Team

Jessica Hudak is a rising third year student at Stanford Law School and brings with her over eleven years of patent prosecution experience as a patent agent with law firms, start up companies, and most recently her own patent strategy consulting firm. Prior to becoming a patent agent, Jessica was a Biodesign Innovation Fellow, received her M.S. in Mechanical Engineering from Stanford University, and her B.S. in Mechanical Engineering magna cum laude from the University of Michigan.

April Yu is completing her M.S. in Computer Science at Stanford University, after having received her B.S from Stanford in the same field. With concentrations in artificial intelligence and human computer interaction, April contributes her experience in designing and implementing data-driven, but incredibly user friendly, applications. Prior to this, April interned at Palantir Technologies and conducted research under Professors Oussama Khatib and Steven Cooper.

Problem

Technology companies value having strong patent portfolio. Often, patents represent a company’s most important asset. Patents allow companies to create new industries and to be the market leaders within those industries. However, the patent management ecosystem is complex, inefficient, and fragmented.

Patent portfolio management is a complex undertaking, which requires—even for small portfolios—tracking hundreds of deadlines and documents at once. The vast majority of patent owners do not have in-house patent counsel and rely on their outside law firms to manage their portfolios at a great expense.

While most patent information is publicly available, it is not usable—it is distributed across many platforms and presented in many different formats. For most patent owners, the only tools at their disposal to manage their multimillion dollar portfolios are static excel documents and email inboxes which are regularly inundated by law firm reporting emails which are designed to protect the law firm rather than provide actionable information. This leads to important decisions being rushed and made without consulting what should be readily available information. Often portfolio managers are making multi-million dollar decisions with either partial or outdated information.

In an attempt to close this information gap, companies are often forced to “throw money at the problem” and are receiving an unclear return from the huge investments into managing their patent portfolios. Some companies with very large portfolios have been forced to seek out a solution and have either developed their own complex custom databases or use the currently available large enterprise solutions which cost hundreds of thousands of dollars, are overwhelmingly complex, and require tremendous amounts of manual data entry.

Patent portfolios are far too valuable and the stakes are far too high for the status quo to continue.

User Interviews

Before delving into solving this problem, we validated the need that Jessica had personally experienced in her eleven years of patent prosecution and working with dozens of clients and hundreds of patent applications. Over the course of the quarter the team interviewed more than fourteen patent professionals. Those professionals ranged from Engineers tasked with managing their company’s portfolios, to CEOs, to Patent Attorneys with decades of experience. The users that we spoke with had portfolios that were as small as a dozen patents all the way up to hundreds of thousands of patents, distributed across several subsidiaries. We also spoke with people who had built internal custom patent management systems for large companies that manage thousands of patents.

The users that we spoke with strongly validated the need that we identified. Some of the key insights we gained during these interviews included the following:

• IP Managers need an improved way to gather information and communicate the status of their portfolio with their investors and Boards of Directors.
• IP Managers need a way to identify and prioritize upcoming tasks.
• Because IP managers are often in charge of other programs within the organization, they need a quick way to get back up to speed on any given patent application that requires attention.
• Companies often do not possess information on the status and size of their portfolios and must request (and pay for this information) from their law firms.
• There are a lot of inefficiencies within the system, especially on the administrative side.
• IP Managers want to know what tasks are coming due in the near future.
• IP Managers want to know more about a patent applications history and other corresponding information while responding to office actions.
• IP Managers want to have a better sense of what their budget has gone towards and what the costs are coming up.
• IP Managers want a better sense of what their patents protect, specifically with respect to their technology and product lines.
• IP Managers need a better way to tag and sort their patent applications within their portfolios.
• IP Managers do not want to have to hire a person to manage software and want to minimize data entry.

Existing Solutions

In addition to understanding the user need, we also worked to gain a sense of the software solutions that are currently on the market. Within the patent management ecosystem, there are many companies that each focus on 1, maybe 2 aspects of the patent management process. Those aspects include prosecution and examiner statistics, patent searching, patent strategy, invention disclosures, workflow and file management, docketing, annuities, and others. This leads to a fragmented market place as shown below in Figure 1. The problems that a fragmented marketplace pose for the IP manager include spending a tremendous about of time and money researching and on-boarding a number of independent systems that any one company will need to manage the patent process. These systems require multiple log-ins and often the data cannot be shared between the disparate systems. Leading to an inefficient use of time and resources even after the systems have been installed.

Figure 1. Patent Management Technology Solutions

Over the course of the quarter we demoed six commercially available patent management systems. These products are primarily focused on managing workflows and patent personnel relationships. Importantly, at least currently, most, if not all, of these systems do not auto-populate with data provided by patent authorities (discussed in more detail below) and only use first party data that requires manual entry and verification. Therefore, for each of these systems, their set-up and use is very expensive and time-consuming. Additionally, all but one of the systems we demoed start at tens of thousands of dollars per year.

