Glossary

CAGR (Compound Annual Growth Rate)

Annualized growth rate of patent filings between two years. Formula: (last_year / first_year)^(1 / n_years) − 1. A CAGR of 23% means filings grew on average 23% per year over the period. Primary driver of the Momentum Index™.

HHI (Herfindahl-Hirschman Index)

Market concentration index, computed as the sum of squared market shares of all assignees in the dataset. Range 0 to 1. Below 0.15 the market is competitive (many players), 0.15 to 0.25 moderate concentration, above 0.25 high concentration. Input to the Openness Score™.

YoY (Year-over-Year)

Growth rate between one year and the previous one. YoY = (year_n − year_n-1) / year_n-1. Used alongside CAGR to detect recent trend reversals that long-term averages can hide.

Silhouette Score

Cluster quality metric in the range −1 to +1. Measures how well each patent fits its assigned cluster versus other clusters. Above 0.7 means excellent separation, 0.5 to 0.7 good, 0.3 to 0.5 moderate (read with caution), below 0.3 weak (clusters may be unreliable). Stratensight computes the silhouette on classified patents only, excluding noise.

S-curve

Technology adoption lifecycle model. Technologies typically evolve from slow initial growth, to rapid acceleration, then plateau at maturity. Stratensight uses the shape of the annual filing curve to position the technology on this curve.

P50 / P90

Statistical percentiles. P50 is the median (half of patents fall below, half above). P90 is the 90th percentile (only 10% of patents exceed it). Used in citation velocity benchmarking.

UMAP

Uniform Manifold Approximation and Projection. Dimensionality reduction algorithm that transforms 384-dimensional patent embeddings into a 10-dimensional space suitable for clustering. Preserves local and global data structure.

HDBSCAN

Hierarchical Density-Based Spatial Clustering. Identifies clusters based on data density. Unlike k-means, HDBSCAN does not require a fixed number of clusters and handles noise natively. Used by Stratensight to detect technology clusters.

peak_age

Age of the technology domain at its peak filing year: peak_filing_year − first_filing_year. Below 5 years the domain is young (emerging). Between 5 and 10 years it shows active growth. Above 10 years the domain is mature. Used to validate the Lifecycle Position™.

PCT (Patent Cooperation Treaty)

International treaty letting inventors file a single international patent application valid in 150+ countries. PCT applications are published about 18 months after the priority date. Widely used by multinationals for global IP protection.

IPC (International Patent Classification)

Hierarchical classification system created by WIPO. Divides technology into 8 sections (A to H) and around 70,000 subdivisions. This is the classification Stratensight reads first for the technology diversity that feeds the Lifecycle Position™; CPC is used only as a fallback, when a document carries no IPC code.

CPC (Cooperative Patent Classification)

Joint classification system developed by the EPO and the USPTO. More granular than IPC, with over 250,000 codes. Acts as the fallback for technology diversity when a document carries no IPC code, and feeds technology mapping.

Priority Date

The earliest filing date in a patent family. Establishes the legal priority of an invention. Critical for computing expiration (priority date + 20 years = theoretical maximum protection). Required to activate the Patent Expiration Timeline.

Filing Date vs Publication Date

Filing date: when the application was submitted to the patent office. Publication date: when the application became public, roughly 18 months after filing. Always use Filing Date for the Momentum Index™, since publication dates introduce an 18-month lag that compresses the innovation curve.

Patent Family

Group of patents filed in multiple countries for the same invention. Example: one US patent, one EP application and one PCT filing form a single family of three documents. Deduplication by family ID reduces noise and avoids over-counting a single invention.

Forward Citations

Number of times a patent is cited by later patents. High forward citations signal a foundational patent that influences subsequent innovation. Used to weight the Momentum Index™ signal.

Legal Status (INPADOC)

Patent legal event history from the INPADOC database (EPO). Records events: grant, opposition, appeal, lapse, revocation. Used by Stratensight to detect litigation signals. Opposition and appeal are procedural events, not judicial litigation.

FTO (Freedom to Operate)

Legal analysis determining whether a product or process can be commercialized without infringing existing patents. Stratensight does not provide FTO analysis. Consult a patent attorney for FTO assessments.

IPR (Inter Partes Review)

Post-grant review proceeding before the USPTO Patent Trial and Appeal Board (PTAB). Lets third parties challenge patent validity based on prior art. One of the strongest indicators of real patent litigation in the Stratensight classifier.

Momentum Index™

Composite score of filing dynamics. Computed from CAGR (35 %), filing peak recency (30 %), recent year-on-year changes (20 %) and corpus size (15 %). Range 0 to 100: above 70 the pace is high, below 45 it is low. The score states a level, never a direction — direction is carried by the CAGR alone. Not computable below 50 documents (upload) or 150 (Query Engine), below a 3-year span, or when the annual series has a gap. When the CAGR cannot be published, its component takes a neutral value.

