Metrics & Methodology

Present fair value

The headline fair valueis a consensus estimate of what the stock is worth today, recalculated daily. It blends roughly a dozen methods across four families: peer-relative multiples (P/E, P/S, EV/EBITDA, price-to-free-cash-flow), self-history (the stock's own typical multiples), intrinsic models (discounted cash flow, residual income, dividend discount, Graham), and analyst price targets. The methods are combined as a family-weighted average, and a family is down-weighted when it structurally can't price a name (for example, peer multiples for a business with no real comparable).

The 12-month figure is the present-day read. The 3-year and 5-year figures add a forward growth path and are shown as context. The discount rate is macro-aware: it moves with the live equity risk premium, the stock's beta, a credit-spread estimate, and the 10-year Treasury yield.

How to read it.Treat fair value as a reference point, not a target, and read it asymmetrically. In a long-run backtest (2014 to 2025), names trading well below fair value moved toward it about 77% of the time within a year, while names trading well above it did so only about 25% of the time. So “cheap versus fair value” is a more reliable signal than “expensive.” Over three years the gap matters more: names that were expensive underperformed the market by roughly 22% (median) versus fair or inexpensive names.

Caveats. The model tends to under-value franchise leaders at peak margins (it normalizes earnings toward a through-cycle level), so a great business can sit below its fair value with the thesis fully intact. When the peer multiple is a poor anchor, the confidence read (below) says so.

How a stock is priced (multiples)

The multiples block shows the current valuation multiple (P/E, EV/EBITDA, free-cash-flow yield) next to two benchmarks: the peer median(the same multiple across the stock's comparison group) and its own history (its typical multiple over the past several years). A premium versus peers means the market pays up for this name relative to comparable businesses; a premium versus its own history means it is expensive relative to how it has usually traded.

How to read it.Check both. A name can look rich versus peers but sit in its own normal range (often the case for a durable franchise), or look in line with peers but stretched versus its own past. For certain businesses the standard lens is wrong: equity REITs are anchored on price-to-FFO (funds from operations) and banks and insurers on price-to-book, so a raw P/E there can mislead. The multiple alone does not predict direction; quality, growth expectations, and the market regime all shape what “expensive” means.

Fair-value confidence & the bridge

The valuation bridge grades how much to trust our fair value (low, medium, or high confidence) and explains the gap between price and fair value in words. Confidence reflects whether the anchor multiple is reliable for this name, whether good peers exist, whether our earnings estimate lines up with the Street, and what is driving any divergence (cheap versus peers, aligned, a durable premium, a re-rating, or optionality).

How to read it.High confidence means the fair value is a solid reference. Low confidence is often a feature, not a flaw: it flags names where a peer multiple genuinely can't capture the business (a wide-moat compounder, or an optionality play), and the narrative tells you why. In those cases lean on the explanation rather than the number. The bridge grades the estimate and explains it; it never changes the headline fair value or the cheap/expensive label, and an analyst range shown alongside is a sanity check on our number, not a target.

What's priced in

This decomposes the current price into the expectations it embeds. The key number is implied growth: the forward earnings growth rate that would justify today's price given the peer multiple. We compare it to the consensus analyst forecast to get an expectations gap. If the price requires more growth than analysts expect, a lot of good news is already in; if it requires less, expectations are modest.

How to read it. This is transparency, not a forecast. Knowing a name needs, say, 40% annual growth to justify its price tells you what you are betting on, which helps you size conviction. It does not tell you whether the bet pays off. The expectations gap is deliberately a context read: in testing it did not beat the raw price-versus-fair-value gap at anticipating drawdowns, so we surface it to explain the setup, not as a trade signal.

What the price requires

A reverse discounted-cash-flow read on what assumptions the price bakes in, built on normalized owner earnings (net income plus depreciation minus capital spending, measured at a through-cycle margin rather than a noisy trailing quarter). It surfaces three levers:

• Price versus no-growth value: how the price compares to the business valued with zero future growth.
• Growth runway priced in: how many years of today's normalized earnings it takes to justify the price. “Beyond any horizon” means a flat business never gets there, so the price embeds real growth.
• Steady growth required: the single perpetual growth rate that makes the math balance.

