Patent statistics for decision-making.
Analysing patent statistics
There are three steps in the path from raw data to insight and knowledge:
- gathering the data
- analysing the results
- visualising the outcome.
For the first step, the old "garbage in, garbage out" maxim holds true. You must make sure that you are gathering the right data for your analysis. Defining your goals, in terms of quality and coverage, will help you select the right database and user interface (from free-to-use collections to commercial offerings). You must also set up your queries to filter the data as required, using classification information, country of applicant, etc.
In step two, you first need to clean the data and remove noise. Spelling variants of applicant and inventor names can be a challenge, and statisticians have been investing considerable efforts to improve the results in this area. You may still have to do some manual sorting, especially when dealing with large international groups spanning various legal entities with
different names. Free tools like Google refine can be useful because they include history tracking and algorithms for identifying matching data elements.
Step 3 is the visualisation step. While bar and pie charts show basic trends, bubble diagrams, 3D charts, word clouds and landscaping techniques can reveal unexpected patterns and allow you to visualise more complex dependencies. Graphical representations of technology growth rates or how an area of technology is shared among the players can help with understanding competitors' positions (see graphic).
Inventions typically have more than one publication both within a patent office and internationally and it is important to be aware of these in patent analyses. Failure to distinguish between patent applications and grants can lead to errors in your analysis.
The type of date used will also strongly influence the results. The priority date is close to the date of invention, whereas the publication date defines the state of the art and is often suitable for trend analysis. Counting patent applications is not always a good proxy for assessing the
underlying value of patent portfolios.
The PATVAL study for the EU looked at some 8 000 European patents and concluded that 90 percent of their combined value was in the top 20 percent of patents. More advanced
statistics therefore go beyond simple counting and take quality factors (e.g. citation rate) into consideration.
Espacenet can be a good starting point for an overview of the patent situation. You can then switch to more advanced tools like the Global Patent Index or PATSTAT. Until recently PATSTAT was only available as raw data and you needed advanced skills in structured query language (SQL) to use it. The new online version, currently in a beta test version, is available in two flavours. The public version provides some global statistics free of charge. The subscription version offers users a means to run some basic queries and to visualise the results without having to load the raw data into a database of their own.
Autor: European Patent Office News
Referencia: European Patent Office News, 1 (2012) pp.: 6