Making the most of the latest data analytics tools
We are only scratching the surface with what can be achieved with an array of new offerings, leaving IP in-housers with plenty to consider with regard to finding the best platform
The data services sector has grown by leaps and bounds in recent years. The level of deal activity and the money being raised by some leading players reveal how the space is being targeted for investment as rights holders seek out the best platforms to drive more value from their intellectual property.
In the past year we have seen Clarivate Analytics snap up Darts-ip, IPlytics raise millions of euros in funding and UnitedLex take over iRunway. Top-level activity is focused on expansion, but there are also plenty of smaller, bespoke patent data services to choose from.
While tech-savvy lawyers are increasingly utilising these tools to inform strategic decision making, many others are yet to capitalise on them. Part of the reason for this is the general lack of awareness regarding patent analytics, how it can be used and to what benefit, as identified by two academics at the University of Cambridge in their paper, “Exploring the Future of Patent Analytics”.
As it stands, the industry is still in its early stages and – at a basic level – helps attorneys to streamline their strategies in order to boost efficiency and save costs. Nevertheless, patent analytics is expected to advance significantly and rights holders with experience using these tools will be well equipped to overcome the major challenges that they face when building and maintaining IP portfolios.
But with a dizzying array of options available, what should companies look for? Regardless of whether you are an expert in data analytics, there are certain things that everyone should know before they begin. We asked industry experts for their advice on getting the most out of patent data analysis, including how to optimise value from the various options on offer and what to expect from platforms in the future.
The middle ground
As a medium-sized public company with a few hundred granted patents and applications, how should you use data analytics to grow your portfolio? To get expert insight into the issue, we asked three leading thinkers how best to approach the challenge.
Jay Yonamine is the head of data science and operations, global patents and Michelle Park is senior patent counsel at Google. Over the past few years they have emerged as leading thinkers in the developing patent analytics space. Here is what they had to say:
- Collect key data:
- Analyse your existing portfolio and that of your competitors – The two most important attributes are general composition metrics (eg, the number of assets in which jurisdictions and trends in filings and expiration or maintenance fee payments) and how well the specific assets relate to specific business needs around key technologies. All of the aforementioned portfolio composition metrics and many estimates of technology relevance are made available by various IP-focused data and SaaS providers.
- Analyse the data to devise a strategy:
- Establish a general portfolio size model – While no magic, one-size-fits-all formula exists, it is still important to have a general framework in place to determine optimal portfolio composition. This helps to justify strategy to executives, explain rationales to other team members and provide year-over-year consistency.
- Understand business goals – The general model framework should map directly to business goals, which generally involve protecting core differentiating technology while continuing to invest in growing more speculative, long-term emerging technologies, as well as maintaining a sufficiently strong position relative to third parties.
- Refresh analysis and strategy on an annual basis – Portfolio management requires constant attention to keep pace with many impossible-to-predict changes, including case law, market dynamics and internal innovation rates. The data should be regularly refreshed and applied using the consistent model framework to account for these changes and update the strategy accordingly.
- Build technical resources as required:
- Adopt appropriate tools – As you move from a few hundred to a few thousand assets, an off-the-shelf modern IP management system (IPMS) will be necessary. This takes time and energy to initially set up and optimise, but it is critical to portfolio management at scale. Additionally, it is important to standardise analytics and reports to increase accuracy and reduce turnaround times. Most IPMS systems will have business intelligence capabilities that allow for analytical dashboard creation.
- Hire an expert – It is important to have at least one person – preferably in-house, although a full-time contractor will work if the person is highly experienced and empowered – who understands product management and analytics to manage the software, drive ongoing data quality initiatives and maintain continuous analyses.
- Establish consistent workflows – As a portfolio grows, attorneys will need to establish one or more taxonomies and enforce consistent annotation procedures, such as labelling assets within the taxonomies. This will allow for quick analytics on the number of patents with important specific business attributes. It will also dramatically reduce the time taken to identify key assets, thus enabling faster decision making and more informed portfolio management decisions.
Ahsan Shaikh is a partner at McDermott, Will & Emery and in recent years has become a leading voice on patent data issues, helping to organise a Silicon Valley group of in-house counsel that specifically addresses best practices in data analytics. Here is his advice on how a medium-sized business can use analytics to grow its portfolio:
- Generate a listing of the company’s registered patents that maps priority relationships between them and categorises these according to technology classifications. This gives a sense of the business products’ coverage.
- Apply the mapping technique against:
- a mapping of your top competitors’ patent assets;
- your company’s product roadmap; and
- a technology trends or market map. This will help to identify areas in which to grow out your patent portfolio.
