> For the complete documentation index, see [llms.txt](https://matterhorn-doc.mometic.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://matterhorn-doc.mometic.com/reference/glossary.md). # Glossary Plain-English definitions of Matterhorn scores, CAGR, ROIC, backlog, RPO, ARR, dilution, evidence quality, ML ranking, and historical validation. Use these definitions to interpret the output, then inspect the metric-specific source and calculation where a decision depends on precision. Company-defined metrics can differ even when they share a familiar label. ## Acceleration A change in a growth rate. Growth rising from 15% to 25% is acceleration of 10 percentage points; a higher growth rate than another metric is not automatically acceleration. ## Adjusted / reported measures Reported measures follow the stated accounting basis. Adjusted measures exclude or modify specified items. Compare like definitions and retain the reconciliation; “adjusted” does not mean inherently more reliable. ## ARR Annual recurring revenue: a company-defined annualized measure of recurring business. It is not necessarily recognized revenue, cash received, or noncancellable contracted demand. ## Available at The earliest supported public-availability time for an observation. Historical features must not use information that became public after their observation date. ## Backlog A company-defined amount of business remaining to be delivered. Scope, cancellation rights, and conversion horizon determine what the balance can tell you. ## Baseline A simpler or existing method used for comparison, such as standard financial screening, growth, quality, momentum, or a broad-market benchmark. ## Beta In “ML enhanced Beta,” the product release status. In a risk metric, beta measures sensitivity to a benchmark. These uses are unrelated. ## Book-to-bill Orders or bookings divided by recognized revenue for a specified period. Definitions and time windows must match before comparison. ## CAGR Compound annual growth rate: (ending value / starting value) raised to 1 / years, minus 1. Requires an appropriate positive starting and ending basis. It smooths the path and does not reveal drawdowns. ## CAN SLIM A seven-part growth-stock framework associated with William O’Neil: current earnings, annual earnings, new developments, supply/demand, leadership, institutional sponsorship, and market direction. Matterhorn labels the available measurements and gaps. ## Cash conversion How reported operating progress turns into cash. Working capital, capital expenditure, tax, financing, and definition choices can cause cash flow to differ from earnings. ## CIK Central Index Key: the SEC issuer identifier. It helps distinguish a company from a ticker string that can change or be reused. ## Completeness The coverage of the evidence requirements being assessed. Check the denominator and weighting; standard completeness differs from ML family input coverage. ## Compounding rate In Matterhorn, a calculated rate based on the stated capital-return/reinvestment framework or documented growth fallback, with per-share handling where supported. It is not a stock-return forecast. ## Confidence / evidence quality A description of support for the evidence or inputs. It is not a probability that a stock will rise. Analyst confidence is a separate qualitative judgment. ## Deferred revenue A liability associated with consideration received or due before revenue recognition under the applicable accounting treatment. It is not interchangeable with backlog or all future customer demand. ## Delisting Removal of a security from an exchange. Historical outcome analysis must consider the reason and treatment, including acquisition proceeds, losses, or unresolved data. ## Dilution An increase in the share count that can reduce existing holders’ participation in aggregate growth. Comparable share definitions and corporate-action treatment are essential. ## Drawdown The decline from a prior peak to a subsequent trough. Maximum drawdown is the deepest such decline within the measurement window, not the worst possible future loss. ## Durability Persistence of favorable economics. Standard Matterhorn assesses current supporting conditions; the ML target has a narrower four-quarter operating-income/gross-margin definition. ## Emergence Development of a potentially attractive opportunity. Standard Emergence uses rate, durability, and headroom; ML emergence is a learned rank signal. They are different measures. ## EV / enterprise value A valuation concept combining equity value with debt and other relevant claims, net of cash as defined. Inspect the actual calculation before comparing multiples. ## EV/sales or EV/EBIT Enterprise value divided by sales or earnings before interest and tax. Period consistency and unusual items matter; a low multiple is not automatically attractive. ## FCF / free cash flow In the common Matterhorn financial calculation, operating cash flow less capital expenditure. It is useful for cash-generation analysis but is not immune to working-capital timing or classification differences. ## FCF yield Free cash flow divided by the specified valuation denominator, typically equity market value. Check whether the metric is levered or enterprise-based before comparison. ## Filing date / reporting period The filing date describes submission/public availability; the reporting period describes the business activity measured. A quarter-end date is not the date the market learned the result. ## Forward label A later outcome used to train or evaluate a model, such as a 24-month return. It must be separate from the features available at the starting observation. ## Gross profit / gross margin Gross profit is revenue less cost of revenue under the stated definition. Gross margin is gross profit divided by revenue. Growth in dollars and change in margin are different