AmberBlog

Here at Ambercite we believe in the value of patent citation data to predict patent quality, but would not claim that this ability stretches to predicting Olympic medal counts.

In any case, there are a number of other analysts that are attempting to predict Olympic medal counts, see the excellent graphics found here. In this graphic, the predicted medal count for the US ranged from 82 (USA Today) to 113 (Price Waterhouse Coopers). These different groups in use a range of formulas are used, as discussed by The Economist. 

But who will be the best analyst?

While avoiding the risk of trying to predict medal counts ourselves, Ambercite might go out on a limb and make one prediction - that the average of the predicted values may be better, or similar to, the best of the individual predictions. This is based on the principle of collective wisdom, where the predictive ability of experts and non-experts alike can be improved by collecting a range of independent opinions and then averaging these opinions. While each opinion may be flawed in itself, the errors in each opinion is likely to be randomly distributed and an average of these opinions may be close to the truth.

So what has this got to do with patent analysis? Network Patent Analysis (NPA) is a method of ranking patent quality based on the citations connections of the patent to other patents around it, being forward and backward citaitons (and the forward and backward citations of the connected patents, and so on and so forth). Patent citations are not randomly or computer assigned, and instead each citation is an opinion that two patents are similar in some way. While individual citation opinions can be imperfect like any other individual opinion, if we collect enough of them the result should give us a very meaningful result.

NPA can be contrasted with other approaches. These typically tend to be:

• either a count of the forward citations from individual patents, which is only accessing part of the available citation data in an area of technology and so missing a lot o potentially valuable information.  
• combining these forward citation counts with what I have called 'prosecution measures' such as the number of family members for the patent, and whether the these patents have been renewed or not. This thinking behind this is is that prosecution measures reflect the value that the patent owner places on their patents.

The limitation of using prosecution measures is that they are based on the opinion of one person, namely the patent owner (or the IP manager within the patent owner). These patent owners are likely to be very careful, objective and relatively well informed in the decisions they make. On the other hand, they are just one opinion, and an opinion that may be influenced by other factors besides the quality of the patent, such as budget presssures, and imperfect information about the likely success of the protected invention. For example, consider the management of the same possibly valuable patent by a) a manager in a well-resourced and very profitable company and b) a smaller university. These two different types of owners are likely to use different management strategies, and for good reasons. These different approaches in turn would mean that a analyst using a prosecution type measure might draw very different conclusions about the same patent.

This is why we prefer network based measures such as NPA. The management of the patent will have an impact on the network, in particular the decision to filing similar family members. On other hand, NPA will look at a range of other opinions besides those of the patent owner, leading to a more rounded final assessment of the quality and influence of the patent. 

Interested in a robust and independent opinion of the quality of a group of patents of commercial interest to you? Please contact us for further information. 

Ambercite was honoured to be invited to present on NPA at the Patent Statistics, Innovation management and IPR Conference in Paris, held on the 5th of June 2012. I presented Network Patent Analysis and foreshadowed our upcoming new products for the first time to a European audience.

The conference also included a number of presentations on the role of patents in the ICT space. Jim Bessen from the Boston University School of Law in his presentation Patents and Possible failures in the Digital World argued the net effects of patents in the ICT space was negative, as the value destroyed due to the effects of patents was greater than the value created through higher profits. Simon Forge from SCF Associates in his presentation "The Tragedy of the Patent" argued that Europe had an opportunity to steal a march on US companies due to the exclusion of business method and software patents in Europe.

The counterview was provided by Claudia Tapia Garcia from RIM, Monica Magnusson from Ericsson, and George Whitten from Qualcomm, including in a panel discussion.  Perhaps the most thought provoking of these presentations was a presentation from Monica Magnusson and about how the 3G Partnership Project (3G for the rest of us) mobile broadband standard works. A series of working committees decide areas of improvement for 3G, and the various companies working in this area are invited to develop competing solutions. The engineers on these working committees then evaluate these solutions , and pick a winner based on technical merit, and make this solution part of the now revised standard. Successful vendors are required to make their solution available to all comers on the now almost infamous Friendly, Reasonable And Non Discriminatory (FRAND) conditions, and one suspects that peer pressure helps to ensure that licenses are indeed FRAND.

