sort every company into its turbulence band, then look at the actual range of 24‑month forward returns that historically followed — the 28‑year test. As turbulence rises the typical outcome drifts down, but the spread explodes: the Extreme band holds the worst wipeouts and the biggest moonshots alike. The number under each band is how many of today’s 2,695 names sit there.
bars span the 10th–90th percentile of 24‑month forward return; dot is the median · counts are today’s universe (1.27M company‑months, 1998–2026)
Today the mass is split — 1,612 names in Low/Elevated, 1,100 in High/Extreme. That’s the door: a high band tells you the range of what could happen is wide, not which way you’ll go through it. A second “company health” axis sharpens the downside within each band; that’s why the analyzer reads the two together, never turbulence alone.
Regime by sector
How wide the door is, by sector
The same door, projected onto each sector by where its turbulence sits today. A tall bar means a wide range of outcomes historically followed companies at that sector’s current turbulence level; the dot is the median. Descriptive — a wide door doesn’t say which way a name goes, only that the range is large.
24‑month forward-return range implied by each sector’s current median turbulence · illustrative mapping from the band test
Health Care and Technology carry the widest doors — their typical company already runs hot, so the range of what could follow is large. Financials, Real Estate and Utilities sit tight and low: narrow doors, contained ranges. Width is the range of outcomes; the next chart is which way the range is moving.
Where it’s rising
The heat, and which way it’s moving
A wide door is a standing condition; a warming one is developing. Bar length is how many names in a sector run hot; the orange segment is how many of those are warming. Health Care’s pile is large but cool; Technology’s is smaller but mostly lit — that’s the sector to watch today. Click one to open the screener pre‑filtered.
names in High/Extremeof those, warming
Information Technology
230 · 62%
▲▲ Accelerating
→
Health Care
372 · 32%
▶ Stable
→
Industrials
158 · 52%
▲ Rising
→
Consumer Discretionary
121 · 44%
▲ Rising
→
Materials
64 · 55%
▲ Rising
→
Financials
42 · 38%
▶ Stable
→
Comm. Services
41 · 41%
▲ Rising
→
Energy
31 · 32%
▶ Stable
→
Whittle down → 230 names in Information Technology
Screener filter: Information Technology · band High/Extreme · trend warming
MUAMDLRCXMRVLANETPLTR
Market
2,695
40% hot
→
Sector
Technology
62% warming
→
Industry
Semiconductors
71% warming
→
Sub‑industry
Memory
roadmap
→
Company
Micron
Analyzer
Macro backdrop
The environment it’s all reacting to
The regime the universe is playing out against, read as a trend. Each indicator shows its last three readings and the actual change, so the direction is the point.
Indicator
3 mo ago
1 mo ago
Latest
Δ 3‑mo
Trend
Rates
10Y–2Y curve
−0.08
+0.18
+0.39
+0.47
steepening
Fed Funds
4.30%
3.88%
3.63%
−0.67
eased, now on hold
Inflation
CPI (YoY)
3.90%
3.54%
3.46%
−0.44
cooling, still > target
Core PCE (YoY)
3.42%
3.19%
3.14%
−0.28
cooling slowly
Labor
Unemployment
3.9%
4.1%
4.2%
+0.3
drifting up
Initial claims
214K
208K
208K
−6K
flat
Credit
High‑yield spread
3.15%
2.80%
2.73%
−0.42
tightening (benign)
Financial stress
−0.42z
−0.79z
−0.88z
−0.46
easing
Global drivers
US dollar index
116.5
120.2
120.5
+4.0
firming
WTI crude
$88
$82
$79
−$9
falling
Industrial production
+0.6%
+1.0%
+1.14%
+0.54
expanding
What it means for the economy: the through‑line is disinflation without a recession. Prices are easing on both gauges, the yield curve has un‑inverted, and credit has relaxed — while the labor market softens only gently and never trips the Sahm rule. That mix is historically the profile of a late‑cycle soft landing, not an imminent downturn. The one counter‑current is a firm dollar against falling oil, which quietly tightens conditions on exporters and commodity producers even as domestic credit loosens.
last three readings per series; illustrative pending a stored release history in the nightly FRED job.
