Sources for technology and business insights explained, part 3/7: Patents

Estimated reading time: 12 minutes

About this series of articles

This is the third in a series of seven articles. In the first two articles, I talked about how you can use venture capital news and market analyses for technology, business, and innovation insights.

This article is about patents. Specifically, I’ll talk about:

  • Where you can find patents
  • Why patents are odd ducks
  • What kinds of insights you can expect to get
  • Things you might not find
  • Strategies for searching patents so you get useful insights

This article is not about freedom-to-operate analyses. Rather, it is primarily intended for people who operate at the intersection of science, technology, and business, and who see patents as one (but not the only) source of information.

Where to find patents

There are a number of free services you can use to find patents. Google Patents, for example, or Espacenet, which is operated by the European Patent Office, or EPO. The EPO has the biggest patent database in the world. Its database collects and aggregates patent documents from patent offices worldwide.

Mergeflow gets its patent data from the EPO. Our algorithms include several analytics capabilities that the free services mentioned above do not have. For instance, Mergeflow tags all incoming patents with a range of emerging technologies from across several technology sectors and industries. I will come back to this below.

Furthermore, since Mergeflow collects and analyzes a range of other data sets across science, technology, and business, you get an “un-siloed” view. For example, if you find an inventor in patents, you can then check if this inventor has written scientific publications, or has started a company. And Mergeflow assigns patent classes to all non-patent documents that it collects. This gives you another way to cross-reference data from patents with other findings.

Patents are odd ducks

There are a few things that set patents apart from most other types of information.

Patents typically don’t get published right away. When someone files a patent, it usually doesn’t get published right away. In most jurisdictions, you have up to 18 months between filing and publication date.

Patents have families. You can file a patent in different jurisdictions. For example, you can file a US patent, or a German patent. You can also file a worldwide patent. Or you can file a patent in the US first, then in Canada, and then worldwide (there are restrictions on what you can do here, but I won’t go into this here). Let’s say you file a patent in three different jurisdictions. Then these three filings constitute one patent family. This is important for counting patents. Do you count individual filings, or families? (Mergeflow counts families)

Patents may use obscuring language. When you write a scientific publication or a news article, you try to write so that your audience understands you. With patents, the opposite may be the case. Here is why: When you file a patent, your goal is to exclude others from making, using, or selling your innovation. Now, others may come and try to challenge your patent. For example, they may say that you didn’t actually invent what you claim to have invented. But in order to challenge your patent, they first have to find it. And you make it harder for them to find your patent if you use obscure language. There are legal limits as to how far you can take this, but it’s important to be aware of this when searching patents.

What you can expect to find in patents

Like I said above, my goal here is not to explain how you can do freedom-to-operate analyses. Instead, I will talk about two other questions that patents can help you answer:

  • How to find people who work on a certain technology, either generally or at a specific company.
  • Find out more about what a certain company does.

Let’s dive in.

Who works on [x] at company [y]?

Like the iPad Pro, the iPhone 12 Pro also has LiDAR in it. You can use LiDAR to create 3D maps of rooms, for example.

Now, let’s say I want to know who works on LiDAR technology at Apple (we explored this in a previous article in our blog). They don’t tell you this in their website. And LinkedIn does not really help here either (I just tried). So instead, I simply did this in Mergeflow:

  1. Search for LiDAR AND Apple.
  2. Create a “person” network graph. Mergeflow extracts the names of inventors from patents, and a graph then shows me who co-authors patents with whom.

Here is the result (click on the screenshot to see a larger version):

A "person" graph generated by Mergeflow. The graph shows inventors who work on LiDAR at Apple. Stronger connections between any two inventors mean that they co-authored more patents.
A “person” graph generated by Mergeflow. The graph shows inventors who work on LiDAR at Apple. Stronger connections between any two inventors mean that they co-authored more patents.

Now, with this information, I can go back to LinkedIn, for example, and look up inventors to whom I want to reach out.

Of course, I could also run more sophisticated searches. There might be patents that do not use the term “LiDAR” but that are still relevant. In order to expand my search, I could use (some of) the patent classes from my results above. Since patent classes are a bit cryptic to non-experts, we use more human-readable labels in Mergeflow instead:

Patent classes for "LiDAR AND Apple" patents in Mergeflow. "Measuring length" corresponds to G01B, "measuring distances" to G01C, and "optical systems" to G02B (Mergeflow shows you the patent classes when you hover your mouse over a tag).
Patent classes for “LiDAR AND Apple” patents in Mergeflow. “Measuring length” corresponds to G01B, “measuring distances” to G01C, and “optical systems” to G02B (Mergeflow shows you the patent classes when you hover your mouse over a tag).

I marked three patent classes that you could explore further. For example, you could run another search like this:

Apple AND optical systems (G02B) AND measuring length (G01B) OR measuring distances (G01C).


Related: How Mergeflow’s search syntax works


Here is the person tag cloud that I get when I run this search in Mergeflow:

Person tag cloud for a search on Apple AND optical systems (G02B) AND measuring length (G01B) OR measuring distances (G01C).
Person tag cloud for a search on Apple AND optical systems (G02B) AND measuring length (G01B) OR measuring distances (G01C).

Next, let’s see how you can use patents to find out more about what a company of interest does.

What does company [z] do (works best for big companies)?

In the previous article of this series, I described how you can use market analyses to find out where (= in which markets) a company is active. You can also use patents to discover more about what a company does.

