Imagine a company planning to develop a new battery, an industrial robot or a water treatment technology. The team faces the question: which direction should we move in next?
You can study competitors’ websites, scientific publications and industry news. But a significant part of technical information appears already at the patenting stage — long before a new development turns into a mass-produced product.
This is exactly where patent analytics is used.
It allows you not just to find similar patents, but to see a broader picture: which technical solutions appear more often, who patents developments most actively, which areas are growing and, conversely, where technological activity is declining.
The World Intellectual Property Organization WIPO defines patent analytics as the use of patent information to identify innovation patterns and trends in specific technology fields. Such data are used in decision-making in R&D, commercialization of intellectual property, licensing and technology policy.
Patent analytics with AI becomes particularly interesting. Algorithms can analyze large arrays of documents, group similar solutions, find connections between technologies and help an engineer see patterns that are hard to detect when manually reviewing hundreds or thousands of patents.
Let us look at how such work differs from ordinary search and how technology development can be analyzed on the basis of patents.
What patent analytics is and how it differs from ordinary patent search
Patent search and analytics start from the same data, but answer different questions.
An ordinary search for patent information is most often conducted for a specific development.
For example:
- Does a battery with such an electrode design already exist?
- Has a similar folding mechanism been patented?
- Are there valid patents for a similar method of liquid purification?
As a result, the specialist receives a certain set of relevant documents and analyzes them in relation to a specific technical solution.
Patent analytics looks wider. Here the question may sound like this:
- How has sodium-ion battery technology developed over the last ten years?
- Which companies are increasing their patent activity fastest?
- In which countries are most developments concentrated?
- Which technical approaches are becoming more popular?
- Which areas are still practically unoccupied?
WIPO describes the patent analytics process as a sequence of obtaining data, assessing it, finding patterns and presenting results. A separate tool of such analysis is patent landscape reports — patent landscapes showing the situation in a particular technology field.
This is how a technology patent landscape emerges.
It may include:
- number of applications by year;
- main applicants;
- geography of patenting;
- distribution of developments by technology area;
- most active IPC classes;
- dynamics of the emergence of new solutions;
- large patent families;
- links between developers and technologies.
For example, if the number of patent applications in a certain technology subcategory has been growing for five years in a row, this does not yet prove the future commercial success of the technology. But such patenting dynamics by year can be one of the indicators of growing research and investment activity.
Therefore, a patent landscape is not a forecast in itself. It is an information base on the basis of which an engineer, R&D director or investor can make better-grounded decisions.
How technology trend analysis is conducted from patent data
Technology trend analysis begins with correctly defining the area under study.
This is more important than it may seem.
For example, the query “robotics” is too broad. It will include industrial manipulators, medical robots, autonomous transport platforms, machine vision systems and dozens of other areas.
Therefore, the technology is first broken down into semantic blocks.
Suppose a company is analyzing unmanned warehouse platforms. Then the separate areas could be:
- navigation;
- route planning;
- collision avoidance;
- computer vision;
- grasping and moving loads;
- battery charging;
- interaction of multiple robots;
- warehouse management systems.
After that, a search strategy is formed.
It includes:
- keywords;
- synonyms;
- technical terms;
- patent classifications;
- time period;
- countries;
- applicants;
- inventors.
Then a search of patent documents is conducted and a data set is formed.
At the next stage, the documents are cleaned of irrelevant results, related publications are merged, and they are distributed across technology categories.
After that, it is already possible to analyze:
- Dynamics. How many applications appeared each year?
- Companies. Who forms the largest portfolios?
- Geography. Where do people most often try to obtain legal protection?
- Technologies. Which technical solutions occur more often?
- Overlaps. Which companies work in the same areas?
- Gaps. Which technical problems are still relatively weakly covered?
The result is a visualization of patent data: charts, tables, technology maps and groups of related solutions.
This is exactly the approach used in WIPO patent landscapes. For example, the organization analyzes global patent activity, its growth, the main rights holders and individual technology areas.
Searching for patent information by IPC classes
The same technology can be named completely differently in different patents.
For example, one developer uses the term “machine vision system”, another — “optical object recognition unit”, and a third describes the individual components of the system and does not use the expected wording at all.
Therefore the International Patent Classification — IPC — is used.
The IPC organizes patent documents according to technical fields. Rospatent (Роспатент) states that patent search, as a rule, is conducted using the International Patent Classification. WIPO maintains and regularly updates the IPC system at the international level.
In simplified form, the structure looks as follows: section → class → subclass → main group → subgroup.
For example, instead of searching by words alone, a specialist can identify several related classes and view documents regardless of what terminology their authors used.
Therefore IPC classification analysis is useful for two tasks at once.
The first is to improve the completeness of the search.
The second is to see the structure of the technology field.
Suppose the number of patents within a general area is growing moderately. But after distribution by IPC it turns out that one narrow subgroup has grown significantly faster than the rest over five years.
It is precisely such changes that may indicate the emergence of a new technology direction.
In practice, effective search usually combines classification indices and keywords. WIPO also uses a combination of patent classes, keywords and search concepts when forming data sets for patent analytics.
Searching for competitors’ patent documents
The second important part of the research is competitive analysis based on patents.