Macro events · what just printed
The data behind the backdrop
FRED releases are the market’s real news — public, scheduled, and where the regime actually moves. Public‑domain data — no license, unlike company news.
Jul 15
CPI, June · 3.46% YoY ▼ cooler than 3.5% prior
Third straight month of easing — keeps policy patient, but still above the ~2% target.
Jul 11
Nonfarm payrolls, June · +57K▼ below the ~110K trend
Hiring is slowing, though initial claims held flat near 208K — softening, not cracking.
Jul 09
Core PCE, May · 3.14% YoY ▼ edged down
The Fed’s preferred gauge — sticky and still above target, the reason rates stay restrictive.
Jun 27
High‑yield credit spread · 2.73%▶ near 13th %ile
Credit stayed relaxed — no funding‑stress signal into the print.
Next prints · Jobs report Aug 1 · CPI Aug 12 · FOMC decision Sep 17
This week’s data reinforced the calm‑but‑watchful backdrop rather than changing it. Cooling CPI and soft payrolls ease pressure on the rate‑sensitive corners — and it shows: Financials and Real Estate carry the tightest, lowest doors of any sector, consistent with a curve that has steepened rather than inverted. What the macro prints don’t explain is the heat in Technology, warming on AI‑infrastructure demand. The data tells you the weather; the sector doors tell you where it’s actually raining.
Sector Screener
Cross-sectional cohort view — one transparent metric per column. Not a ranking or a recommendation.
Free tier
Free
Current Plan
Free
$0
Teaser · No account needed
Ticker or Company Name
Data Source
Pulls directly from SEC EDGAR filings (10-K/10-Q). For names without filings, such as foreign issuers or brand-new listings, it automatically reads Tiingo price history instead.
Or type any US ticker or company name in the sidebar →
-
--
● REGULATED
⚠Not investment advice. The Filter Lab provides structural analysis tools for educational use only. The Filter Lab Research Group LLC is not a registered investment advisor. Past patterns do not predict future outcomes. Full disclaimer →
View
View
Financials - the Analyst view reads the statements; the Quant view reads the distress models.
Run an analysis on the left to see the company overview — what it does, where it sits, and why its readings are worth a look.
Run an analysis on the left to see the plain-language read.
7-Layer Financial Cascade
Stress mapped across seven layers, liquidity to management credibility. L4 (capital structure) is the primary turbulence target. Click a layer for detail.
Valuation · the Price lens
What the market is currently paying for this company's earnings, as a measurement. The third lens alongside the Ride and the Company.
The Ride Over Time
How wild the ride has been over time. Higher means bigger swings, not a higher or lower price.
Shape of the swing over time, computed from daily closes. The current band comes from the full model; this line shows trajectory, not the exact scored signal.
Revenue vs Operating Income
Quarterly trend across the most recent 12 quarters of EDGAR filings (where available). The gap between top-line revenue and operating income is the operating margin - how much of each dollar in stays in after running the business. For banks, top line is Net Interest Income; for insurance, Premiums Earned; for REITs, Rental Income.
Cash Flow Components
Four cash-related metrics quarterly: Operating Cash Flow, Free Cash Flow, CapEx, Net Income. Healthy companies have CFO and FCF tracking together; divergence is a warning.
Asset Allocation
Where the company's value lives on the balance sheet. Goodwill-heavy = M&A risk. Cash-heavy = optionality. PP&E-heavy = capital intensive.
Cash Flow Bridge
From Net Income to Free Cash Flow. Each step shows how earnings convert (or don't) into cash. Big negative bars on operating items signal accounting risk.
Series Trend Spread
Distribution of trend strength across this company's financial series. Each dot is one series (revenue, op income, etc.). Magnitude near 1 = strong trend; near 0 = noise.
Want the full breakdown? Switch to Quant view for the variance gauge, health snapshot, and the three-regime explainer.
90-day rolling variance
Health Snapshot
This is a descriptive risk tool, not investment advice. It describes how a stock’s price has behaved and how solid its fundamentals look. It does not recommend buying, selling, or holding anything, and it does not predict returns.