Let’s look at an example of how to do this. As in my article on market analyses, I use Honeywell as an example company. As the basis for my analysis, I used patents from the last five years where Honeywell is the applicant–more than 3,000 patent families. Now, I don’t want to read all these documents in order to find out what Honeywell is doing.

Discovering emerging technologies and the inventors behind them

In Mergeflow, I have several options for how to structure and analyze these data. I decided to use the set of emerging technologies that Mergeflow assigns to its contents. These emerging technologies are from across industries and tech sectors, such as computing, materials, energy, life sciences, and manufacturing. We define “emerging” as “showing strong momentum”, not as “brand-new”.

As a first step, I get a tag cloud of emerging technologies. Bigger font size means more patents, and I underlined in red the most-frequently-mentioned emerging technologies:

Tag cloud of emerging technologies, assigned by Mergeflow to Honeywell patents from the past five years.
Tag cloud of emerging technologies, assigned by Mergeflow to Honeywell patents from the past five years.

As a next step, I could now zoom in on any of these emerging technologies. For example, I could click on “3D Printing”, and then bring up a network graph of (some of) the inventors:

Network graph of 3D printing inventors at Honeywell, screenshot from Mergeflow.
Network graph of 3D printing inventors at Honeywell, screenshot from Mergeflow.

Stronger lines between any two inventors mean that they are co-authors of more patents.

Next, I could click on any of the inventors to see their patents. Or I could look them up on LinkedIn (turns out that some of them don’t work at Honeywell anymore).

Discovering interrelations between emerging technologies in patents

As another example, I used a set of Toyota patents from the past five years. But now, instead of a tag cloud, I used a network graph of emerging technologies. Such a network graph provides provides more insights than a tag cloud.

Below is the resulting network graph. Whenever a patent is relevant to two or more emerging technologies, links are added to the graph. For example, if a patent is relevant to “biosensing” and “biometrics”, Mergeflow draws a line between these two technologies. Thicker lines mean that more patents are assigned to the technologies. For example, more patents are assigned to “solar energy” and “solar panels” than to “fuel cells” and “microfluidics”.

How emerging technologies relate to each other in Toyota patents from the past five years. Screenshot from Mergeflow.
How emerging technologies relate to each other in Toyota patents from the past five years. Screenshot from Mergeflow.

What you might not find in patents

Above, in the section why patents are odd ducks, I said that patents usually get published only up to 18 months after being filed. In other words, patents aren’t the best source for the most recent information. Depending on where (= in which jurisdiction) you file a patent, you can publish your invention first and then file a patent, within certain constraints. This means that scientific publications may be a more timely source, for example. Of course, if someone doesn’t publish and only file a patent, then you have no other choice.

Also, the deal with patents is that you get exclusive rights to make, use, or sell your invention–but you have to disclose your invention. This means that in order to enforce your patent, you have to be able to tell whether someone might violate it. And there are types of inventions or technologies where discovering patent violations is difficult. Software or manufacturing processes often fall into this category, for example (and yes, there are exceptions).

If disclosing an invention is too risky because it’s too hard to detect potential violations, the inventor may decide against filing a patent. And if this is the case, it might be easier to find a company providing the technology than details about the technology itself. For example, if somebody has figured out a way to automate bookkeeping and builds a company around this idea, you’ll probably find the company but no patents (assuming the inventor determined that detecting patent violations would be too hard).

Good search strategies for patents

I have said this before, but it is worth repeating:

Searching is not about what you want to find. It’s about where and how other people most likely talk about what you want to find.

In this article I talked a lot about how you can use patents to find information about companies. In other words, if the starting point of your search is a company name, this will likely work quite well.

Above, in the “patents are odd ducks” section, I said that patents don’t necessarily describe things in clear terms. Rather, patents often circumscribe a technology, rather than just naming it.


Related: Using patent classes for concept search


For example, let’s say you are interested in flexible displays. Now, it is quite likely that there are patents that are relevant to flexible displays, but that don’t use this term explicitly. How can you find these patents? Here is a possible strategy for a search in Mergeflow:

Start with a simple search for “flexible displays”. Switch the tag cloud to “patent class”. Your result will look like this:

Patent classes tag cloud for "flexible displays" in Mergeflow.
Patent classes tag cloud for “flexible displays” in Mergeflow.

For patent class tag clouds in Mergeflow, we use short labels that are easier to read than patent classes. For example, rather than using “G02F” as a tag, we use “displays”. But when you move your mouse over a tag, you get the patent class description (we use CPC patent classes):

Mergeflow uses mouseover texts to display the full description of CPC patent classes.
Mergeflow uses mouseover texts to display the full description of CPC patent classes.

Next, search for an interesting patent class. In our case, “displays”, or “G02F”, is an obvious candidate. Of course, this by itself would be too broad, but you could search for this:

G02F AND flexible

This will also give you patents that are likely relevant but that don’t explicitly mention “flexible display”. For example…

Display module, display device and driving method of the display module

…or…

Novel flexible TFT array substrate structure and manufacturing method thereof

You will probably have to go back and forth a little bit. But in general, such a combination of patent classes and very general search terms works quite well.


Featured image of The National Inventors Hall of Fame in the Madison Building of the USPTO from Wikimedia.

One thought on “Sources for technology and business insights explained, part 3/7: Patents

Leave a Reply