The role of patents in this process is provide participants with a currency for exchange during these negotiations, and means of ensuring that the successful developers are rewarded for their effort, and not surprisingly Qualcomm, Ericsson and RIM all defended that patent system for this reason. In their view, the current increase in patent litigation in the ICT space, and complaints about 'patent trolling', are all a healthy and to-be-expected outcome from the patent system.

Regardless, one of the representatives of the above companies then admitted in discussions afterwards that their company does sued on a regular basis by other patent holders, but prefers to deal with such assertions by defending itself in courts where it has been quite successful, and was starting to build up a reputation for being so. This representative's key observation was that there are many patents of dubious quality out there, and this may be part of the cause of the recent trend in litigation.

And there may be a lot of truth in this. Allison, Lemley and Walker have shown that only 9.2% of patents asserted by Non Practicing Entities (NPE) eventually win their cases in US courts. This is lower than the 46% figure commonly quoted for all US patent litigation (and dating from 1998, and so which may be out of date). However the recent litigation between the Practicing Entities Apple and Google has seen just 5 out of the 21 patents (a little under 25%) originally asserted survive review by the Judge Posner.

The reason for some of these high patent invalidity rates may due to the difficulty of searching for prior art patents during their initial examination. Current patent searching practices do have significant limitations (including having to read through long lists of patents to find the patents you might be looking for), and this has encouraged Ambercite to apply its Network Patent Analysis principles to develop a new method to improve patent searching. The results of this development are being finalized at the moment, and we look forward to demonstrating this in due course.

I have previously written a blog article on patent quality, and covered various aspects of patent quality in other blogs. I have also been lucky enough to have had many discussions with other people about what patent quality is, both online and in person. This includes recent discussions at the "Patent statistics, innovation management and IPR" conference in Paris last week, for which I will separately report on. And over these discussions, a greater understanding of patent quality is starting to emerge, namely that:

When we think about it, this should not be that surprising. The view of the quality of a particular patent may be very different if you are patent owner, licensee, licensor, potential infringer, follow on inventor or ligitator. Even a litigator may have a different view of patent quality depending on the alleged infringing product. 

Another definition of quality is given by ISO 8402-1986, which defines quality as "the totality of features and characteristics of a product or service that bears its ability to satisfy stated or implied needs", which is a long winded way of saying "fitness for purpose". The same could easily apply to patent quality.

Here at Ambercite, we have defined a measure of patent quality, based on where a patent sits within a network of its peers. However given that patent quality is dependant on its purpose, and different people have different purposes, it may have been a bit ambitious to claim that our ranking system was a measure of patent quality per se - because how can we claim to measure something that means different things to different people?

For this reason, we will from now on refer to the results of our patent ranking system as the relative patent 'influence'. We have chosen the word 'infuence' because the highest ranked patents do have the most influence in the patent network. And regardless of the name change, we remain proud of the ability of patent influence to be a more than useful predictor of patent quality (however quality is defined) , as evidenced by how Network Patent Analysis (NPA) was able to identify the:

• key patent for the promising phase III bapineuzumab as the highest ranked (in terms of influence) patent in our study of 48,000 patents for Alzheimer's treatments, despite its low forward citation count
• relative importance of the highest ranked patent US7,663,607 in the smartphone patent wars white paper, as confirmed by leading smartphone patent wars commentator Florian Mueller.
• the importance of the successfully litigated Paice patents in our study of 72,000 hybrid car patents.

So what does this means for patent analysts and our clients? We still stand by everything we have done and said to date, apart from this change in terminology.