Turbulence - Full Analysis
Want the full breakdown? Switch to Quant view for key drivers, per-series trajectory, and the 7-layer cascade.
Per-Series Trajectory
Turbulence is computed from price-return dynamics. Per-series fundamental trajectory below shows how each financial line item is trending across the analyzed quarters - rising, flat, or falling.
What The Four Ride Bands Mean
✓ Low
Rolling-window variance is in the framework's baseline range (variance ≤ 0.10). Equity dynamics absorb shocks without amplifying them across the rolling window. No active early-warning signal.
⚠ Elevated
Variance is between 0.10 and 0.25 - above the baseline range. Whether the trajectory warms or cools over subsequent quarters is the analytical question; outcomes in the 28-year test history varied.
▲ High
Variance is between 0.25 and 0.50 - well above baseline. In the framework's 28-year test history, signatures in this band appeared in cases of both structural stress (Lehman, AIG, SVB, Enron) and rapid positive scaling (NVIDIA, early Tesla). The turbulence reading does not predict which trajectory applies to any specific company.
✗ Extreme
Variance is at the top of the framework's scale (≥ 0.50, annualized vol ~70%+) - its highest measurement band. High dispersion alone is a measurement, not a verdict. The turbulence reading does not predict outcomes for any specific company.
⚠Reference data, not a prediction. The Backtest Reference shows what the turbulence model would have classified for this ticker at a historical observation date, alongside the actual market outcome that followed. This is descriptive analysis of past data — not a forecast for any future period and not investment advice.
📊
How to read this tab: Each ticker in the backtest universe has a single Q8 observation point from its 20-quarter EDGAR history. We computed five frameworks (Turbulence, Altman Z, Piotroski F, Beneish M, Composite) at that date, then tracked the realized return over the next 1Q, 2Q, 4Q, 8Q, and 12Q. The “closest-aligned framework” is whichever signal best matched the realized outcome — not the “winner”, just the one whose classification landed closest to what actually happened.
⏳
Loading backtest data...
If this message persists, click Run Full Analysis on the left to fetch backtest data for this ticker.
Key Ratios · current value · recent-quarter trend
How to read the income statement: Follow the money from the top (Revenue) down to the bottom (Net Income). Each layer subtracts another type of cost. A healthy company should have strong margins at each layer. Watch for gross margin vs operating margin - a big gap means high overhead costs.
Income Statement - Trailing Twelve Months
Source: SEC EDGAR 10-K / 10-Q · All figures in millions USD
How to read the balance sheet: Assets = Liabilities + Equity. Always. Think of it as a snapshot of what the company owns vs what it owes on a single day. High debt relative to equity means the company is leveraged - manageable when profitable, dangerous when earnings fall.
Balance Sheet - Most Recent Quarter
Source: SEC EDGAR
Cash flow is the most honest financial signal. Net income can be manipulated by accounting choices - depreciation schedules, revenue recognition timing, write-offs. Cash flow is much harder to fake. A company that consistently generates more cash than it reports in earnings is almost always high quality.
Cash Flow Statement - Trailing Twelve Months
Source: SEC EDGAR
These are the standard financial-health metrics most tools rely on - Altman Z (a distress-zone score), Piotroski F (a 9-point fundamental-quality check), and the liquidity and coverage ratios. They describe the company's accounting condition at the filing date. Turbulence measures something different and uncorrelated: the dispersion in its equity dynamics. Reading the two side by side - accounting state next to structural state - is the point of this tab. Neither is a prediction.
Traditional Financial Health Metrics
Altman Z - Decomposition
The single Z is built from five weighted ratios. Seeing the parts shows where the strength - or stress - comes from.
Piotroski F-Score - 9 Signals
Forensic accounting screen. The Beneish M-Score is a peer-reviewed model (Beneish, 1999, Financial Analysts Journal) for flagging firms with characteristics of earnings manipulation. It examines eight year-over-year ratio changes - receivables growth vs. revenue, margin trajectory, accruals, leverage, and more - and combines them into a single composite score. A high M-Score is not a verdict - it is a screening flag for closer examination. The original 1999 study reported approximately 76% sensitivity and 17.5% false-positive rate against SEC enforcement actions on US firms.
