Patent Quality Metrics: Finding Reliable Metrics Linked to Patent Value

Patent valuation is a topic that features often on this blog. After all, IP represents an important part of a company’s assets. Unfortunately, intangible assets in general, and patents in particular, are simply not easy to value. In a recent series, IPWatchDog explored questions like: What is a good price for a patent? Are you getting a great bargain or are you being overcharged? Are you overcharging or selling yourself short? Whether as a buyer or a seller, those are hard questions to answer without comprehensive pricing analysis. Various strategies have been proposed for pricing patent portfolios. Known pricing strategies are either free but unreliable, or more reliable but quite expensive.

Pricing strategies within the free-but-unreliable category include:

• Pricing based exclusively on size of a portfolio (number of patents), times a fixed multiple
• Pricing based on number of applications filed per period, times a fixed multiple
• Cost-based valuation, which is based on the premise that each patent is worth what you paid for it (in R&D, patent office fees, and legal fees)

These pricing strategies are within reach of people outside the patent field, whereas the more detailed analyses described next require knowledge of the patent system.

The other kind of pricing is expensive to obtain. You basically have to hire an expert. But the resulting valuation is much more accurate. These pricing strategies include:

• Market-based valuation, which attempts to use a like-for-like analysis of patent prices. For single patents or small portfolios, it can be difficult to obtain a “benchmark” based on an earlier sale of a comparable product. But market-based valuation is effective for larger portfolios.
• Income-based valuation, which is the most reliable and most used valuation strategy. It works by first determining a royalty rate by mapping the patent onto a product and deciding what fraction of the product is covered by the patent (typically a fraction of a percent). The royalty rate is then multiplied by unit price and units sold to arrive at a valuation.

Without hiring an expert from the patent field, it is difficult to get meaningful information about the value of a particular patent at a glance. As a result, investors often merely look at the size of a company’s portfolio or its rate of filing new patent applications, at least as a preliminary indicator. This provides an inexact but fast way of estimating the value of the company’s patent portfolio. However, the use of size as a quality metric among investors has some unfortunate side effects.

Particularly for mid-to-large sized companies, the resulting competition to have the biggest patent portfolio can lead to “portfolio bloat”. In other words, “everything and the kitchen sink” that is happening at the company’s R&D is packaged into patent applications — not just real, groundbreaking inventions. Some companies even have yearly quotas for number of filings or number of patents granted that they aim to achieve. This results in “filler applications”, filed primarily to achieve a quota. Often, these “filler applications” do grant as patents, but with a scope of protection that is factually nil. In other words, the scope of protection is so small that the patent has no deterrent effect on competitors. Some patents are granted with claims that are so detailed and specific that you couldn’t infringe them without willfully aiming to. However, to unsavvy investors, “a patent is a patent”. And so companies that follow this kind of strategy do sometimes get rewarded with a higher valuation than they perhaps deserve.

Although prosecuting the applications that issue as these “filler patents” provides an easy source of income for the USPTO and law firms, it is unsatisfying work. Furthermore, “portfolio bloat” represents an inefficiency in the economy, as thousands of meaningless “filler patents” grant each year. This contributes to information overload and makes the patent world even more opaque to outsiders than it already is. Also, the practice of “portfolio bloat” ultimately undermines trust in the patent system, as “filler patents” dilute the average value of patents overall.

These are known problems. Over the years, researchers and investors have continually attempted to streamline patent valuation, to offer more transparent yet reliable metrics to investors. For example, it has been proposed that patent value could be tied to back-references — meaning citations as prior art by another patent — similar to scientific papers. This turned out not to be true. Without adjusting for various factors, back-references are at best weakly correlated to patent value. References mentioned in the background of a patent application are given much less ‘weight’ than, for example, in the world of academic publishing, as there is no requirement to explicitly reference earlier patents in patent applications. Furthermore, it is easy to “game the system”. Since patent applications are not subject to independent quality control in the form of peer review before filing, it is easy to “back-reference” unrelated patents within the prior art section of a later patent application by the same company, in order to boost the earlier patent’s ranking.

In a similar vein, it has been proposed to investigate the types of words used in a patent claim, following the premise that rarely-used words translate to more “original” (and therefore, more valuable) patent claims. This approach has also failed to gain traction. Companies can simply instruct their patent drafters to use unconventional words. This is easily possible, as patent drafters are not bound to any particular lexicography.

Lacking a better alternative, the status quo remains: in patent valuation, you get what you pay for. However, if metrics could be found that were inexpensive or even free to obtain, and that provided highly predictive information about the quality — or even the specific valuation — of a portfolio or patent, that would shake up the entire patent valuation industry. Investors would rejoice. And without changing the way that patent prosecution works, the number of “filler patents” and the resulting “patent bloat” could be significantly reduced.

In order to come closer to this goal, we need to ask ourselves: what are the defining features of a “filler patent”? At least two things stand out. First, “filler patents” go through more rounds of prosecution than other patents. Secondly, the independent claims of “filler patents” are longer (have higher word counts) than other patents.

Let’s look at multiple rounds of prosecution first. A “round of prosecution” means an Office action from the USPTO and the applicant’s response. It is typical for “filler patents” to go through multiple rounds of prosecution, such as six or more rounds. At each round of prosecution, the claims are tailored, so that the scope of protection of the resulting patent is whittled down until essentially nothing is left. Then the application is allowed to issue.