Due to these significant costs, for most patent owners, the only tools at their disposal to manage their multimillion dollar portfolios are static excel documents and email inboxes. There is room for improvement over and above the existing software products.

Patent Perfect

Patent Perfect is a patent portfolio management platform that will transform the portfolio management experience for large enterprises and, for the first time ever, will bring automated technology solutions to small enterprise and solo inventor patent portfolio managers. Patent Perfect is a delightfully intuitive portfolio management system that automatically presents accurate and prioritized information to patent portfolio managers.

Patent Perfect will allow patent owners to visualize and manage their portfolios more effectively and efficiently—replacing email, excel, or clunky software with a clean, clear way to organize and optimize portfolios. Furthermore, Patent Perfect future features will bring about significantly decreased costs by automating much of the patent prosecution process.

The current features of Patent Perfect, described in more detail below, include (1) a prioritized list of required patent prosecution actions–which is automatically updated and verified using our proprietary algorithms and custom rule sets–and (2) proprietary deadline time-lines and prosecution history time-lines that appear on each page and are updated to reflect the relevant information from that page. As described in more detail below, Patent Perfect is automatically driven by vast amounts of data.

Patent Data

By utilizing data from patent authorities, we will be able to automatically populate and update our system without requiring any manual data input. Additionally, we will be able to utilize patent and office action text to provide insights and patterns to improve strategic decision-making. One example of a patent authority data set is the United States Patent Office (USPTO) Patent Application Information Retrieval (PAIR) system, which displays information regarding patent application status. This system was developed by the USPTO to provide current patent application status electronically via the Internet. PAIR provides real-time status information for all action taken by the USPTO for a given application.

We have secured access to a full patent dataset for short-term use through a partnership with a local software company. This company’s database, through an API, gives us direct access to all of the data provided by patent authorities and in the same format. These data include the current patent data feed plus back files. The database provides, among others, US full text patents and applications, US patent and application images, PAIR US XML records, and PAIR US images (IFW).

Because our system is populated and runs automatically, driven by vast amounts of data, we have more freedom in our business model as we require almost no set up time or up front investment from the user. In the future, no data entry or sales support will be required to on-board a new customer, and therefore we have the flexibility to begin users with a free subscription, following a classic freemium model. Patent Perfect accurately and automatically populates the system with a user’s information, as described in more detail below. Because the patent portfolio information is auto-populated from the entry of basic information such as inventor name or company name, a user will see immediate value and proof that the system works before any sign up is required. This will allow us to establish trust in our product early on. The importance of establishing trust was a recurring theme during our user interviews.

Initial Prototype

By the middle of the quarter, we created a working prototype mock-up. The prototype consisted of detailed screen shots created with Adobe Photoshop. The screen shots were created using actual patent portfolio data for a local start up company. We automated the screen shots in a prototyping tool called Marvel (marvelapp.com), such that our users could click through the prototype and understand how the data was presented and how the different features worked. The initial prototype can be accessed here: http://marvl.in/4028fa.

As shown below in Figure 2, a user will first land on the dashboard page. This screen will present a visual, holistic view of the portfolio in the left panel. The visual may be downloaded as an image file and added into documents for easy presentation of information. The dashboard will also provide a list of the most immediate actions required in the right panel. Across the top of the page is a time-line. On the dashboard page, the time-line presents all of the upcoming deadlines that the user has over the next three months. We chose a three-month window based on USPTO deadline practice and the consistent user feedback we received with respect to deadline tracking windows.

Figure 2. Patent Perfect Dashboard

A user may access more details on the pending action by clicking through the list or through the icon on the timeline. Once a user clicks through on one of the pending actions, they will be directed to an action detail page, as shown below in Figure 3.

In the left hand panel, the user will be presented with key facts about this particular application. On the right hand panel, the user can click through to see the documents for this application. They can also click through to see more details and statistics on the patent examiner for this application, and other cases within the portfolio that the examiner may also be handling. In the center panel, Patent Perfect will summarize the key information from the pending office action for that case, such as which claims have been rejected, the reason for the rejection, and the prior art references cited by the Examiner. As with the dashboard, there is also a time-line across this page, however this time-line is specific to the prosecution history of this particular case.

Figure 3. Patent Perfect Office Action Summary

When we showed this mock-up to one of our users, we received positive feedback. His comment was “This is really cool, I wish I had this!” He also shared with us that, without a system like Patent Perfect, he was managing his patent portfolio reactively through his email inbox. When he saw the time-line and the number of times they had submitted responses to the patent office (as shown in the green and yellow time-line, above in Figure 3), he told us that he would have altered either his investment in this case or his response strategy. Patent Perfect presents patent information in a way that managers have never seen before.

User Feedback

As noted above, we shared our working mock-up prototype with our user group through a combination of live demos and virtual demos. For the virtual demos, we created a video of our working prototype, which can be seen here: https://vimeo.com/128526188. (The password is “patent.”) We also asked our users to fill in a survey after watching the video so that we could gather their detailed feedback. The survey can be seen here: goo.gl/forms/MBlH8k1rCG. In general, the feedback was very positive with four out of our five respondents indicating that they were very likely to use the software. We will utilize this feedback and future feedback to prioritize features in our future versions of Patent Perfect.

