Portfolio Value Index (PVIX) Methodology

Summary

The Portfolio Value IndeX (PVIX) score is an internally developed tool designed to measure patent portfolio value based on published academic studies. The score is based on three components: (1) Family, (2) Market, and (3) Reputation. Users can use the patent search tool or input custom lists to see a selected PVIX score.

See PVIX in Action!

Methodology

Family: PVIX measures size based on a per-family basis rather than a per-patent or application basis, using the European Patent Office’s (EPO) Simple Patent Family ID, a unique number assigned to a group of patents or applications covering the same technical invention.

Every currently-active patent family is assigned a PVIX score, and the overall PVIX value for a given portfolio is simply the sum of the individual PVIX scores for each family within the portfolio. Thus, for a portfolio containing N unique families, the overall PVIX value is:

Evaluating on per-family basis permits PVIX to assess the geographic market and technological impact for each fundamental invention disclosed within a portfolio without inflating due to multiple patents or applications covering the same invention. As explained below, the geographic market of a patent family is captured by its Market score while its Reputation score captures the technological impact of its disclosed invention.  For a given patent family (i), its PVIX score is essentially the product of its Market and Reputation scores.

Once each patent family has been assigned a PVIX score, the raw scores are renormalized so as to fall within 0 and 100.  The overall distribution of PVIX scores for patents families world-wide is depicted below:

PVIX scores are updated every week as new citations are made and new documents are published.

Market: Academic studies have shown a significant correlation in patent value to national GDP.[1] Therefore, to assess the economic dimension for a given patent family, the Market score uses the most recent national GDP figures for each country in which that family has one or more currently-active issued patents.

For example, take US-8697359-B1 entitled "CRISPR-Cas systems and methods for altering expression of gene products" and assigned to the Massachusetts Institute of Technology.  This patent belongs to simple family ID 50441380, which in turn has 35 family members (including 10 US publications). The family's active issued patents range across 26 different countries, with its EP patents contributing 20 of those nations.  The corresponding GDP figures, normalized with respect to the GDP of the United States, are listed below.

The Market score for simple family ID 50441380 is simply the sum of these numbers, which in this case is 2.450.

Reputation:  Studies have also shown that forward citations are a reliable indicator of patent value. [2]   Hence, using forward citations, the Reputation score captures the amount of recognition a family’s disclosed invention has received from other patent publications over time.

First, for each publication in our global patent database, the number of citations received is normalized according to a hierarchy of tiers: (1) time of publication, (2) country, and (3) technology field (represented by the IPC classification codes).  Continuing with the example above, the ‘359 patent was published in (1) 2014 by the (2) US patent office, and its main IPC code is (3) C12Q 1/68. The corresponding hierarchy of tiers for this patent is highlighted in red in the figure below.

At each tier, the number of citations received by each category is discounted by the overall geometric mean (GM) of that tier.  As an example, for the tier highlighted in blue below, the total number of citations received is 146,864 and the overall GM is 2.14.  The total number of citations received by the C12 (“Biochemistry”) IPC class is 37,147 and its GM is 2.10. The two means are then divided to arrive at an adjustment factor for C12 (2.10 / 2.14 = 0.98).  Finally, these horizontal adjustment factors are aggregated together vertically (see the portion highlighted in red below) by taking a weighted GM to arrive at an overall adjustment factor for the number of citations received.

In this example, for US patents that were published in 2014 and that have a main IPC classification of C12Q 1/68, the adjustment factor for the number of citations received is 1.06. In other words, citations received by this group of patents (again, in red below) are given slightly more credit by PVIX.

The exact same treatment is applied to the number of citations made. Take, for consideration, CN-106637421-A entitled "Method for constructing double-sg RNA library and method for applying double-sg RNA library to high-flux functionality screening research" and assigned to Peking University.  This application was published in 2017 by the Chinese patent office, and its main IPC classification code is C40B 50/06. Applying the same approach as above, for Chinese patents that were published in 2017 and that have a main IPC classification code of C40B 50/06, the overall adjustment factor for the total number of citations made is 0.98.

The Chinese application also happens to cite to the US patent from this example. However, instead of tabulating this citation as 1.0, the Reputation score multiplies the received and made adjustment factors together to arrive at a count of 1.06 x 0.98 = 1.03 for this individual forward citation.

This is similar to the CITX rating developed by Unified to predict patent value. However, in contrast to the CITX rating for individual patents, Reputation determines these citation ratings on a per family basis.  For example, continuing with simple family ID 50441380, this family has received a total of 461 citations from other publications. These 461 citations were made by 326 different simple families. For example, simple family ID 49624232 (University of California) made 71 citations to simple family ID 50441380 (MIT).  Pioneer Hi-Bred International Inc (simple family ID 55361965) made 25 citations, Sangamo Therapeutics Inc (simple family ID 51899008) made 7 citations), etc.