The other defining feature of “filler patents” is that their independent claims have very high word counts, as compared to other patents. Overall, the average word count of independent claims of granted patents has been found to be very constant over time. Average word count is relatively constant for all technology areas, as well (with IPC class C being the only major exception, having an average word count of about 30 words less than the other classes). The distribution of word count of the independent claims is presumed to be similarly unchanging.

With the help of Paul Dougherty from The Patent Box, a group of 3,131 patents issued on December 27, 2016 was examined for independent claims’ word count, with the resulting distribution being shown in Figure 1. As expected, the word count of independent claims of these granted patents creates a bell-shaped curve with a right-shifted distribution. The median claim length is 173 words. Longer independent claims reflect narrower scope of patent protection. For “filler patents”, it is not unusual for the main independent claim to be 1.5 pages (about 350 words) or longer, by the time the patent is allowed to issue.

In fact, it is impossible to define the exact boundary, in terms of wordcount and rounds of prosecution, at which a patent’s scope is “guaranteed” to be factually nil. However, it is true that the higher these numbers, the more likely a patent is to be just “filler”. For the purposes of defining a cutoff, therefore, it is believed that patents with independent claims averaging over 350 words, particularly ones granted after 6 or more rounds of prosecution, are extremely likely to be “filler patents”. However, lower cutoffs (such as 250 words and 5 rounds of prosecution) could also be reflective of a high probability of a particular patent being “filler”.

“Filler patents” aside, it is also generally the case that rounds of prosecution and word count of the independent claim reflect patent quality, making it hard to “game the system” and to force an earlier allowance with a shorter independent claim, simply by prioritizing improvement of these metrics themselves. Patent practitioners are already trained to try to get an allowance with as few rounds of prosecution as possible. At the same time, patent practitioners constantly hone their skills at drafting the shortest possible independent claims that still meaningfully reflect the invention. Therefore, it is basically impossible to “game the system” and to magically bring prosecution to a swift close with a very short independent claim, unless the underlying invention really is foundational.

Here, it is important to note that prosecution rounds and claim length are closely interrelated. In order to swiftly bring prosecution to a close, an option is often to lengthen the independent claim by adding a feature. This is true especially if it is debatable as to whether the claim without that feature would be patentable. On the flip side, applicants can choose to risk extending prosecution by another round, but fight for the existing scope of protection (that is, by declining to add features to the independent claims). Patent prosecutors are already weighing these two competing goals at every step and trying to satisfy them both, in what turns out to be a tricky balancing act with no easy shortcuts to success.

The rare patent application that issues with really short independent claims after just one or two rounds of prosecution is rather likely to be a ground-breaking invention that is the “first of its kind” in a particular field or area. This is because broad scope of protection is only available for really revolutionary inventions that open up a new “space” of technology.

As an aside, there are some patents that issue right away with short claims because the Examiner at the USPTO failed to do a good search or examination (i.e., the patent is invalid). However, that is believed to be the exception, as Examiners are also trained to be very wary of short claims. Unofficially, some patent practitioners joke that USPTO Examiners apply the “three-finger rule”: if you can cover the independent claim with three fingers, it’s still too short to be allowed to issue.

Taken together, the two interrelated factors of: (1) Rounds of prosecution and (2) Word count of independent claims, could serve as a much more reliable immediate indicator of quality. The information about rounds of prosecution and claim word count of granted patents is already “out there”, as the prosecution history of published applications (indicating the rounds of prosecution) and the contents of published patents (including their independent claims) are available online from the USPTO. Unfortunately, it currently requires data mining the USPTO’s public files to produce these metrics for any particular patent.

It might change the world of due diligence if these pieces of information were more accessible, for example if the USPTO provided them in the electronic file or on the cover of the granted patent itself. This would give investors outside the patent world much more reliable immediate indicators that are closely correlated to the scope — and consequently, the value — of a patent. Over time, these indicators might replace the more simplistic metric of merely evaluating patent portfolios based on size (meaning number of patents). As a consequence, companies would be incentivized to abandon patents with very high word counts and long prosecution histories, in order to keep their portfolio averages (meaning word count and rounds of prosecution) in check.

If this standard were adopted, and companies accordingly reduced the practice of “portfolio bloat”, IP practitioners would still have plenty of work to go around. In the exemplary dataset shown above, only 7.3% of patents have independent claims with 350 words or more, indicating that they are likely to  be “filler patents”. Disincentivizing such “filler patents” might lead to a slight loss of business for patent law firms and the USPTO. However, this loss would be marginal, and in any case could likely be offset by the continuing positive trend in patent filings which has been seen over the past few years. Furthermore, surplus patent budget that is freed up by abandoning “filler patents” can be redistributed back into a company’s R&D. This could result in an increase in real innovation, thereby leading to more meaningful patent filings.

Also, adopting this standard would certainly not replace the work performed by firms specialized in patent valuation, as the average word count and number of rounds of prosecution of a patent portfolio only serve as a preliminary measure of the portfolio’s overall value, and cannot be translated directly into a dollar-estimate of patent value. Therefore, the demand for detailed market-based and income-based valuations of portfolios would be largely unaffected.