Richardson Oliver Insights on the challenges of IP Teams

Erik Oliver, COO at Richardson Oliver Insights, talks about the challenges of IP teams tasked with managing high value portfolios and the increased demand for both better data and better ways to communicate the value of patents to executives. ROI supports the increased use of economic models and how Cipher provides the essential data to make this feasible. Erik is deeply connected with patent value and how you extract value from your portfolio. His work at Richardson Oliver Insights helps companies figure out how to use their patent portfolios strategically, advises them on various investment choices as well as works around patent buying and selling.

You have a separate business, Richardson Oliver Insights. Can you tell me more about that?

One of the things we started to realise about 2½ – 3 years ago was there were people who wanted to get data about the patent market independent from being working with a law firm.  We took the data portion of our business and moved that into a standalone entity so that any company in the world could come for data about the patent market and purchase that data and access that data and do that without having to have a law firm relationship.

We’ve expanded the products being offered. For example, we now offer a subscription service called Real Prices where we’ve have created a community of companies anonymously sharing data about their actual patent transaction closing prices. Anyone can buy that report now. So, we’re expanding the range of patent data related products.

What are the trends in quality and quantity of patent data?

In terms of the underlying patent data there is more high-quality data becoming available and some of that is free like Google Patents and some of that are paid services and analytic tools, such as Cipher. Also becoming more available is the second level of data, things about transactions, litigations, IPRs. Our clients are looking for more data and looking for ways to ground their decisions in the data. This seems to find favour over putting your finger in the air to make decisions.

What kind of problems or challenges that IP teams are trying to solve?

With transaction data, they’re looking to benchmark their patent acquisitions and sales, either for a specific transaction or for internal transfer pricing and accounting choices.

The consistent point is that they’re being asked by their CFOs for rigour and not just say, “Well, this was the cost of prosecution”.  But something relating to market transaction pricing: “What would this go for in the open market?” Or, “What if we decide to sell a part of our portfolio?  What could we expect to get if will sell?”  All of those type of questions.

We spend a good deal of time discussing portfolio optimisation. Do you think that’s a new conversation or has that been going on for a much longer time?

I remember in the early 2000s being in-house and being asked, “Why do we have so many patents? If we get rid of a third of our patents, then what would happen?” This is a very challenging conversation without some sort of model.

People have been asking those questions and wanting to have a framework for a long time.  What has changed in the last couple of years is newer tools with machine learning, for landscaping – classifiers – they make it possible in a way that was never possible before, to ask questions about the universe of all patents.

I remember at one time hand-coding patents owned by our company and everybody in our industry. That was a massive undertaking. That enabled us to answer some of the same types of questions we’re asking now, but we could only answer them about the three or four major companies in our industry.  We had no idea if there was somebody else with patents that might be relevant.

What’s changed is the time it takes to do it, the ability to do it repeatedly and the ability to do it in a way that is less influenced by human bias – in the sense of one person doing the categorisation and then getting different results compared to another team member doing the categorisation. That’s just the reality of the nature of humans. To be able to do it systematically, to be able to get some of those answers in a couple of hours instead of weeks, maybe even months, is super valuable.

How has Cipher helped take away some of that inefficiency?

We now have several clients where we have helped them build a classifier or worked with the Cipher team to build a classifier to try to understand the landscape. For example, we’ve just done a project in the last week or so where a client wanted to select a manufacturer to help with a specific technology. The candidates included 5 companies all owning tens of thousands of patents. There’s was no easy way using conventional search software to get a good read-out of what did they own in or adjacent to the focus technology.  With Cipher we built a classifier to understand what was in this space and not just for the 5 companies, but for everyone. We were able to get the whole universe of everybody out there that has these patents.

And did that reveal things that were unknown prior to the project?

It did.  As it turned out, none of companies had many patents in this focus technology. This was not what we expected to find. It suggested fresh avenues for the client. We were able to confirm that no company had all that many patents in this space and started to focus on non-patent reasons to evaluate the vendors. Could we have done pieces of this project before Cipher?  Sure.  But there’s no way we could have done it in a week. And reach conclusions about not just the 5 but to have some perspective about the world.

You referred to classifiers which is one of Cipher’s core capabilities.  How would you describe using supervised machine learning to someone who isn’t familiar with these new approaches?

I firmly believe in models, but recognise that despite all models being wrong, some are extremely useful. When we use classifiers, we know that there will be inaccuracies but it’s a reliable model that helps us understand what’s out there. We’re going to miss some patents that should be in and we’re going to have some patents in it that have no relevance to it.

What classifiers enable you to do is look at the universe and figure out things that are simply unknown. We focus on the fact that there was no way to systematically get to the same level of patent intelligence previously. It’s a big deal when you find a company with a significant patent holding that you have never even thought of.

We also find one surprising advantage of the supervised machine learning approach is that it also builds internal consensus within the client about what is “in” vs. “out” so there is more buy-in to the results.