- Generate a ranking of the company’s patents that are most frequently cited as blocking references against competitors. This will give a sense of which assets are valuable and against whom. Continue to pay maintenance fees for these patents. If they include family members, then keep filing continuation patent applications off them.
- Determine which references are most frequently cited as blocking references against your pending patent applications. If possible, acquire them anonymously through a middle-person to free up your path to allowance and your position against competitors. If you do not plan to acquire these assets, develop a response strategy to prosecute around them.
- Identify which inventors are most commonly cited against your high-value patents or patent applications and determine whether you can hire those inventors in order to boost your valuable technology patent pursuits.
Left to right: Ahsan Shaikh, partner at McDermott, Will & Emery, Bill Harmon, Director of intellectual property at Uber and David Andrews, Chief IP analytics officer at Aon.
Top tips to follow
Know what you want
Starting with first principles, users should have a clear idea of what they are trying to accomplish. This may appear obvious, but each data provider has its own strengths and weaknesses and it can be difficult to determine which is best suited to your needs. Having a clear strategy will help to narrow down the options.
Question the data
Patent analytics is not an exact science and there are multiple ways of modifying and cleaning the data. “Many people would think it is fairly straightforward to figure out which patents a particular company owns, but that is definitely not the case,” says BigPatentData founder Chad Gilles. The raw data can be noisy due to misspellings, differing entity names and mergers and acquisitions. Further, there are times when the data is missing (eg, when an assignment is not recorded) and the overall process requires a level of subjectivity. “The result is that different tools will often arrive at different conclusions about who owns what – and I have seen people surprised by this,” reveals Gilles.
Conduct a background check of the platform
Similar to any other purchase, data users must do their due diligence when seeking out analytics products. “The technical proficiency of providers varies considerably and many data and software as a service (SaaS)/analytics providers overstate their degree of automation or use of AI and machine learning,” points out Jay Yonamine, head of data science and operations, global patents at Google. “In a perfect world the patents department would have someone with sufficient technical and substantive expertise to vet data and SaaS providers.” For those without the resources to do so, it is worth seeking support from an external contractor or consultant, or from a member of an internal engineering team. What is more, attorneys should request references from the provider and – where possible – from companies that chose not to purchase the data or service.
Ensure that the visualisations deliver value
One question that Bill Harmon, director of intellectual property at Uber, asks is: “Can data analytics providers harness and visualise the information in a way that makes it usable to me and my team?” Aside from the analysis itself, it is crucial that the data be presented in a way that delivers insight and enables people to take action.
When in doubt, hire an expert
Data analytics is not everyone’s strong suit, but that should not prevent people from benefiting from the technology. “The average patent attorney will not really grasp the strengths and weaknesses of the bajillion pieces of software out there and, more importantly, the strengths and weaknesses of the underlying data (although to a certain extent they all work off the same data (eg, that from the EPO’s documentation database)),” says Gilles. “Further, a software provider that repackages the data is unlikely to be eager to explain its flaws or how hard certain things are (eg, deduping or normalising assignees).”
For those in-house attorneys who are not data experts, Gilles recommends hiring an independent consultant. After gaining an understanding of the business, these individuals can select the appropriate software for its goals. That said, it is important to ask any outside hires to explain each of their platform recommendations, as well as the so-called ‘gotchas’ in the data.
Figure 1. Network virtualisation example
The argument for an automated, data-first approach
Over the course of his career, David Andrews, chief IP analytics officer at Aon, has played a prominent role in shaping the patent analytics landscape. After spending 20 years at Microsoft, where he worked as a software developer and then as a patent attorney, Andrews eventually left to launch a data science and legal services company called Legal Analytics, which specialises in patent acquisition, valuation, landscape and portfolio management. Here he explains how to adopt a data-centric approach to portfolio development:
Step one: where are you going and how do you get there with data?
The first thing to do is to take stock of what you have and then create an idea of what is needed. Then, I would create metrics and tools to ensure that I build a portfolio that is sufficient to satisfy the business’s goals and at the best cost possible.
Each patent in the portfolio represents a variable value to the company, but that value can be difficult to measure through traditional one-at-a-time methods. There are a number of approaches to automated valuation, but one way to determine value is to estimate the chances of prevailing at trial on both infringement and validity, and then understand what sort of damages, as well as any ongoing royalty revenues, may be available.
It is possible to conduct this on a patent-by-patent basis for a few hundred assets, but when people focus on a single patent at a time, they often become susceptible to cognitive biases that make judgements unreliable. Data analytics allows us to see the forest and the trees in one glance, so the portfolio can be built with the combined view of the patents you own, as well as those held by competitors. Data analysts can build machine-learning models that identify patents that address specific technologies, and the collection of these represents a product or revenue stream which the patents in this space protect. These models are useful because they can be re-run and will produce the same results given the same set of data.