How to read it.Compare the required growth to what you (or analysts) actually expect. These levers are best read as a cross-sectional rank (how demanding this name is versus others), not as precise absolutes. We deliberately do not cap the raw outputs: an extreme runway is the model telling you the price is hard to justify on normalized earnings. The leftover between price and the no-growth base is optionality, a bet with real downside that can also pay off; it is not the same as “overvalued.” Financials and REITs are excluded because a cash-flow DCF is the wrong tool for them.

Value realization (convergence)

For a stock trading away from fair value, this is the calibrated historical probability that the gap closes (price moves toward fair value) within 12 months, with a confidence interval. It is read off a lookup of how similarly-placed names behaved in a backtest spanning 2014 to 2025.

How to read it.The signal is strongly asymmetric. Deeply cheap names converged up about 77% of the time; names near fair value closed the gap only around 17% of the time; expensive names converged down only about 25% of the time. So a high convergence probability on a cheap name is a meaningful floor read with convex upside (frequent large gains, rarer large losses), while a low number on an expensive name should be read as “low downside probability at this distance,” not “due for a pullback.” The probability grades reliability, not size of move, and it assumes broadly normal markets (the edge is weaker in euphoric or crisis regimes).

Quality standing (winners screen)

A research-tested measure of business quality, scored relative to sector peers on four durable signals (higher is better on each):

• Share dilution: change in diluted share count year over year (less dilution is better). Carved out for REITs, which raise equity by design.
• Return on capital: clean return on invested capital (debt-adjusted).
• Free-cash-flow yield: free cash flow relative to market value.
• Free-cash-flow margin: free cash flow relative to revenue.

Each leg is a sector-relative percentile. Names in the top of the investable universe (market cap roughly above $2B) are tiered higher quality, average, or lower quality; smaller names or those missing too many legs are left unrated rather than guessed.

How to read it. This is a forward quality tilt, not a price prediction. In backtesting (2014 to 2024, sector neutral and placebo-validated), the highest-quality names tended toward better typical outcomes than the lowest, by roughly 6 percentage points over 12 months and around 20 over three years (median), and the signal was stable in 8 to 10 of 11 years. The edge is mainly fewer bad outcomes and a better typical result, not a higher average (the lower-quality bucket still holds the occasional lottery winner). Use it as standing and context, never as a buy or sell.

Factor signals

Factor signals are parallel diagnostics, each with its own plain read and historical track record. They are deliberately never combined into one score; mixing them dilutes what each one says. On the stock page you will see:

• Fundamental momentum: the direction and strength of the business's recent fundamentals.
• Earnings yield (value): how cheap or expensive the name is on trailing earnings relative to its cohort. Read as a characterization, not a return forecast.
• Earnings quality: the most predictive factor we have. In testing, the top earnings-quality group grew earnings about 19 percentage points faster over the following year than the bottom group, and the ranking held its direction every year from 2021 to 2025. Note this predicts future earnings growth, not price.
• Management: signals on capital allocation and execution, kept as narrative context.
• Macro sensitivity: how the stock has tended to move with specific macro drivers, with the strength of each relationship.

How to read it.Each rank is shown with its cohort (for example “top 20% in Technology”) and a plain why-line. Read the level, not the week-to-week change: changes in these ranks do not predict returns. And combined, these factors do not predict 12-month price direction; that is a coin flip, consistent with decades of research.

The value setup (quality × valuation)

Combining the value read and the quality read gives the kind of setup you are looking at:

• Cheap and robust: inexpensive with strong fundamentals, the most constructive setup.
• Cheap and fragile: inexpensive but weak quality, the classic value trap (cheap for a reason).
• Expensive and robust: a quality name priced for perfection; the premium can be deserved, but it has to deliver.
• Expensive and fragile: rich and low quality, the highest-risk profile.