measurements. ## Headroom A model-based plausibility assessment of remaining scale, using starting market value and implied future size. It is not a formal market-size forecast. ## Holdout A reserved set of observations kept out of model development for final evaluation. Repeatedly tuning to it destroys its independence. ## Institutional sponsorship Evidence of ownership or accumulation by institutions. Holder-register dates and reporting lags matter; it is not inferred merely from a price increase. ## LambdaRank / LightGBM LambdaRank is a learning-to-rank objective; LightGBM is the tabular tree-model implementation used for the emergence ranker. Raw outputs primarily express ordering, not return units. ## Leverage Borrowing or other financial obligations relative to a stated financial base. Debt/assets and net debt/EBITDA measure different things; cash flow and maturities provide essential context. ## Margin change The difference between margin readings, commonly expressed in percentage points. Moving from 20% to 25% is a 5 pp increase, not a 5% relative increase. ## Market capitalization Share price times the applicable shares outstanding. It changes with price, issuance, buybacks, and corporate actions. ## Missingness Information about which inputs are absent. It may be useful to a model, but the absence does not become a reported financial value. ## Momentum / relative strength Momentum measures price performance over a window. Relative strength in Movers and CAN SLIM is a percentile of measured-company returns; it is not the RSI oscillator. ## Money-weighted return An annualized return that accounts for the dates and amounts of contributions and withdrawals. Often calculated with an internal-rate-of-return method. ## NDCG Normalized discounted cumulative gain: a ranking metric emphasizing placement of more relevant outcomes near the top. It is not a portfolio return. ## Net retention / gross retention Company-defined measures of recurring-business retention. Net retention can include expansion; gross retention generally excludes expansion. Check what cohort, revenue basis, and period are used. ## OOD / out of distribution An indication that some inputs fall outside the model’s observed training support. The available diagnostic is limited; an in-range observation is not guaranteed reliable. ## Operating leverage The way operating profit changes as revenue changes, influenced by fixed and variable costs. A model feature may use a specific formula; inspect its definition. ## Percentage point / pp The arithmetic difference between two percentages or percentile readings: 30% minus 20% equals 10 percentage points. ## Percentile A relative position within a specified population. The 90th percentile describes standing, not a 90% probability or 90% return. ## Point-in-time A reconstruction using only information supported as available by the historical observation date. It guards against learning from the future. ## Provenance / lineage The trace from a source document through a fact and calculation to an output. It allows a result to be inspected and rebuilt. ## Purged walk-forward validation Time-ordered evaluation that trains on earlier mature labels, tests later observations, and removes overlapping outcome intervals that would contaminate the separation. ## Rankability The application’s classification of evidence readiness under configured rules. It is not a business-quality rating or investment guarantee. ## Reinvestment rate The portion of earnings put back into the business on a specified basis. R\&D-led and plant-intensive businesses require different accounting considerations. ## ROIC Return on invested capital: operating earnings after tax relative to the defined invested-capital base. Different treatments of cash, debt, goodwill, and R\&D affect comparability. ## RPO Remaining performance obligations: the transaction price allocated to unsatisfied or partially unsatisfied obligations under the relevant revenue accounting disclosures. Exclusions and timing matter. ## Sector-and-size-adjusted return percentile The relative position of a company’s return after comparison with observed peers of its starting industry and size. The population and missing-return treatment are part of the definition. ## Shadow mode / frozen bundle Shadow mode retains a separate model view without automatically changing production selection. A frozen bundle fixes predictions, identities, versions, dates, and integrity fingerprints for publication. ## SHAP A method for attributing a model prediction to its input features. Contributions explain that fitted model’s output; they are not proof of economic causation. ## SIC Standard Industrial Classification, used in SEC company records. It supports checklist and peer grouping but does not define an exact competitive peer set. ## SPY SPDR S\&P 500 ETF Trust, commonly used as a broad U.S. equity comparison. Use consistent dates and dividend treatment when benchmarking. ## Survivorship bias Bias caused by analyzing only companies that survived or remained in today’s universe, omitting failures, delistings, or acquired businesses. ## Time-weighted return A return measure that links subperiod performance while separating the effect of external contributions and withdrawals. ## Total return Return including price change and distributions under the stated reinvestment convention, with appropriate corporate-action handling. ## TTM Trailing twelve months: the latest twelve-month period, often assembled from four comparable quarters. It is not necessarily the last reported fiscal year. ## Volatility Variation in returns over a defined window. It measures variability, not every form of business risk or permanent capital loss. Return to [column definitions](/reference/columns.md), [methodology](/reference/methodology.md), or [your first investigation](/start-here/first-investigation.md).