Saturday, June 12, 2010

The illusory concept of a "very high quality patent"

In a recent post, the IAM blog features work by one Rahul Jindal. Prior to working at CPA Global, Jindal was at Pangea3 and, before that, at EvalueServe.

Of the work in question, concerning patents sold by Micron to Round Rock, the IAM Blog writes:

In work done exclusively for IAM, Jindal finds that while just 5% of the entire universe of USPTO patents are of very high quality, that description applies to 20% of the patents Micron sold to Round Rock, the NPE set up by the company's former outside patent counsel John Desmarais. What's more, a further 29% fall into the high quality band, as compared to 15% of all US patents. Jindal assesses just 22% of the sold Micron patents as being of low or very low quality; by contrast, 50% of all US patents fall below the medium quality threshold.

There is no mention within the post of "how" a patent becomes of "very high quality". In past assertions of patent quality, one finds bogus indicators like patent citations. In commissioning Jindal to develop this stuff, IAM is almost at the level of "manufacturing" an issue. To get the real scoop on what's happening in patent citations, check out:

Edlyn S. Simmons and Nancy Lambert. "Patent Statistics: Comparing Grapes and Watermelons" In Recent Advances in Chemical Information, Proceedings of the 1991 Montreux International Chemical Information Conference & Exhibition, H. Collier, Ed, (Royal Society of Chemistry, Cambridge; CRC Press) pp 33-78 (1992).

Simmons, Edlyn S.; Lambert, Nancy. "Comparing grapes and watermelons." ChemTech 23 (6), 1993, p. 51-59. ]

**See also

There they go again: patent citation studies

**See separately

Different takes on patent reform
, including

Comment to IAM blog on May 11-->

Of the rhetorical question -- What else could the USPTO Director have said? --, the situation that Kappos faced in the House in May 2010 was quite foreseeable in March 2010. [See ] Within the IT industry, there is not much traction for the current form of S.515, as was evinced by the Coalition for Patent Fairness indicating it was worse than the present state of affairs. Kappos and Locke could have easily predicted where Lofgren was going to go, and Kappos could have said "yes, in the absence of the other proposed reforms, please end fee diversion so the USPTO at least has enough money to examine patent applications." Holding onto S.515 when key House members are not buying in does not advance the interests of patent office customers, who are really suffering. Symbolically, the strategy of the USPTO officials should be rejected under KSR for failing to recognize the obvious.

In light of the efforts by Zoe Lofgren since May 11, and in light of the non-action on S.515 since May 11, one suspects that there is lot more that Director Kappos could have said, but didn't. While political games are played, the Patent Office, and its customers, suffer. That problem is more significant than the deal between Micron and Round Rock.

**Separately, of interest

Judge Ward to Plaintiff: Accept $51 Million Remittitur or Face New Trial on Damages
. Concerning damages: At best, [plaintiff's expert] testified that almost all computers sold in the retail market include optical disc drives and that customers would be hesitant to purchase computers without an optical disc drive. This evidence notwithstanding, there was no evidence from which the jury could conclude that the patented features of the invention formed the basis for the customer’s demand for the entire computer. [One could also query Jindal as to what "proof" he has that something he says is a very high quality patent actually IS a very high quality patent.]

On Dan Ackroyd's skull-shaped Vodka bottle. Within there is mention of a bottle shaped like a casket with the word "poison" and a skull emblazoned on the front. Every time someone talks of metrics for "high quality patents," think of poison and a skull.

Further to the comment below, Rahul Jindal is invited to a post a link to any site giving details of his methodology and an explanation of why it is useful.

As an update on June 17, 2010, Rahul Jindal has provided no link. One wonders if there is substance to the work.

UPDATE: It's June 18 and no link from Rahul Jindal.

UPDATE. On June 21-22, the additional comments appeared. At to the link:

The patent scoring has a significant meaning, I believe, because it has been validated on numerous occasions (hundreds of entries in my validation tracker). Some of them (all based on publicly available patent numbers and related cases) are in the table below. There are several others which cannot be revealed in a public forum like this.

The relevance of the scoring system has also been validated on a large set of abandoned patent data set, containing nearly 350,000 patents. It has been found that the system categorizes nearly 86% of the recently Abandoned patents (of last 5 years) into Low value (bands Low, Very Low) segments.

There's no there, there.


Blogger Rahul said...


You are more than welcome to learn about how we rate patents. Just ask!

We spent several man years researching for and coming up with our methodology employing PhD statisticians working with my team of IP experts.

However, you and all your readers should know that we ourselves ensure that no discussion on our patent scoring methodology is complete without mentioning that it is only a filtering tool to enhance the efficiency of any portfolio analysis endeavor. It gets you from point A to point B faster; and yes, there may be false positive and false negatives. Or else, there will be no need for human expert analysis. Our methodology does not replace human expert analysis; only enables a much better use of it.

I would be glad to send a comprehensive whitepaper to you or any of your readers if that will be of interest.


1:21 PM  
Blogger Kyle Jensen said...

I would love a link to a whitepaper! Very interested to see details.

4:22 PM  
Blogger Rahul said...

Here's some detail on validation of patent scoring on real publicly available data:

The system is designed to highlight outliers. If a patent's characteristics are such that it scores among the top 5% of the set of ALL US granted patents, then it is labelled "Very High", next 15% as High, next 30% as Medium, next 30% as Low and the last 20% as Very Low.

All that a statistically modeled patent scoring methodology can do is to highlight patents that exhibit characteristics which are statistically similar to the characteristics exhibited by patents known to be good (in the form of high sale values, victory against invalidation, etc). There will be false positives and false negatives - which is true for any statistical modeling based system, that is why we recommend using the patent scoring as a first filter; an efficiency tool directing where to deploy human assessment first.

11:21 PM  
Blogger Rahul said...

Hi Kyle,

Happy to mail you the whitepaper - please let me have your mail id.


11:24 PM  

Post a Comment

<< Home