Introduction

One of the core competencies in patentability search is building a patent search strategy for locating prior art documents that may destroy novelty or inventive step. However, this is far from simply typing a few keywords into a search box. Globally, there are over 150 million patent documents, plus an immense body of non-patent literature — if the search strategy is inadequate, you will either be overwhelmed by massive results or, worse, miss critical documents that lead to fatal misjudgments.

This document establishes a complete patent search strategy methodology, covering everything from keywords to classification numbers, from search query construction to multi-database coordination, helping you achieve a search that is comprehensive, precise, and traceable.

Three Core Principles of Patent Search Strategy

Before unfolding the specific methods, let us first establish three principles:

Principle 1: Prioritize Comprehensiveness

The essence of patentability search is “finding negative evidence” — missing just one critical document can overturn the entire patentability conclusion. Therefore, a patentability search should err on the side of a slightly broader scope with somewhat more noise, rather than over-refining for the sake of “cleaner results.”

Principle 2: Multi-path Cross-validation

Do not rely on a single search method. Keyword search, classification number search, citation search, competitor search — multi-path cross-validation is the only way to minimize the risk of omission.

Principle 3: Full-process Traceability

Record the search query, database, date, and hit count for every step of the search. This is not merely a mark of professionalism — in subsequent patent examination or invalidation proceedings, a complete search record itself constitutes important evidence.

Methodology 1: Keyword Strategy

Keyword search is the most fundamental yet also the most error-prone step. The core challenge is this — the same technical concept can be expressed in vastly different ways across different patents.

Six Major Sources of Keywords

Building a Keyword Matrix for Patent Search Strategy

For each core technical concept of the invention, establish a keyword matrix:

Core technical concept: Heat Dissipation

Using Truncation and Wildcards

• * (multi-character truncation): cool* matches cool, cooling, cooled, cooler, coolant
• ? (single-character wildcard): wom?n matches woman, women
• Note that truncation/wildcard syntax may differ across databases

Boolean Operator Combination Strategy

(heat dissipation OR thermal management OR cooling OR 散热 OR 热管理 OR 冷却)
AND
(graphene OR 石墨烯 OR carbon material OR 碳材料)
ANDNOT
(PCB OR printed circuit OR 电路板)

• OR is used between synonymous concepts (expands coverage)
• AND is used between different concepts (increases precision)
• ANDNOT/NOT is used to exclude obvious noise areas

Methodology 2: Classification Number Strategy

The inherent limitation of keyword search is that you can only search using “keywords you know.” If a prior art document describes the same technology using entirely different terminology, the keyword search will fail. Classification number search exists precisely to solve this problem.

Three Major Classification Systems

IPC (International Patent Classification)

• Scope: Global
• Structure: Section → Class → Subclass → Main Group → Subgroup (hierarchical)
• Example: H01L 23/34 (Cooling arrangements for semiconductor devices)
• Advantages: Broadest coverage, globally unified
• Disadvantages: Coarser granularity, slower to update

CPC (Cooperative Patent Classification)

• Scope: Based on IPC but more granular
• Example: H01L 23/473 (Cooling by liquid flow)
• Advantages: More refined (CPC has approximately 250,000 entries, IPC approximately 70,000), updated more frequently
• Disadvantages: Not all countries use CPC indexing

FI/F-Term (Japan-specific)

• Scope: Japanese patents
• Characteristics: Even more granular than IPC/CPC; multi-dimensional classification (indexed from multiple perspectives such as technical purpose, function, material, etc.)
• If the target market includes Japan, FI/F-Term search must not be overlooked

How to Find the Correct Classification Numbers

Method 1: Start from known relevant patents

1. Perform a preliminary search using a few core keywords
2. Locate 3–5 of the most relevant patents
3. Examine the classification numbers of these patents
4. Confirm their meaning and hierarchical relationships in the official classification definition system

Method 2: Use classification number search tools

• WIPO IPC search: use this to browse IPC classifications and confirm official classification scope
• USPTO CPC search: use this to browse CPC classifications and related definitions
• Search directly for keywords you understand and observe which classification numbers are returned