How to read it. This tells you what kind of bet a name is, not whether to make it. It pairs naturally with the fragility read below.

Management scorecard

A read on how management runs the business versus sector peers, across capital allocation, earnings discipline, margin discipline, balance-sheet discipline, guidance credibility, and how the stock reacts after calls. Each pillar carries a plain label (strong, solid, or weak) and shows only when there is enough to score it; a provisional flag means some inputs are missing.

How to read it. The guidance track record is the anchor: how often management met or beat its own guidance, and by how much. It is a read on execution and credibility, not a forecast of the stock.

What management is focused on

The objectives management has stated in recent earnings calls, filings, and press releases, each tracked over time with a status (on track, mixed, behind, or newly stated), the supporting quote, and the source. As new disclosures land we mark each priority confirmed or disproved against the evidence.

How to read it.This is the difference between reading a filing once and keeping score. Watch the status transitions (“was behind, now on track”) and the proof statements. Numerical claims that the evidence does not support are flagged rather than repeated.

Earnings setup

A qualitative read on positioning going into the next earnings print (from strongly bullish to strongly bearish), built from recent analyst revisions, rating actions, price-target changes, and how tightly estimates agree.

How to read it. It describes the bar, not the outcome. A bullish setup means expectations are high, so it takes a bigger beat to move the stock; a bearish setup means a normal quarter can feel like a beat. It does not predict whether the company beats or misses. Most companies clear their own guidance most of the time, and in testing this read did not reliably forecast surprise direction, so use it as context for the print.

Thesis state

A living state on whether the reason to own a stock still holds:

• Intact (green): the fundamentals still support the thesis.
• Watch (amber): something durable slipped (for example, performance fell below the peer cohort) but has not freshly broken.
• Weakened (red): the thesis has durably deteriorated. This is the state that triggers an alert.

How to read it. States are computed from fundamentals (quality, earnings quality, factor regime), never from price moves, and each carries its since-date and a plain-English reason. An alert never says sell; it says the reason you own this changed, so re-read it while you have time to decide.

The thesis (bull, bear, breakers)

For covered names, the standing, falsifiable read: the archetype (for example, growth at a reasonable price), a conviction level, the bull case and bear case, what is already priced in, what has to stay true, and what would break it (with how severe each breaker would be).

How to read it. The value is in the breakers. They are the specific, checkable conditions that would change the story, so you know in advance what to watch for.

What to watch next

Forward-looking, falsifiable catalysts, each with a direction (upside, downside, or inconclusive), a confidence, a time horizon, a proof signal (what would confirm it), and an inverse signal (what would disprove it). Resolved items from the last 90 days are kept so you can see the track record.

How to read it. These are testable expectations, not predictions. Each one names in advance what would prove it right or wrong, which is what makes it honest.

Fragility

A risk tier (standard, elevated, or high) that flags names sitting in historically dangerous setups. It fires on three independent conditions: the price is demanding (it implies high forward growth), quality is weak (bottom half), and the sector is in a turbulent regime. High fragility means all three are true; elevated means the price is demanding plus one other; otherwise it is standard.

How to read it. This is historical cohort risk, not a forecast that a name will fall. In a 2021-onward window (a drawdown-heavy period), high-fragility names saw a roughly 49% rate of 20%-plus drawdowns versus about 33% for standard. The regime condition is load-bearing: the very same expensive, weak-quality names are far less fragile during a euphoric melt-up, so the tier is capped in those regimes. It is context for sizing and patience, not a sell trigger; the base rates are shown so you can judge for yourself.

News catalysts

Recent news graded against the company's own stated objectives: whether each item reinforces, challenges, or could break the thesis, alongside a credibility read on the source.

How to read it.The grade is relative to the thesis, not generic sentiment, so “reinforces” means it supports the specific reason to own the name. Weigh higher-credibility items more.

Material updates (8-K)

Recent SEC 8-K filings (the form companies file for material events such as leadership changes or major agreements), ranked by likely impact, confidence, and recency, with the item codes and a short summary.