However this does help clarify our value proposition for patent analyst and potential clients who may be using other metrics to define patent quality. Just like a doctor might use multiple tests to determine the health of a patient, we would recommend that our 'patent influence' test may provide a unique and very valuable perspective on the 'health' of the patents they are reviewing that may be missed by other techniques.

If you were relying on your doctor's advice for your health, would you want them to base this advice on incomplete data? No? In that case, why would you accept incomplete metrics for what may be a financially significant decision in the patent space?

Please contact us to find out how we can help you understand the full picture (literally in our case, unlike some of our competitors) on patent quality and value.

Ambercite's Ben Palmer, our Director of Software Development, presented on 'Big Data" to the Churchill Club this week (a group focused on IT and startups). Ben was standing in for our Managing Director Doris Spielthenner, who was unavoidably detained elsewhere working on another big data project for the Australian government.

Ben was one of three excellent speakers. Ian Holsman, consultant & Former Chair of the Apache Hadoop project, spoke about the work he had done on data mining for a large consumer news website. This at one stage included real time feedback from the newspieces they were posting on the website, allowing the website to fine tune articles of interest for their audience.

Peter Buckingham from Spectrum Analysis Australia spoke the work he did in predicitng the best location for new franchise outlets, based on sophisticated data analysis. He also discussed a recent court case in Australia, where a franchisor was successfully sued by a franchisee over misleading representation made by the former over likely sales for a new franchise. When asked on what basis these predictions were made, the franchisor admitted to 'scientific guessing'. The judge was not impressed.

Ben spoke about some earlier work he and his team had done for the Australian government on predicting air traffic movement all over Australia, and in particularly deviations from expected movements, in real time, i.e. while these planes were in the air. Ben also previewed recent developments in our patent searching software. 

Despite the variation in areas of interest, some common themes emerged from all three speakers and the followup discussion:

• The real value in big data is not the data itself, but what we do with it. Visual presentation of the resulting analysis can help end users make the most of this value.
• Data quality, even from reported reliable sources, can be very patchy. There will be lots of hole. This in turn creates opportunities to identify this missing data, which can be very important.
• Timely analysis can be very important

All of these concepts have been taken on board in our development program:

• We have been told that our visualisations are the best in the business, but we aim to go beyond anything we have done before. 
• Like others, we also need to clean up the data we are using. We are tackling the problem of missing data in ways that we think will surprise and delight many people. Certainly our internal panel of patent attorney testers are more than excited by what we are showing them, and the additional information being made available when compared to other patent data sources.
• And while we don't necessarily need to process patent data faster than planes can fly, we understand that clients want results in real time when use patent analysis software, and that new patent data is published every day that our clients will want to know about.

Interested in knowing more? Come back to us, and we will keep you fully informed about future product releases.

A study of 14 of the largest subway networks around the world has discovered that they all tend to end up in common network structure, once you consider the distribution of lines, stations and total distances. For example, the number of stations was proportional to the square of the number of lines. The core of the networks had a similar number of neighbors in the network, with about half of the stations found outside the core. Most interesting, the different networks all appear to converge to this optimal structure over time, regardless of how the network began.

So what has this got to do with patents? Here at Ambercite, we believes that patents are best asssessed in terms of their role in the network. We have only recently republished the image below from our smartphone report, but it remains an excellent example of how individual patents can be assessed in terms of the patents around it.

In this particular case, we can see how this particular Motorola patent is connected to a broad range of backward citattion patents, but also there is a broad range of foward citations links to a strong clusters of key Apple patents.

However not all patents have this same degree of valuable connections. In fact, virtually all of the networks we have studied have converged to  a similar distribution of patent value, (again previously presented) in which only a small number of patents have the majority of the value. 

So when the academic study found that subway networks too converged to a common network structure, this did not surprise us, and in fact we would expect to see convergence in a variety of network structures, such as for example the Delta  airline route map, which is a great example of networks in daily life . We await with interest to see the details of these other types of networks being published.