Figure 1 is an example that shows network virtualisation portfolios of the competitors in this space. The view is dynamic and in this instance has been filtered to show the priority dates and expiration dates of three companies in the overall field. The size of the box and the height of the line in the graph represent a patent count weighted by quality factors where patents that are predicted to be above average in district court litigation are weighted more heavily than average or below average patents. From this graph it is clear that Nicira is a relatively new filer whose portfolio will last for a long time. If Juniper is worried about a conflict with Nicira, it would make sense to file more patents.
Once the universe of patents has been identified, natural language processing and statistical techniques can be used to parse the claims of each patent and family to identify the superstars and the assets that may not be worth continued investment. It is also possible to use this tool to see how technologies, products and revenue streams are protected by the patents, and which are threatened by competitors through a quality and quantity viewpoint. Further, it shows what areas need more investment and where competitors are expanding. If an acquisition is necessary, then this technology is capable of finding suitable candidates that might be available without having to wait for patents to hit the market.
The ability to see the bigger picture allows attorneys to make decisions that are individually appropriate while supporting the larger mission of the company. Automated analytics makes this practical through real-time decision making. This is how to use data analytics to understand what the company should be working towards – the next step is to measure efficiency in getting there and commit to continuous improvement in both cost and quality.
Step two: keeping costs to a minimum
The only way to improve both cost and quality at the same time is to have continual measurements of each. Cost is relatively easy to track, using either financial data or prosecution information. Quality is more difficult and patent professionals often do not have a starting place to understand the quality of their portfolio in the aggregate. Machine-learning technologies can provide an immediate measurement of quality while humans review the portfolio and add to or replace computer-derived measures of quality. Every time a human adjustment is made the computer learns how to improve the estimate. While computers and algorithms can greatly reduce the burden on patent attorneys, ultimately, the decisions should always rest with humans. Informing decision making with data and analytics makes attorneys faster and better at their jobs.
Following this data-first approach, the new portfolio should be shaped to how you want it to support the goals of the business and you should be able to articulate why the money spent on each case is worth it given the overall portfolio and business aims.
The future of patent analytics
The key to utilising patent analytics effectively is to first understand what you want to achieve with your portfolio. For Uber’s Harmon and Aon IP Solutions’ chief IP analytics officer, David Andrews, the goal is to pursue strong patents that satisfy business needs – and at the best price possible. The pair are experts in the field, having worked together to develop a bespoke service that measures the strength of granted and pending patents. Their tool pinpoints the attributes that correlate with a powerful patent, thereby helping to highlight the patents that will have an impact at the negotiation table or in the courtroom.
Given that IP owners now typically place far more focus on the quality of their patent portfolios, it is worth asking data providers the question: “Can you tell me what attributes of the patent will help me to convince people that it is valuable and, if I present it to people, will they be dissuaded that they have a good likelihood of challenging its validity?” As it stands, many data platforms provide objective information about a patent, but say nothing about whether a patent should be pursued or abandoned – and this can be a major drawback. “They don’t necessarily tie the data that they have to giving you decisions that are proven to yield results,” argues Harmon. “The data platforms do not try to pick the winners or losers, so they do not fundamentally nudge me in the direction of what I should do. It is still left up to the attorney to wade through the data, and that’s not quite as helpful.”
Patent analytics is yet to experience the breakthrough that occurred in the sports analytics industry years ago, when several Major League Baseball teams began to adopt a far more sophisticated approach to number crunching in order to gain an edge. This may be due to the general mindset of data providers and attorneys. “People are not trying to optimise value, they are trying to minimise risk,” Harmon explains. “I think the phenomenon that occurs when people are trying to get patents is that they are working just to get that allowance for the invention and they do not want to part with it.” Ideally though, attorneys should take a moment to consider whether the invention is giving more marginal value or whether they should be investing resources elsewhere.
It is worth reiterating that patent data analytics tools are still in their early stages and these technologies are likely to advance significantly in the coming years. “Exploring the Future of Patent Analytics” used a survey to delve into some of the problems in the field and how these could be solved. Respondents argued that a more harmonised approach to open source patent data would increase quality, while many felt that a lack of understanding of existing tools and techniques – and how to use them effectively – calls for a general increase in awareness. Finally, there was strong interest in improving patent data visualisations. In the long term, this would mean that graphics become more interactive, intelligent and personalised.
Data is not an end to a means, but a means to an end. While analytics are designed to simplify the decision-making process, having a confirmed strategy and a clear idea of what you want to accomplish is essential before you start to play around with figures.
To get the most out of patent analytics platforms, attorneys should:
- ask questions to determine exactly what the data shows and what the potential limitations are;
- check that the visualisations deliver value and help to inform decision making; and
- hire an expert to point them in the right direction.