Suppose two companies produce the same equipment. From publicly available information it seems that they are moving in roughly the same direction.
Patent data may show a different picture.
Company A has been actively patenting new materials over the last three years.
Company B is increasing the number of developments in the field of automation.
Company C has started filing applications in new jurisdictions.
This does not automatically mean that tomorrow they will release exactly such a product. Patents may not reach commercial implementation, and individual applications may lapse.
But taken together, the information helps to see a competitor’s research direction earlier than it becomes obvious from advertising or finished products.
You can analyze:
- number of applications;
- years of their filing;
- technology classes;
- inventors;
- subsidiaries;
- patent families;
- countries of patenting;
- joint applicants;
- cited documents.
It is especially interesting to compare changes in a portfolio over time.
If a large manufacturer used to patent mainly mechanical designs, and in recent years has been increasing the number of applications for control algorithms and machine vision, this may be a signal of a change in R&D focus.
In this way patents become a tool not only of legal protection but also of technology intelligence.
Patent analytics with AI: what the neural network automates
The main problem of classical analytics is the volume of information. Suppose that after the initial search 5,000 documents have been found.
The engineer needs to:
- Review the titles.
- Study the abstracts.
- Separate out irrelevant documents.
- Determine the technical features.
- Distribute the solutions across categories.
- Find related applicants.
- Compare the technologies.
- Draw conclusions.
Doing this manually can take a considerable amount of time.
Patent analytics with AI makes it possible to automate some of these operations.
Modern algorithms are able to:
- analyze the meaning of patent text;
- identify technical features;
- group semantically close solutions;
- classify documents;
- find recurring technological approaches;
- compare developments;
- generate short descriptions;
- highlight potentially important documents;
- help structure large search results.
WIPO also uses AI tools in working with patent classifications. For example, IPCCAT automatically suggests IPC classification categories based on the content of a patent document.
But AI should not turn patent analytics into automatic prediction of the future.
If an algorithm has detected growth in a certain category of patents, this does not yet mean that the technology will become a market leader.
For technology foresight, it is advisable to compare patent data with:
- scientific publications;
- investments;
- the market;
- regulatory changes;
- technology readiness;
- cost of implementation;
- production constraints.
This is precisely why WIPO Technology Trends supplements patent analytics with scientific, economic and expert data, rather than basing a forecast solely on the number of applications.
Another useful task is searching for white spaces in the patent field.
Imagine a technology matrix: vertically — material processing methods; horizontally — types of equipment.
Most combinations are already being actively patented, but a few areas are almost empty.
This does not automatically mean that a free place for a patent has been found. The cause may be technical impossibility, lack of commercial interest or an incomplete search strategy. But such areas become points for additional research. An engineer can ask the question: Why are there so few solutions here? It is from this question that a new development sometimes begins.
How “POSINT” conducts patent analytics and builds a trend map
“POSINT” combines international patent search, analysis of technical solutions, patent research and technology reviews.
The service’s current plans separately provide for technology reviews and patent landscapes, as well as patent and technology research.
The work begins with a description of the technology.
The user does not have to compose dozens of complex search combinations on their own. A description of the technical task or development is uploaded to the service.
Next, the AI assistant helps to:
- Identify the technical elements of the object.
- Form search directions.
- Conduct a search of patent information across Russian and international sources.
- Find technically close solutions.
- Analyze the content of the documents.
- Structure the results.
- Identify recurring approaches and directions.
- Use the results to form a technology review or a technology patent landscape.
On our “POSINT” website, an example of research on a technology for deep processing of petroleum residues is shown. The service analyzed not only matching words but also individual elements of the technical solution: extraction, cavitation treatment, ozonation, filtration and other processes, after which it compared the development with the patents found.
This approach is especially useful for complex technologies, where a single solution combines elements from several technical fields.
After the array of patents has been formed, the data can be analyzed along several axes:
Time → patenting dynamics by year.
Technology → distribution of documents by direction.
Applicants → competitor activity.
Classification → distribution by IPC.
Geography → countries and territories of patenting.
Links → intersection of technologies and companies.
On the basis of this structure, a map of the patent field is formed: which directions are already saturated, which are only beginning to grow, and where potential white spaces in the patent field may be located.
Important: the system speeds up the collection and structuring of information, but strategic conclusions remain the task of a specialist. The results of the “POSINT” service are recommended to be used as a basis for analysis, and final decisions must be verified by the user.
Conclusion
Patent analytics makes it possible to turn an array of patent documents into a source of information about technology development.
An ordinary search answers the question: “What similar solutions exist?”.
Analytics goes further and helps to understand:
- who is developing the technology;
- which directions are growing;
- how patenting dynamics change by year;
- where competitors are working;
- which IPC classes are becoming more active;
- where potential white spaces in the patent field are located.
AI significantly speeds up this process: it analyzes large arrays of documents, identifies technical features and helps to structure the results.
Find out where your technology is heading
If you are planning R&D (research and development work), launching a new product or choosing a promising direction of development, conduct a patent and technology analysis before starting costly work.
Upload a description of the technology to “POSINT”. The AI assistant will help conduct an international search of patent documents, analyze analogues and form a basis for a technology review and patent landscape.
Try “POSINT” and look at your technology not as a separate idea, but as part of the global patent field.м