Method 3: Analyze the classification hierarchy
Once a main class is found, browse its subordinate classifications to confirm whether a more precise classification number exists. For example:

• H01L 23/00 (Details of semiconductor devices)
• H01L 23/34 (Arrangements for cooling, heating, ventilating or temperature compensation)
  • H01L 23/473 (Cooling by liquid flow)

Combined Classification Number + Keyword Search

Best practice is to combine classification numbers and keywords:

IPC=H01L23/34 AND (graphene OR 石墨烯 OR carbon nanotube OR 碳纳米管)

First use classification numbers to define the technical field (ensuring nothing is missed), then use keywords to focus on the specific technology (improving precision).

Methodology 3: Citation Search

Citation search leverages the citation relationships between patents to discover relevant documents.

Forward Citation Search

Identify a key patent and examine which later patents cite it. This can reveal:

• Subsequent improvements building on that technology
• The level of competitor interest in that technology
• Patents in the same field that may be relevant to your invention

Backward Citation Search

Examine which prior documents a key patent cites. This can reveal:

• “Foundational” patents in the technical field
• Prior art that the examiner referenced when examining that patent

How to Perform Citation Search

Most patent databases support citation search. In Google Patents, view the “Cited By” and “Citations” sections; in Espacenet, view “Cited documents” and “Citing documents.”

Methodology 4: Competitor / Patentee Search

Sometimes, rather than searching broadly, it is better to go directly to the entities most likely to hold relevant patents.

Identifying Key Patentees

• Direct competitors: Companies making similar products to yours
• Technology leaders: Leading enterprises in the technical field
• Active applicants: Entities with the highest filing volumes under the relevant IPC classification
• NPEs/PAEs (Non-Practicing Entities / Patent Assertion Entities): Entities that specialize in holding and asserting patents

Search Approach

Most databases support searching via the applicant/assignee field:

AN=("Tesla" OR "Panasonic" OR "CATL") AND IPC=H01M*

Multi-database Coordinated Search Strategy

No single database covers all global patent documents. The following coordinated strategy is recommended:

Iterative Refinement of Search Queries

In practice, multiple rounds of iteration are typically required:

Round 1: Broad preliminary search

• Use 2–3 core concept groups, connecting synonyms with OR, without too many restrictions
• Observe hit counts, language distribution of returned documents, and patentee distribution

Round 2: Targeted deepening

• Based on preliminary search results, lock in the most valuable classification numbers and keywords
• Search each classification number + keyword combination separately

Round 3: Precision supplementary search

• Conduct targeted supplementation for knowledge gaps identified during comparative analysis

Round 4: NPL supplementary search

• Beyond patent documents, search non-patent literature

Common Search Errors

Error 1: Using only keywords without classification numbers

Risk: The same technology is described using different terminology in different patents, and keyword search will miss them.
Solution: Combine classification numbers + keywords.

Error 2: Overlooking synonyms and broader/narrower concepts

Risk: Keyword coverage is too narrow.
Solution: Build a keyword matrix for each core concept.

Error 3: Insufficient search time range

Risk: Missing early foundational patents or recently published applications.
Solution: The search time range must cover the full span from the technology’s origins to the present.

Error 4: Searching only one database

Risk: Different databases have different coverage scopes and data quality.
Solution: Cover at least Espacenet + the patent office database of the target market country.

Error 5: Ignoring non-English-language documents

Risk: Patent documents from non-English-speaking countries such as China, Japan, and Korea contain a large volume of high-quality comparative technology.
Solution: Any global search must cover key non-English patent document sources such as China, Japan, and Korea.

Key Takeaways: Patentability search requires a four-pronged approach — Keywords (synonyms + broader/narrower concepts + multilingual), Classification Numbers (IPC/CPC/FI/F-Term), Citations (forward and backward), and Competitors/Patentees. Employ multi-database coordinated search and record the iterative process throughout. A strong patent search strategy is built on three principles: prioritize comprehensiveness, multi-path cross-validation, and full-process traceability.