IP Signals · Data note

How we count

Standing methodology for every IP Signals issue. Official registers and EPO/UPC publications always prevail.

Bulletin coverage

Patents. Counts come from indexed TÜRKPATENT patent bulletins in Validas Patent Intelligence (currently 2015_06 onward). MoM and “vs 12-bulletin average” use the last twelve indexed bulletins, not a calendar year.

Trademarks. National publication totals use Brand Intelligence bulletin indexes. Coverage expands as older issues are parsed; each issue states the bulletin range in force (e.g. 384–501). Gaps are disclosed, not interpolated.

OCR normalization

Bulletin list layouts are OCR-noisy. Party names, addresses and transfer pairs can fragment across strings. We report clusters as approximate until cleaned; we do not invent merged legal entities from OCR variants.

Applicant alias logic

Company spotlights and “who accelerated” lists use matched applicant/owner keys in the index. Extreme ratios can reflect alias splits, late bulk publications, or incomplete prior windows — not only real filing surges. Unverified name clusters are anonymized (IPC / sector only) until a human check confirms the string.

What “activity” means

Patent “activity” is observed event volume in the indexed bulletin window (publications, grants, transfers, restorations, and related lifecycle events as classified in Patent Intelligence) — not a complete portfolio inventory and not a recommendation.

Trademark “publications” are national (and separately Madrid) records appearing in the stated bulletins. Concept tags (e.g. AI/ML) come from the Brand Intelligence concept layer at the tier named in the issue.

UPC coverage methodology

The UPC section uses the Validas research layer (parsed dockets and events), not the court’s official statistics feed. Each issue states dataset version, case/event counts, and coverage versus the official catalog. Until coverage is complete, we prefer durable metrics — most active division, infringement vs revocation, median events per active case, busiest dockets, first-instance → appeal transitions — over raw monthly event totals. Layer activity is never presented as official court volume.

Law & procedure strip

External items are official sources only (EPO Selected, Abstracts, Official Journal, EBA pending referrals; later EUIPO / PCT-Madrid when wired). Each candidate carries title, date, type, why it matters, official URL, confidence, and include_candidate. The published strip holds 2–4 short signals; empty slots stay empty.

← IP Signals · October 2026 issue

Not legal advice. Methodology may tighten as indexes deepen; material changes will be noted here.