Patent intelligence for policy and business decisions.

Three ways to work with us. Two of them cost exactly what it says here; the third we quote after one conversation.

Every report on this site is readable in full — no form, no email, no trial. Read one now

queries.sql — the SQL behind every number ships with the report. Disagree with a figure and you can re-run the statement that produced it.

EPO PATSTAT Global, Spring 2026 · counted at DOCDB family level

Report Start here

We run it. Once.

from

1,500 €

one commissioned analysis

Book a scoping call

Toolchain

You run it. Continuously.

flat

6,000 € / year

or 500 € per month

See what is included

Custom Solution

We run it with you.

On request

after one conversation

Describe the problem

20 published reports.

Read any of them end to end. No form, no email, no trial.

All published reports

Analysed subjects — not customers

Patent families analysed. Every name above is the subject of an analysis, never a client of ours — we do not publish client names, and nothing on this page implies one.

What you can buy

Three mechanisms, not three sizes of one thing. A Report answers the question once; you give up asking the next one without us. The Toolchain gives you that ability and you take on running the analysis. A Custom Solution is a cadence, and the only one of the three we cannot put a number on. Four reports cost what a year of running the Toolchain yourself costs — that is the whole choice, arithmetically.

Commissioned analysis

Report

We run it. Once.

One question, answered against EPO PATSTAT Global and written up so it can go in front of a board. Scope and price fixed in writing before the first query runs.

What is included

Applicant scope rebuilt across every name variant, including non-Latin office renderings
Classification search built and checked layer by layer, not typed into a keyword box
Eight to fifteen sections, every chart with an accessible data table behind it
queries.sql handed over with the report. English or German.

The stack behind the reports

Toolchain

You run it. Continuously.

The same stack that produced every report on this site, connected to your own AI agent.

In development
A white-label platform hosting your reports under your own brand. It is being built, it is not part of the subscription today, and we would rather tell you than sell it to you.

What is included

patstat-mcp — EPO PATSTAT Global on BigQuery, over MCP
DPMA register and EPO OPS — the German national route, plus legal status
Classification and applicant reference data: full IPC and CPC hierarchies, definitions, reclassification chains
The report generator (question → multi-layer search → SQL → sections → QA), methodology guides, onboarding and support

Ongoing programme

Custom Solution

We run it with you.

For when a single report is the wrong shape: several subjects, one subject tracked over time, or the toolchain run alongside your team. Quoted after one conversation, in writing.

This is the one price we cannot print, because it depends on a question we have not heard yet.

What is included

Several reports, or one subject followed across PATSTAT editions
Toolchain access with us working alongside your team
Scope, cadence and price agreed in writing before work starts

You already pay for a patent database. Here is what the 1,500 € buys on top of it.

A subscription gives you a search box. What it cannot give you is the thing that decides whether the answer is any good: whether the search itself was right. Every case below is measured, and every one comes out of a report published on this site.

Case 01

HARTING Technologiegruppe · applicant name variants, across 40 han_name groups

One company, sixty-five spellings

PATSTAT does no entity resolution: a company exists in it as a set of spellings. HARTING has 65 of them, scattered across 40 of PATSTAT’s own han_name groups — so picking the largest group, or even the largest three, returns a fraction of the portfolio. Sum the variants naively and you get 1,422 families; sum the groups instead and you get 1,239; deduplicated at family level the portfolio is 704.

Six to seven of those families are reachable only through the Chinese renderings of the name — six under 哈廷电子有限公司及两合公司 and one under its machine transliteration, which the report treats as the ceiling rather than the count. A further variant, the typo “HARTIN Electric”, no %HARTING% search reaches at all.

Read the HARTING report

Case 02

40.1%

Machine learning from Germany · 6,361 families carrying no core AI classification

Two fifths of the field, outside the obvious codes

A landscape built on the core AI classifications finds 9,479 German families. A further 6,361 carry no core code at all — they are tagged where the machine learning is applied: engine control, adaptive control, speech recognition, radar signal processing, wind-turbine control, and over 200 further CPC symbols. That hidden set is 40.1% of the true total, it misses at least three families in ten in every single year, and the report is explicit that 6,361 is a floor, not a total.

Read the machine-learning report

Case 03

478

Berlin 2019–2024 · families recovered from the German national register

The applicants European-route data cannot see at all

PATSTAT geocodes applicants only along the European filing route. Of 3,778,695 German person records at nuts_level = 0, only 4.4% carry even a postcode, and around 70% of German national families never take the EP or PCT route. Reading the German register directly and bridging back by file number, not by name, recovered 654 Berlin families, 478 of them (73.1%) outside the European-route view. For IAV alone: 14 families visible on the European route, 69 in the register. The report calls 478 a floor.

Read the Berlin report

13: numbered QA checks per report — four before a file is written, including a hallucination check that requires every claim to trace back to a query.
15: past defects written down in our own repository with a prevention rule each — one of them inflated annual family counts by 37% before it was caught.
queries.sql: ships in every report. If you disagree with a number, run the statement that produced it.
Named editions: every report states the PATSTAT and IPC/CPC editions it was computed on, and keeps them.

Open one and check the method

All 20 reports published here, newest first, with the numbers they produced. Every one carries a methodology section — where the argument above is settled in public, for anyone to pull apart.

Technology Landscape 2000–2016 in German

Erst Patent, dann Paper? Neuheitsschädliche Vorveröffentlichung an deutschen Hochschulen

3.03% self-collision · Fraunhofer 1.15% · −9.2 points on grant rate

Patent intelligence for policy and business decisions.

20 published reports.

What you can buy

Report

Toolchain

Custom Solution

You already pay for a patent database. Here is what the 1,500 € buys on top of it.

One company, sixty-five spellings

Two fifths of the field, outside the obvious codes

The applicants European-route data cannot see at all

Open one and check the method

Erst Patent, dann Paper? Neuheitsschädliche Vorveröffentlichung an deutschen Hochschulen

Apple Inc. — Patent Filing Strategy 2020–2024

Drohnen (UAV) — Globale Patentlandschaft 2014–2024

Water Electrolysis — Patent Landscape 2000–2024

VEGA Grieshaber KG — Patent Portfolio & Competitor Landscape

Autonomous Commercial Scrubber-Dryers — Patent Landscape 2019–2024

HARTING Technologiegruppe — Patent Portfolio Analysis 2014–2024

Adjustable Aerodynamic Systems (Spoilers) for Motor Vehicles — Technology Landscape 2014–2024

University Patent Activity in Saxony & Thuringia (2013–2024)

Patent Activity in Berlin (2019–2024)

TU Dortmund — Patent Portfolio in Transition (2013–2024)

Machine Learning Patents from Germany — Technology Landscape 2014–2024

TU Dortmund — Patent Portfolio Analysis 2019-2024

Rare Earth Elements — Patent Technology Landscape

Assistive Robotics for People with Special Needs — Patent Landscape

TU Chemnitz — Patent Portfolio Analysis 2019-2024

B. Braun Melsungen AG — Patent Strategy Analysis 2014–2024

Medical Technology — Technology Leaders Europe

Airbus SE — Patent Filing Strategy 2014–2024

AI Diagnostics MedTech — Patent Analysis

The questions we actually get

Two numbers and one conversation.