How to read it. These are the official, primary filings as they land. The direction label (positive, negative, neutral) is a first read on impact, with the source linked so you can check it yourself.

Risk: how a stock moves

Volatility translated into dollars on a $10,000 position: what a typical day swings, what a rough day (around the worst 5% of days) has looked like, and the deepest peak-to-trough drop over the last year. A plain volatility label (for example, moderate or high) summarizes it.

How to read it. This is about the bumpiness of the ride while you hold, measured from past behavior. It is not a forecast; it is there so the size of the position matches your tolerance.

What actually moves stocks

Useful background for everything above. We studied the large single-day moves in a set of names and attributed them to their cause. The sector cycle drove about 26%, earnings results about 16%, and broad market or macro moves about 12%. Genuinely company-specific news (analyst actions, product, customers, deals) accounted for only around 6% combined.

Why it matters.Most big moves come from sector, earnings, and the market, which is why we put so much weight on sector regime, the earnings setup, and macro context, and why we are skeptical of tools that claim to time company-specific news. The slow “business is picking up” story rarely shows up as a single jump; it is a gradual re-rating.

How the screener works

The screener filters thousands of US stocks by the same transparent reads used across the product: sector, market cap, quality standing, the value setup, factor grades, thesis state, macro sensitivity, and value realization. Most columns are explained in the sections above; the screener only repackages those reads for browsing.

How to read it.It is characteristic-based exploration, not a leaderboard of predictions. We do not claim these names will go up; we tested that directly and no honest tool can. Use it to find names matching a profile (for example, “robust quality, reasonable valuation, sector with a tailwind”), then read each one on its stock page.

Small-cap quality + value rank

For names below roughly $2B (where the main quality tier is left unrated), the screener shows a dedicated cross-sectional rank that blends gross profitability and cheapness (with checks like the Piotroski F-score and a financial-distress flag). It is shown as a percentile: higher is more attractive on these characteristics.

How to read it. It is a relative rank within the small-cap universe, not a dollar fair value. A distress flag is a caution worth heeding.

Macro sensitivity

A label (low, moderate, or high) for how much a stock has tended to move with macro forces such as growth, inflation, and rates.

How to read it. High sensitivity means macro surprises tend to push the name around more, so the economy pages and the macro pulse matter more for it. It describes past behavior, not a forecast.

Why a name surfaced

The short chips on each row (and in the detail panel) explaining what made the stock stand out given your filters or the active preset, for example a cheap valuation gap, improving fundamentals, or strong quality.

How to read it. They are a quick orientation, not a thesis. Treat them as a starting point and confirm on the stock page.

Preset screens

Curated starting points (for example durable compounders, cheap and improving, revision leaders) that apply a sensible combination of the filters above. Selecting one shows its members with the reason each was included.

How to read it. Presets are convenience, not recommendations. They encode a profile worth exploring; the read on any single name still lives on its page.

The AI field contract

Economy pages carry four AI-written fields, and each label always maps to the same underlying field: Headline (one sentence on the current state), Summary (the why behind it), Momentum (direction and pace of change), and Risk (what could go wrong).

How to read it. If a field is missing you see a dash, never another field quietly swapped in. Each explanation is generated only from the numbers shown next to it, so it cannot invent sources or figures.

Sector regime

Each of the 11 sectors carries a regime read based on its recent performance relative to the S&P 500 (roughly the last 60 days): tailwind (leading), neutral, or headwind (trailing), plus a lifecycle view of where the sector sits in its cycle.

How to read it. Because the sector cycle drives a large share of big stock moves (see what actually moves stocks), a sector tailwind or headwind is real context for a name in that sector. It is a relative-strength read, not a price call.

Industry Business Cycle

Industries move in repeating boom-and-bust cycles. In good times the companies in an industry ramp up (hire, build capacity, stock inventory), eventually overshoot, then pull back until demand bottoms and recovers. We track this at the sub-industry level (for example Semiconductors) rather than the broad sector, because a sector averages the swings away: Information Technology looks smooth because volatile Semiconductors is blended with steady Software.