Intelligence Grade™

Meta-score grading the available material, never the technology. Three terms: corpus size (37.5 %), temporal coverage — years with data over the full span (37.5 %) — and completeness of date, applicant and classification metadata (25 %). An academic fallback source reduces the grade by a quarter. Labels: HIGH ≥ 70, MED 45 to 69, LOW < 45. Below 45 findings are preliminary. This score does not measure how separable the corpus is: a dataset with unusable groupings can still show a high grade. Not to be confused with evidence strength, which qualifies the conclusion rather than the corpus.

Openness Score™

Measures how filings are distributed across applicants. A composite of four terms, dominated by the share of the top five applicants (45 %); the HHI carries 20 %, new entrants 20 %, and the number of distinct applicants 15 %. On highly fragmented corpora — more than 300 applicants and a very low HHI — the weight shifts to the HHI, which is more reliable in that regime. Scale 0–100: higher means less concentrated. OPEN ≥ 80 (no dominant player), CONTESTED 55–79 (active competition), CONCENTRATED 30–54 (few dominant actors), DOMINATED < 30 (one or two players control the field). The score describes concentration: it does not say whether you can enter, since freedom to operate depends on claim scope, which this score does not read.

Lifecycle Position™

Places the corpus on five maturity rungs: Inception, Early, Intermediate, Advanced, Late. Composite: age factor (30 %), filing deceleration (25 %), a growth signal derived from the CAGR (25 %), IPC classification diversity — CPC as fallback — (20 %). A decline push is added when the last three years recede, and when the peak sits in the first half of the period. The value is then blended with citation density for 40 % when available; when the measured CAGR is below 3 %, it is capped just under the Advanced rung boundary. A rung is a position, not a movement: an Advanced position on a corpus whose filings are receding is not a contradiction.

Decision Engine™

Produces the final strategic attention level using AND logic on three individual scores: Momentum, Openness and Intelligence Grade. The scale has four nested rungs — CONVERGING SIGNALS, PARTIAL CONVERGENCE, ISOLATED SIGNAL, LOW COMPOSITE PACE — plus two off-axis states where no rung is issued: INSUFFICIENT EVIDENCE (the dataset cannot support a reading) and RE-RUN REQUIRED (the analysis predates the confidence calibration and must be re-run). The certainty axis does not lower the rung by one step: it adds the Conditional marker, or withholds the rung entirely. The Intelligence Grade™, however, is itself one of the three rung thresholds, so corpus quality does enter the level shown. A level describes what the filing signals show; it is not an instruction to act. The Decision Score is a weighted display value (Momentum × 0.35, phase weight × 0.25, Grade × 0.25, Openness × 0.15): the 0.25 term is Inception 20, Early 45, Intermediate 75, Advanced 90, Late 35 — not the Lifecycle Position™ value. This score is not a gate.

Conditional verdict

Surfaces when evidence_certainty is LOW: the dataset is rich enough to compute every score, but Signal Integrity warnings limit the audit-grade confidence. The rung is kept and rendered « {level} — Conditional » in amber (for example PARTIAL CONVERGENCE — Conditional). A low certainty never moves the domain down the scale: the rung says what the signals show, the certainty says how far to trust it — two quantities, never merged into one word. Resolve the flagged Signal Integrity warnings to lift certainty. Distinct from INSUFFICIENT EVIDENCE, where no rung is issued at all because the dataset itself is unusable.

INSUFFICIENT EVIDENCE

Reserved label emitted only when the input is genuinely unusable: typically fewer than 10 patents, temporal coverage below 1 year, or structurally corrupted data. No attention level is issued — no signal can be defended. Distinct from a Conditional verdict (which means the dataset is rich but flagged with Signal Integrity warnings).

Signal Integrity

Self-audit layer that runs thirteen deterministic checks on every analysis before display. Surfaces mathematical contradictions, data quality artifacts and scoring inconsistencies. Critical findings do not block the verdict but must be reviewed.

White Space Score™

Measures uncontested technological sub-domains inside a cluster. A high score signals filing corridors with low competitive density — potential entry points for new IP. A score of 100 means no dominant assignee in this sub-domain.

Critical Reader

AI auditor (Claude Haiku) that reviews each analysis for logical inconsistencies, missing data contradictions and methodology gaps. Generates findings in the user’s language with impact assessment and recommended actions. Filtered through hallucination guards before display.

Filing Intelligence™

Competitor filing profiles surfaced inside the Competitive Landscape. Click any patent owner to open their filing trajectory — CPC focus areas, jurisdictions and recent filings. Included in every analysis.

Portfolio Positioning™

Upload a client patent portfolio to see where it sits within each technology cluster, and where the uncontested White Space lies. Matched patents are overlaid on the cluster map. Available on Pro and above.