Consider the 7 citations made by Sangamo Therapeutics (51899008) family to the MIT family in this example (50441380):

Instead of adding the 7 adjusted citations (above) together, the Reputation score looks at the maximum adjusted number of citations from one family to another.  In the table above, the maximum is highlighted in yellow and is 1.513.  Finally, for the MIT family (50441380), its total Reputation score is the sum of all 326 family-to-family maximum adjusted citations.  In this case, that final score is 480.

Examples

The following case studies serve as examples of PVIX’s potential applications.

# 1: Litigated vs. Non-litigated patents

Academic literature has shown that litigation or opposition proceedings indicate higher patent value.[3] Figure 1 shows a histogram of PVIX scores from all US litigated patent families in the since 2000 (blue) overlaid against all world-wide active patent families (red).[4]  Litigated US families demonstrate a PVIX score that is 20 points higher (66) than the average.

# 2: Orange Book (OB) patents and Tech Center 1600

In the US, any Food and Drug Administration (FDA) patents associated with an approved drug must be included in the Orange Book. These listings include many of the world’s most valuable patents.[5]  We used PVIX to compare the Orange Book patents relative to non-Orange Book patents in Tech Center 1600 (biotechnology and organic chemistry).[6]

PVIX scores reveal that Orange Book Patent families have a higher median score (67) than counterparts (58) in the same Tech Center (1600).[7] PVIX scores align with the view that Orange Book listed drug patents are more valuable than non-listed patents in the same technology sector.

#3:  Standard Essential Patents and Non-Essential Patents

While what qualifies a patent as essential to a particular standard is a hotly contested issue in most FRAND litigation, very few would deny the inherent value of such standard essential patents (SEPs).[8] This example compares Samsung’s HEVC patent portfolio against Samsung’s broader set of video-compression patents that are presumably not essential to the HEVC standard. Unified did not employ any subject-matter experts to independently-verify the essentiality of these two sets but instead relied upon its internally-developed Objective Patent Landscape (OPAL). OPAL uses publicly-available essentiality lists provided by patent pools such as MPEG-LA and HEVC Advance to train an algorithm to predict essentiality to a given standard.

As seen in Figure 3, Samsung’s standard essential HEVC patents received a higher median PVIX score of 66 compared to Samsung’s other video compression patents (PVIX score of 58). Additionally, the narrower base of the HEVC patents indicate less variability in value; that is, the HEVC patent families are more aggregated around their higher PVIX value, while the non-HEVC patent families greatly vary in value.

Search Unified's Portal to see PVIX ratings for specific patents

For more information about PVIX or the analytics tools available on Unified’s Portal, view our FAQ. For questions regarding the trial sets referenced in the case studies above, please contact info@unifiedpatents.com.

See PVIX in Action!

References

[1] Ernst & Omland, The Patent Asset Index – A new approach to benchmark patent portfolios. 
[2] Falk & Train, Patent Valuation with Forecasts of Forward Citations
[3] Harhoff,  & Reitzig, Determinants of Opposition Against EPO Patent Grants
[4] Figure 1 Horizontal values show normalized PVIX values with 100 being the most valuable.  Vertical values shows the count ratio of patent families normalized to account for their respective sizes.  The area under each curve totals 1.
[5] Cooper, Higbee & Byrd, INSIGHT: Drug Delivery Devices in the FDA Orange Book After In re Lantus.  see also Winkler, Weingarten; Shana Cyr, Requirements, Benefits, and Possible Consequences of Listing Patents in FDA’s Orange Book
[6] 87% of Orange Book patents are from Tech Center 1600.  A variety of diagnostics and drug-related patents pending FDA approval are also in 1600.
[7] As in Figure 1, the distribution in Figure 2 is presented using percentages in order to account for the different sizes of each sample set and allow for an easier comparison of the Orange Book and non-Orange Book patent lists.
[8] Alexander Galetovic, Stephen Haber & Ross Levine, An Empirical Examination of Patent Holdup, 11 J. Competition L. & Econ. 549-578 (2015), (available at https://doi.org/10.1093/joclec/nhv024); see also Corporate Counsel Business Journal, Standard Essential Patents (SEPs): Costs And Benefits Of Broad Application, The Metropolitan Corporate Counsel (2014). (An interview with Matthew R. Lynde, Vice President of Cornerstone Research).