The European Commission’s proposed essentiality checking and patent counting at the EUIPO is troubling. While parties are entitled to present whatever methods and studies they wish to imply Standard Essential Patent (SEP) portfolio strength in licensing negotiations or to the courts in litigation, the proposed registry with mandatory essentiality checking on random samples of patents will give a false sense of security on the applicability, accuracy and reliability of such checks, measures and any royalty charges derived from them. Essentiality determinations and patent counts provide a poor and unproven gauge of patent portfolio strength. Methods fail a key integrity test for any evaluation or measurement system because results are not reproducible. Despite the EUIPO being ordained the official authority on determining patent essentiality, its checking will be as contestable technically as for private evaluators and their studies that already check essentiality, count patents and invariably disagree with each other. Nevertheless, even though determinations are non-binding they will have significant sway with the courts.
A European Parliament press release issued following a January 2024 Legal Affairs Committee vote to adopt “New rules to promote standard-setting innovation in new technologies” states that “in 5G almost 85% of the standard essential patents are in fact non-essential. The new essentiality test will stop the occurrence of over-declaration”.
Some studies do indeed indicate essentiality rates of only 15% (i.e. 100%-85% = 15%) or even less in some cases—which might well be correct—but there is no evidence to support the latter contention that checking will improve the declaration behaviour of patent owners. There is no shame or sanction for over-declaration. Bias in essentiality checking—that is most severe at such low essentiality rates—means that the effects of over-declaration can only be somewhat moderated by checking. Over-declaration can never be anywhere near eliminated. The bias incentivises over-declaration despite checking. Rather than stopping over-declaration, institutionalized checking by the EUIPO will likely motivate patent owners to game the system by declaring even more patents of dubious essentiality.
Essentiality is subjective and only one among various factors affecting patent strength
Patent strength is a function of validity, infringement and technical contribution, as well as essentiality to the standards. While some parts of standards go unimplemented, are rarely used, become obsolete or are peripheral to where standards provide most innovative value, other parts are fundamental to very significant improvements with new technologies such as 5G. For example, various radio access network technology improvements have increased network speeds and capacities one hundred thousand-fold (e.g. from 10 kbps to 1 Gbps) since the introduction of 2G data in the mid-1990s.
Some characteristics can be objectively, reliably and reproducibly checked, others cannot. Patent essentiality and validity are matters of judgment where different assessors will often disagree about what are ostensibly yes-no decisions. As stated by the judges’ decisions in Unwired Planet v Huawei and TCL v Ericsson, respectively:
“Based on my assessment of both experts, I am sure the disagreement represents cases in which reasonable people can differ.” (Paragraph 335.)
“Given the somewhat subjective nature of these determinations, ‘disagreements’ is probably a more accurate label than ‘error.’" (Footnote 16.)
By way of analogy: on the one hand, selections of beauty pageant and international song contest winners are also subjective tasks that can be swayed by judges’ predilections and do not have reproducible results with different assessors; on the other hand, and in marked contrast to all the above, checks such as the UK’s annual car roadworthiness MOT test is highly objective and reproducible. Two different test centres would reliably come up with the same pass-fail result for the same car after verifying that brakes work, turn indicators flash, and measuring that tyre tread depth is sufficient, among other checks.
Determining true essentiality is made more difficult by the fact that patent counters have very different objectives to those agreed by consensus in Standard Setting Organizations. ETSI merely wants to ensure standards such as 5G are not blocked by demanding patent owners declare whether they believe a patent might be or might become essential. ETSI never checks essentiality and does not want to do so. Essentiality declarations such as those in ETSI’s IPR database were never intended to be used for royalty rate determinations in Fair, Reasonable and Non-Discriminatory (FRAND) licensing, as sought by the Commission with the EUIPO’s registry and additional steps of essentiality checking and patent counting.
Only the courts can definitively determine which patents are truly essential, which are not invalid and valuate portfolios. Cases in litigation illustrate how uncertain everything is and how expert opinions differ. The challenges in assessing essentiality were extensively discussed at the conference. Issues include interpretation of patent claims and that the scope of these can be entwined with validity. Prosecution history can be pertinent in making determinations. As patents are amended to cover the standard they can include what has been contributed to the standard by others. With many patents being found invalid by the courts, validity should not be ignored on the path to determining value, as it is in the Commission’s proposed checking. Validity can be the most significant factor affecting SEP value.
In FRAND litigation, highly experienced top minds including judges, experts and those representing the parties spend many months at multi-million dollar costs evaluating and deliberating—with various disagreements on essentiality and validity of litigated patents prior to judgment—even though typically only a few patents are examined.
If all that work including analysis of claim charts and patent prosecution histories is actually required to do a proper job on only a few patents, how can we trust the accuracy of the EUIPO’s experts examining orders of magnitude more patents and whose determinations ignore the crucial issue of validity? While the courts tend only to have the resources to do the required assessments on no more than a handful of patents in each case, there are many tens of thousands of patents and patent families declared essential to standards such as 5G. It is unsurprising that the UK courts have tended to reject patent counting as a means of determining FRAND royalties, except in some cases as a cross-check for determinations primarily based on comparable licensing agreements.
Unscalable checking
One proposed way dealing with the insurmountable task of checking all a standard’s declared patents is to check only random samples of them. The hope is that it would be possible to check a manageably small number of them very thoroughly and accurately. Consensus is that accuracy can best be achieved with the preparation and use of claim charts.