The read is fundamentals-driven (orders, revenue, capital spending), not based on the stock's price, and runs through six stages: Recovery (off the bottom), Expansion (growing), Supercycle (an unusually strong boom), Steady (high but leveling off), Deceleration (still positive, but slowing), and Contraction (shrinking). A ▼marks a quarter the sub-industry's growth turned down: amber is an early, unconfirmed watch; red is confirmed once the next quarter validates it. These turns often appear before the obvious headlines.

How to read it.A company does not trade in a vacuum: a booming sub-industry is a tailwind for the names in it, a contracting one a headwind, so the stage is context for the company's results (flat revenue during a supercycle is more worrying than during a contraction). It describes the sub-industry, not this stock, and is not a buy/sell call or a price prediction. It is the finer-grained, per-name cousin of the sector regime.

Macro pulse (stance & momentum)

A compact read on the macro backdrop across five drivers (growth, inflation, labor, policy, and risk). Each tile shows a stance (the current direction, for example expanding or cooling) with a short evidence line, and a momentum strip shows how each driver shifted most recently.

How to read it. The Fed read sits under policy, and the risk tile is a market-pressure gauge. Read stance and momentum together: stance is where things are, momentum is which way they are turning.

Economic indicators

The dashboard and the deep-dive pages (CPI, PCE, GDP, labor, jobless claims, Fed rates, yield curve) show the latest value, trend, a risk read, and the next release date, plus history, component drivers, and consensus expectations. Common terms: YoY compares to the same time last year, QoQto last quarter, and consensus is the forecasters' expectation going in.

How to read it. Direction and surprise versus consensus usually matter more than the level. Period labels come straight from the data so they are not subject to time-zone quirks.

Concentration

How much of a portfolio (or a fund) sits in its largest positions, shown as a plain label (for example moderate or high) with the top position and top-three weights, and a concentration index (the Herfindahl-Hirschman Index, which rises as holdings cluster into fewer names).

How to read it. Higher concentration means more of your outcome rides on a few names. It is a structure read, not a judgment; concentration can be intentional.

Portfolio beta

How sensitive the whole book is to the market, measured as a dollar-weighted beta versus the S&P 500 over the recent window. A beta of 1.0 moves with the market; above 1.0 is more market-sensitive, below 1.0 is less.

How to read it.It is reframed in plain terms (for example “about 20% more market-sensitive than the S&P 500”). It describes typical co-movement, not a prediction of either the market or your book.

Diversification & vs S&P 500

The spread of a portfolio across sectors (how many, and the largest), and its excess return versus the S&P 500 over a trailing window. Where a cost basis is recorded, unrealized profit and loss is shown per position and for the book.

How to read it.The versus-S&P read is historical, and is flagged when concentration makes it noisy. It is a scorecard on what happened, not a forecast.

Fund & ETF X-ray

A look-through into a fund or ETF: the real companies inside and their weights, the top-10 concentration, the sector mix, the macro forces it leans on, and the overlap with another fund or with the broad market. Headline facts include the expense ratio (the annual fee) and yield.

How to read it. The point is to see what you actually own and where the same names show up across funds, so a company that breaks its thesis still touches your book even when you hold it through a fund. Overlap with the broad market shows how much a fund simply replicates the index.

How we keep AI honest

• We never predict price direction. We backtested our own signals against future price moves and they are a coin flip, consistent with decades of research. Any surface implying a price call would be dishonest, so we do not ship one.
• Numbers come first. Every AI explanation sits next to the data it describes and is generated only from those drivers. It cannot invent sources or figures.
• Strict field contracts. Labels map to exactly one underlying field. Missing data renders as a dash, never a silent substitution.
• Receipts. Validated claims (such as earnings quality preceding earnings growth) are shown with their sample sizes and time periods, including where we have no edge.
• Precision over noise. Alerts are tuned so that when we speak, it is real. The ambiguous middle becomes a Watch state, never a forced yes or no.

Everything here is context for your own research, not a recommendation, and not investment advice. For how these pieces fit together in daily use, see the main documentation.