How does it work? Taking citation analysis to the next level

Past to present

The problem with current methods

Practiced since the 60's
Researchers have found correlations between citations and expert judgement, auction values, litigation likelihood, and much more, but the signal has always been noisy.
Structural biases cause noise
There are many structural biases in the citation network that artificially manipulate citation counts. To make citation analysis work, we neet to take the finger off the scale.
Wrong conclusions
Without properly correcting for structural biases, any citation based metric will deliver noisy results at best and plain wrong results at worst.

Making citation analysis worth your time

What structural biases are we dealing with?

Tech domain

Citation practices and average citation counts vary strongly per domain. The better we are at determining a patent's topic, the more reliably we can compare it to its true peers. Here, Focus uses NLP models rather than IPC/CPC codes.

Region and size

Because of regulatory differences, U.S. patents tend to get cited more than their EU counterparts. There are similar differences in other regions. Also, the larger the family, the more often it tends to get cited for reasons unrelated to quality.

Examiners

Examiners tend to cite patents they have examined or cited in the past. There are also large differences between individual examiners. To get a fair score, all of this must be corrected.

Applicants

Applicants tend to cite patents from the same applicant and the same goes for inventors. Some cite a lot, and some very little. This must be corrected to fairly compare patents.

Get the research

Beyond counting citations

Correcting citation counts for biases alone already results in a much more reliable metric than what is available today. But why stop there? Rather than counting citations, we measure intergenerational impact. This means that we weigh the importance of a patent's ancestors, the patent itself, and its offspring.

Better, sooner

Altogether, this results in major improvements and allows you to spot important patent families significantly better and years earlier than previously possible. Focus helps you identify important assets before they become obvious to everybody else.

Schedule demo

If subsequent generations die out, only counting direct citations can lead to wrong conclusions.

By also taking indirect citations into consideration, we can measure patent importance more accurately.

How do we compare to the competition?

Focus

The Rest

Intergenerational influence

Topic determination

NLP algorithm

IPC/CPC codes

Corrects applicant bias

Nope

Corrects examiner bias

Corrects inventor bias

Nope

Common pitfalls

For IP teams

Paying maintenance fees for weak patents.
Overlooking important out-licensing candidates.
Signing unfair licensing deals.

For your Business

Missing out on partnering opportunities.
Failing to spot major technological developments.
Misidentifying key inventors.

Common Questions

Let's put your mind at ease.

How quickly can you measure importance?

Citations are added over time and there is a lag between citing and publishing. This means that we cannot predict patent importance, but we can measure it reliably years before any other company is able to. As soon as there is a notable difference between the expected and the measured impact, we will see it.

Why not just use IPC/CPC codes?

At a high level, IPC/CPC codes throw all sorts of technologies into the same basket. At a detailed level, arguably similar technologies end up in different baskets. Natural language processing algorithms have made such good progress over the past few years that they are now much better at determining patent topics than IPC/CPC codes.

Where do you get your data from?

Our patent data is provided by IFIclaims, who provide over 125 million patent records from over 98 countries. IFIclaims also provides their data to services such as Google Patents, Questel's Orbit, and many more.

Can patent importance even be measured?

Patent importance is clearly context dependent. However, given a context, our algorithm can help you identify the most useful patents in any scenario. We can show you the subset of most likely important patents, which will save you countless hours of reading patents unnecessarily. After showing you the subset, it is up to you to identify the patents best suited for the job.

How do we measure importance?