📊
Run Full Analysis to compute the Beneish M-Score
The forensic screen pulls annual (FY) data from EDGAR filings and computes eight year-over-year ratios. Click Run Full Analysis on the left sidebar to generate the breakdown.
Peer comparison: See how this company's turbulence and traditional metrics stack against industry peers. Turbulence reveals structural patterns conventional metrics miss - comparing peer turbulence values shows where this company's variance sits relative to its sector.
AI Analyst — temporarily unavailable
The in-app AI chat is disabled pending a redesign in which you bring your own LLM with the framework’s system prompt. No questions are sent to any AI service from this page.
Pricing
Simple, Transparent Pricing
Two tiers. No setup fees. Cancel anytime. Annual saves ~20%.
Every account starts free
Create an account and you get 14 days of Investor in full — every financial, the plain-English read, the trajectory, the whole workflow. No credit card.
When the 14 days end, nothing is charged and nothing is deleted. You drop to Free, and the header shows how many days are left the whole way through.
Free
$0
free forever · email required, no card
✓Turbulence band on any US-listed company
✓The Market Analysis dashboard — the whole-universe daily read
✓The daily market brief by email
✓The weekly newsletter, in full
✓A 5-name watchlist
The reading is free. The per-company workflow — Screener, Analyzer, financials, the plain-English read — is Investor.
Every U.S.-listed company, every month, for 28 years. We checked whether our turbulence signal actually sorted the companies that later failed from the ones that survived - point-in-time, with no peeking at the future.
Educational research about historical patterns. Not a trading recommendation.
28 yrs
1998-2026 full market history
1.27M
company-months 9,411 failures
0.801
AUC, top-3,000 cohort walk-forward, no lookahead
survivor- honest
bankruptcies at −100% nothing dropped
What we tested
The idea, and the test
The idea: a company’s stock gets choppier in a characteristic way before it gets into real trouble - the same “critical slowing down” that shows up before tipping points in ecosystems, climate, and other complex systems. We measure that choppiness as the rolling variance of daily returns.
The test is deliberately unflattering. We take the entire universe of U.S.-listed companies - not a hand-picked set - and, at every point in time, ask a simple question: did the companies that went bankrupt in the next 12 months rank as riskier on this signal than the companies that survived? Every measurement uses only data available at that moment. Acquisitions count as survivals, not wins. Companies that died are kept in at −100%, so nothing is quietly dropped to flatter the numbers.
The honest number
Why we don’t show you a 0.96
You will see other early-warning work quote accuracy scores up around 0.96. We could too - on a small, curated list of famous collapses, this same signal scores that high. We don’t lead with it, because a curated list is the easy version of the test. The honest number is what the signal does across every company, including the thousands of unglamorous ones, with no cherry-picking.
What was tested
Score (AUC)
What it means
Every company, full 28-yr history
0.789
the broad, honest science number
The top-3,000 study cohort
0.801
the number we stand behind for the product
A small curated list of collapses
~0.96
the easy test - not how we describe the product
An AUC of 0.801 means: pick a company that later failed and one that survived at random, and the failing one ranked riskier about 80% of the time. 0.5 is a coin flip; 1.0 is perfect. This is a ranking statistic about the past, not a prediction about any single company’s future.
The core finding
It sizes the door - it doesn’t point you through it
Here is the single most important thing the data says. Sort every company into four turbulence bands, then look at the actual range of returns over the next 24 months. As turbulence rises, the typical outcome drifts down - but the spread of outcomes explodes. The Extreme band holds the worst wipeouts and the biggest moonshots; they look identical at the start.
Bars span the 10th-90th percentile of 24-month forward return; dot is the median. 1.27M company-months, 1998-2026.
This is why the product is descriptive, not a buy/sell call: a high band tells you the door is wide - the range of what could happen is large - not which way you’ll go through it. A second “company health” axis sharpens the downside within each band (it sorts the trapdoors, not the jackpots), which is why we read the two together rather than turbulence alone.