Final Prototype

The final prototype can be found at http://bit.ly/1NoEhIW. The current features of Patent Perfect include (1) a prioritized list of required patent prosecution actions–which is automatically updated and verified using our proprietary algorithms and custom rule sets–and (2) proprietary deadline time-lines and prosecution history time-lines that appear on each page and are updated to reflect the relevant information from that page. As described in more detail below, Patent Perfect is automatically driven by vast amounts of data.

The driving force behind the implementation was the need for the software to be incredibly user friendly and require little to no maintenance by the user himself. As such, Patent Perfect was built solely as a web application, which ensures compatibility across all platforms and allows for users to instantaneously receive software updates with no effort on their part. Patent Perfect is built with the Django web framework on Amazon AWS servers. The servers allowed for larger computing power when retrieving and analyzing the data, while Django provided a framework to build out the dynamic web pages needed to display the information.

The final prototype deviated from the initial design mockups largely due to the limitations in the data received. The prototype utilized a timeline library, which allowed for more time to customize the information displayed, rather than spending that time writing custom timeline code. The data did not include international patent information, which made displaying the entire scope of the patent portfolio impossible without additional subscriptions to patent authority data, which we will acquire for future versions. Instead, the space initially designated for portfolio statistics was utilized as a quick navigation bar through the entire collection of US patents in the portfolio. The office action summary page was also extremely difficult to implement, as the API only gave access to that information in PDF form. Because of this, we attempted to parse the PDF and extract the relevant information to display. However, because of the variations in how these documents are written, we could not achieve a sufficiently accurate representation of the relevant information to display to the user.

An additional feature that is different from the mock-ups is that we have not yet built the login feature.  Currently the final prototype requires no other user-specific information beyond the initial search for an inventor’s portfolio, which could be conducted by any user.  The missing credentials and thus workflow through a login process allows users to more quickly receive the information they request and begin their analyses immediately without having to pre-register.  The final prototype currently only allows for search by inventor name, rather than assignee.  The data, with respect to the assignee data, was not particularly clean.  Some companies included the address, while others did not.  Because of this irregularly formatted data, we were not able to achieve accurate enough search results and thus have not yet included that functionality.

However, many interesting algorithms embedded in our source code emerged because of the limitations of the data. Filtering the results from the search query is one example of such an algorithm. When searching for patents by inventor name, the dataset gives data for any patent that may contain a part of that provided inventor name. For example, if searching for the patents of “John Doe”, the dataset also provides the patents of “Joe Doe” and “John Smith”. However, we could not simply use filter out the patents of anyone other than “John Doe”. Because the various patent applications of a single portfolio can be written by different patent agents, our inventor “John Doe” (say his full name is “John William Doe”) could appear in the data as “John Doe”, “John William Doe”, “John W. Doe”, etc. Thus, we needed to implement an algorithm that spanned all the above versions of the name “John Doe”. The algorithm splits the inventor name into its individual parts (for this example, the individual strings would be “John” and “Doe”). Then, the algorithm searches for these individual strings in the inventor name provided by the data. If there is a sufficient enough match within a specified tolerance, the algorithm learns that the two names are linked and allows the patent to be associated with the searched inventor name.

Another interesting algorithm was formatting numbers associated with the patents to be displayed. The format of these numbers conveys a lot of information to the user. For example, a number in the format “US XXXX/XXXXXXX”, where “X” is a single digit, signifies a patent publication number, while a number in the format “XX/XXXXX” signifies a patent application number. Despite this wealth of information embedded in the format of the numbers, the dataset returns a string of numbers and letters with no delimiting characters to provide a context of the stage at which the patent currently is. Our algorithm then takes into account the length of the returned string as well as other statistics provided by the dataset, such as filing date, to determine the correct format in which to display the corresponding number.

Next Steps

Patent Perfect future features will include (1) a smart user interface that presents the most relevant information to the user based on their credentials and a clear navigation path through the system, and (2) summarized actions received from the patent office with plain language explanations created using OCR with machine learning. The future Patent Perfect will also automate routine tasks along the patent prosecution process, such as template and strategic office action response preparation, patent office form creation, and eventually automated filing—saving patent owners thousands of dollars per patent each year.

While we have visions of creating a platform on which all patent prosecution is conducted, our immediate next steps are accessing international patent data, parsing the documents provided by the dataset for the office action information, and implementing custom dynamic timelines and charts to be able to display this information with finer granularity.

A key future feature that we will also add is providing actionable insights and future projections through predictive analytics. One example of insights and projections we plan to build is around presenting cost information–both what has been spent and predicting future spending based on identifying patterns in past action and specific Examiner and art unit statistics. Closely tied with this feature is the ability for users to login to a customized dashboard, in which they will be able to conduct all of their analysis, aided by the breadth of information provided by our system.