However, there are several problems with this approach, as illustrated in my empirical research in 2021 and 2022:
- Even when claim charts are used to assist in determining essentiality, different assessors still disagree widely in their determinations with agreement on only around 83% of them. That’s not as good as it might seem when one considers that different assessors can be expected to agree on precisely 50% of them if one assessor was making determinations randomly based on the flip of a coin. If two assessors disagree in their determinations, at least one of them must be wrong. However, if two different assessors are in agreement, that does not mean the determination is correct.
- Inaccurate determinations cause a substantial upward systematic bias in essentiality rates derived after checking. My empirical research shows that the proportion of false positive essentiality determinations will greatly outnumber false negatives at essentiality rates of 15% or less.
- Sample sizes need to be large (e.g. >1,000) if true essentiality rates are at 15% or below and if, for example, accuracy within ± 15% at the 95% confidence level is required. Sampling error as a proportion of true essentiality rate increases at lower and lower levels of true essentiality.
- Sampled patents cannot be appealed and reassessed without destroying the integrity of the sample. For example, if one in ten patents is sampled the determination has a 10x effect implied in the entire population count. With inevitable selection bias in appealed patents, “corrected” determinations will have a distorted and magnified effect implied in the overall population.
- However, it would be to deny justice not to allow some kind of appeals procedure on determinations made by a public authority. This issue could weigh heavily in FRAND dispute litigation.
- It’s very costly. Ericsson testified in TCL v Ericsson that it took 50 man-hours per patent to prepare claim charts.
What the Commission is seeking to concoct at the EUIPO will produce yet more patent counting studies, somewhat like what PA Consulting has been producing regularly for years and that several other firms have published from time to time. PA’s studies are widely used because others use and seemingly take heed of their results—not because they are proven to be accurate and reliable, because it is impossible to prove that. Here’s what Justice Smith had to say in the Optis v. Apple judgment:
“So, as with validity – but for different reasons – making a judgement about levels of essentiality in the stack is unreliable and unsafe.
My conclusion is that – accepting entirely that PA Consulting seeks to do a careful job – for the purposes of a judicial determination of what is fact, the PA Consulting/Optis approach to determining Stack size (or the figure for the denominator) is not to be relied upon.
I accept that were a reliable qualitative assessment to be possible, that might well be preferable. But an unreliable qualitative assessment – especially where even the magnitude of the error is unknown – is not (in my judgement) an acceptable metric to use when seeking to answer the FRAND Question.
I cannot use the PA Consulting data as a metric in answering the FRAND Question.”
Patent counting is simplistic
Even checking both essentiality and validity is woefully insufficient in determining patent value. It’s widely recognized that different patents vary in value enormously—by orders of magnitude from virtually worthless to some real gems. The significance of a patent’s technical contribution to a standard and value in implementation can vary from being seldom used or of marginal worth to being fundamental functionality that might enable major cost savings or increases in customer utility or revenues to be generated. Convenience aside, there is, therefore, no basis for assuming that portfolio value is in proportion to any kind of patent count (i.e. of declared, found essential or found not invalid patents). On the contrary, some patent owners likely have a much larger proportion or number of gems than others. However, even approximately how much more is an unanswered empirical question.
Ministry of Patent Counting and Red Tape
The Commission’s proposals for registering, checking and counting patents, among other demands in the proposed legislation, is also in conflict with the stated objectives of European leaders.
French and German leaders Emmanuel Macron and Olaf Scholz recently co-wrote an op-ed in the Financial Times setting out some laudable objectives:
“With an ambitious industrial policy, we can enable the development and rollout of key technologies of tomorrow, such as AI, quantum technologies, space, 5G/6G, biotechnologies, net zero technologies, mobility and chemicals.
We call for strengthening the EU’s technological capabilities by promoting cutting-edge research and innovation and necessary infrastructures.
We call for an ambitious bureaucracy reduction agenda to deliver on simpler and faster administrative procedures and cutting bureaucratic burdens for businesses of all sizes. We welcome the European Commission’s initiative to reduce reporting obligations for our companies by 25 per cent.” (hyperlink added)
The Commission’s proposed demands for patent registration at the EUIPO, together with preparation and submission of additional information such as patent claim charts will substantially increase administrative burdens for European companies such as Ericsson and Nokia that remain dependent on SEP licensing income. These new burdens will cause friction, delays and diminution in the well established, highly effective and self-sustaining innovation loop in which licensing fees are used to fund further R&D, leading to the creation of yet more valuable new technologies.
Better to have scarce and costly technical experts innovating and prosecuting their own patents, or designing and testing new products, rather than tying up hundreds of them generating additional information for checkers and in doing the checking—at patent owners and at the EUIPO, respectively.
Transparency about what?
There was consensus at the conference that greater transparency could help with FRAND licensing for SEPs. However, rather than burdening licensees with voluminous disclosures on patent claims and with delays while conciliators deliberate about aggregate royalties and technical experts check patents for essentiality, it would be better to have licensors and licensees disclose more about actual licensing. This should include terms in licensing agreements and information on licensed trade including volumes and prices. If parties are unwilling to make such information public, then it could be disclosed to a confidential repository with limited access, information anonymised and other safeguards. Let’s find out more about what’s happening already and rely on that, rather than trying to make things up with top-down rate setting.
In Q&A under the Chatham House Rule, I asked another panel whether a modicum of accuracy and reliability can be achieved in essentiality checking to determine patent portfolio strength. Bad news — no. Good news — it’s probably not necessary because most licences get agreed, regardless.
While I believe it would be best for the Commission to abandon is proposed checking and rate setting, if it does proceed it should consider recommendations about how to do that competently and with recognition of limitations, as explained in my publications cited with hyperlinks in this article.
Keith Mallinson, founder of WiseHarbor, has more than 25 years of experience in the telecommunications industry as a research analyst, consultant and testifying expert witness.
No comments:
Post a Comment