Across the cycle
It lights up at every crisis
The share of the market sitting in the two highest turbulence bands, month by month, for 28 years. The fast (90-day) lens spikes early and sharply; the slow (trailing-year) lens lags and smooths into a regime read. Shaded bands are the 2001, 2008-09, and 2020 recessions.
COVID is the clearest example: the fast lens hit ~97% of the market in weeks, while the slow lens barely moved - a true shock that never became a sustained regime. Both lenses ship, because they answer different questions.
Not a penny-stock trick
Strongest on the biggest companies
A fair worry about any signal like this: maybe it only “works” on tiny, illiquid junk stocks. The opposite is true. Split the market into ten size buckets and the signal is weakest on the smallest names and strongest on the largest - the reverse of a penny-stock artifact.
Turbulence-alone AUC within each market-cap decile. 0.50 = chance.
How it holds up by sector
Most validation pages skip this part. It’s the part that matters most.
The signal is not uniform across sectors. In most, turbulence and company strength score best together, so we blend them. In three we don’t — and the table shows exactly where. In each row, the highlighted value is the basis we actually use.
Sector (top-3,000 cohort)
Turbulence
Company
Two-axis
We use
Consumer Defensive
0.948
0.860
0.940
Turbulence
Basic Materials
0.862
0.888
0.909
Blend
Consumer Cyclical
0.870
0.845
0.899
Blend
Real Estate
0.825
0.784
0.877
Blend
Utilities
0.742
0.742
0.855
Blend
Energy
0.847
0.735
0.836
Turbulence
Industrials
0.445
0.818
0.663
Company (books)
Technology
0.760
0.794
0.817
Blend
Financial Services
0.744
0.743
0.796
Blend
Healthcare
0.762
0.754
0.795
Blend
Communication Services
0.724
0.749
0.782
Blend
Industrials: turbulence alone scores below a coin flip (0.445) here, so we don’t use the price signal at all — we read company fundamentals instead (0.818). Consumer Defensive & Energy: turbulence alone already scores highest, so blending in the company axis doesn’t help — we keep it simple. We also tested bank- and REIT-specific fundamental recipes to sharpen the weaker sectors; none beat the generic company-strength measure, so we use the simple one everywhere and don’t pretend otherwise.
It is not a timing trigger. The signal characterizes risk; it does not tell you when something will happen. Both axes tend to light up together, late, rather than one leading the other. Useful for describing how exposed a company is - not for trading a date.
What this doesn’t prove
This is a retrospective study of historical patterns, not a live, prospective forecast. The signal was measured against the past, not deployed in real time.
Results may not generalize beyond 1998-2026 U.S.-listed companies. Different markets or regimes could behave differently.
Walk-forward testing respects time direction but cannot test against a future regime with no historical analog.
Nothing here is a trading recommendation. Historical ranking accuracy does not imply forward predictive validity for any individual security.
The data and the code
Method
A point-in-time discrimination study over 1.27M company-months and 9,411 bankruptcies (1998-2026), survivorship-honest, walk-forward. Fundamentals use as-reported trailing figures joined as-of their filing date - no restatement lookahead.
Disclosure
The Filter Lab is a research program; conclusions follow the data. Analysis by Ryan W. Malone (independent researcher); AI was used as a drafting and engineering aid, with all results computed and reviewed by the author.
How It Works
Find it, read it, track it.
Filter Lab measures two independent things about a company — how turbulent the stock has been (The Ride) and how the books look (The Company) — and describes where it sits. Here is how the three screens work together.
FILTER LAB PHILOSOPHY
We measure. We describe. We compare. We monitor. You decide.
Every reading is descriptive — a measurement of what has happened, not a forecast or a recommendation. Filter Lab sizes the range of what could happen, never the direction, and never tells you what to buy, sell, or hold.
THE INVESTIGATION LOOP
Screener → find candidates worth a look Analyzer → the hub, where you dig into one name Watchlist → track what you keep, get flagged when it changes
You start at the Screener and spend most of your time in the Analyzer. Every result on every screen can be handed to your own AI assistant with one click.
1 · Start at the Screener
Where you begin — explore by criteria
The Screener is the exploration loop. You set the criteria — ride band, sector, fundamentals — and it returns the companies that match. Sort by turbulence, size, or any column. Use the Divergence filter to surface names where the ride is wild but the books still read solid — often the most interesting cases to investigate.
Convergence (both readings pointing the same way) is a clearer signal than a single reading alone. Click any row to open it in the Analyzer.
You chose the filters; it returned the matches — never “what fits you.”
2 · The Analyzer is the hub
Where you spend the most time — one company, in depth
Type any ticker (or arrive by clicking a Screener result) and run a full structural read. You get the two readings side by side, the risk home the company lands in, its position within its sector cohort, and the turbulence history charted over time. This is a deep read on one company at one point in time, assembled so you do not have to gather it by hand.
Read the plain-English summary first; click any ? to go deeper on a measurement. When the two readings diverge — a wild ride on solid books, or the reverse — both can be true at once; the signal sizes the range of outcomes, not the direction.
3 · Track in the Watchlist
Retention — the return trigger
Save the names you care about. The Watchlist shows the current reading for each holding and, more importantly, flags when a reading changes — a holding moving from a Low to an Elevated ride, for example. That change is your cue to open it in the Analyzer and understand why.
The Analysis tab X-rays your own basket: how it distributes across the risk homes, sizes, and sectors. It describes what you put in — it never assembles a basket or calls one suitable.
On every screen: work with your own AI
Each screen has its own Copy for AI button. It hands the data — plus a reading guide that teaches your AI assistant how to interpret it — to ChatGPT, Claude, or any assistant you use. The three do different jobs:
Screener · Copy list for AI — triage a list of names: which are worth a closer look, which to set aside.
Analyzer · Copy analysis for AI — understand one company: what each reading shows and what to investigate.
Watchlist · Copy changes for AI — review what moved in your holdings and what is worth a second look.
Filter Lab structures the problem. Your AI helps you think it through. You make the decision.
About The Filter Lab
The structural signal Wall Street doesn't measure.
What it does
The Filter Lab is a structural monitor for public companies. It measures when a company's equity dynamics are widening - unstable, dispersing return behavior - and shows that alongside its accounting fundamentals. It describes a company's current structural state. It does not predict prices, rank stocks, or tell you what to do.
Where it comes from
Complex systems under stress tend to behave the same way before they tip - their fluctuations grow and slow, a pattern studied across physics, ecology, and physiology as “critical slowing down.” The Filter Lab applies that lens to corporate finance: it watches whether a company's equity dynamics are settling or dispersing, often before the strain reaches earnings reports or credit ratings.
Descriptive by design
Most tools measure the state of a company - what the numbers say right now. The Filter Lab measures the dynamics - whether the system is settling toward stability or dispersing away from it. That shows up earlier, but it is a description of risk, not a forecast: it sizes the range of what could happen, not the direction.
We also show exactly where the signal works and where it fails - including the sectors it reads poorly. The full 28-year, every-company test record lives on the Validation page.
Use it with your own AI assistant
Every analysis can be handed to an AI assistant with one click. In the Analyzer, Copy analysis for AI copies a structured, plain-text export; in the Screener, Copy list for AI does the same for a whole filtered cohort. Paste it into ChatGPT, Claude, or any assistant and ask it to help you read what the tool measured.
The export carries three things: a short reading guide that tells the assistant how the framework works and what it does not claim; today's market context - the regime and where turbulence is concentrating; and the measurements themselves for each company - the turbulence band, the fundamental models, the seven-layer cascade, the valuation lens, and the sector peers. You can add up to ten names to one copy, so the assistant can compare them side by side.
Prompts that fit what it measures: Where do the turbulence and fundamental lenses disagree, and what in the filings would resolve it? Which two or three questions does the data say matter most before any decision? What would have to change to move the turbulence band?
One honesty note: the assistant has no more view of the future than the tool does. The export describes the measured past, so it is for characterizing a company, not forecasting it. Ask it to read the measurements against the filings, not to predict a price. Not investment advice.
Built in the open
Independent research by Ryan W. Malone. The method and its derived data are archived publicly, so the work can be checked rather than just trusted. AI was used as a drafting and engineering aid; all results were computed and reviewed by the author.