[Gnso-rpm-trademark] REMINDER - Proposed Agenda - RPM PDP WG TM Claims Sub Team - 27 March 2019

Wed Mar 27 16:15:47 UTC 2019

That data would be available via the registry monthly reports, so
someone would be able to compile it. But, I don't see how it could
possibly be relevant to a study of the efficacy of TM Claims notices.
All it would reveal is the fraction of registration attempts that
generate TM Claims notices. Interesting, but not really relevant. It
might be of interest to those who want "expanded matches".

Sincerely,

George Kirikos
416-588-0269
http://www.leap.com/

On Wed, Mar 27, 2019 at 12:09 PM Michael Graham (ELCA)
<migraham at expediagroup.com> wrote:
>
> By "applications" I meant the total number of applications to register second level domain names in the New gTLDs.  I presume this information is available -- but I do not have the ability to access it at this time.
>
> Michael R.
>
> -----Original Message-----
> From: Gnso-rpm-trademark [mailto:gnso-rpm-trademark-bounces at icann.org] On Behalf Of George Kirikos
> Sent: Wednesday, March 27, 2019 8:59 AM
> To: gnso-rpm-trademark at icann.org
> Subject: Re: [Gnso-rpm-trademark] REMINDER - Proposed Agenda - RPM PDP WG TM Claims Sub Team - 27 March 2019
>
> Hi Michael G.,
>
> I'm not sure what you mean by "applications"? I linked to the Analysis Group report, which shows their methodology.
>
> A TM Claims notice is generated when a registrant attempts to register a domain name and the domain is an exact match to a recordal in the TMCH. So, 1.8 million domain registration attempts matched a TMCH entry, generating 1.8 million notices, with 93.7% abandoned (after excluding the 2 registrars, as noted in separate emails, whose abandonment rates were close to 100%).
>
> Sincerely,
>
> George Kirikos
> 416-588-0269
> http://www.leap.com/
>
>
> On Wed, Mar 27, 2019 at 11:46 AM Michael Graham (ELCA) <migraham at expediagroup.com> wrote:
> >
> > George:
> >
> > That's 1.8 million notices out of how many applications?
> >
> > Michael R.
> >
> > -----Original Message-----
> > From: Gnso-rpm-trademark [mailto:gnso-rpm-trademark-bounces at icann.org]
> > On Behalf Of George Kirikos
> > Sent: Wednesday, March 27, 2019 8:38 AM
> > To: gnso-rpm-trademark at icann.org
> > Subject: Re: [Gnso-rpm-trademark] REMINDER - Proposed Agenda - RPM PDP
> > WG TM Claims Sub Team - 27 March 2019
> >
> > Hi Phil,
> >
> > Thanks for feedback. If we consult the Analysis Group report:
> >
> > https://community.icann.org/download/attachments/64066042/Analysis%20G
> > roup%20Revised%20TMCH%20Report%20-%20March%202017.pdf?version=1&modifi
> > cationDate=1490349029000&api=v2
> >
> > Footnote 55 on page 17 documents that "As discussed in Section IV, there are two registrars that averaged downloads of more than 20 trademark strings per download, which is large compared to the average of fewer than five trademark strings in the downloads of other registrars. We also exclude downloads made by ICANN’s monitoring system. The exclusion of the two registrars does not significantly impact our results. Inclusion of the two registrars shows that 99% of registrations are abandoned and 0.5% of completed registrations are disputed."
> >
> > Thus, the 93.7% figure is already an *adjusted* figure. Bulk mining of the TMCH to see whether a particular mark had been recorded likely came from those 2 registrars, where more than 99%+ of registrations were abandoned (probably close to 100%). (i.e. 99% is a blended rate of 93.7% and the rate from the excluded data).
> >
> > Even if there was mining at other registrars, it was likely to be small in relative terms. Remember, on page 17 it says:
> >
> > "As shown in Table 4, we find that 93.7% of the 1.8 million registration attempts that received a Claims Service notification were abandoned."
> >
> > And on page 6, there were 41,601 records in the TMCH itself. Someone who managed to do additional mining of the TMCH and who found *all* of the records would have a small impact on the 1.8 million registration attempts. Likely they wouldn't find all of them, either (i.e. obscure non-dictionary terms that have low commercial value are likely to not be tested).
> >
> > As for the abandonment rates for non new gTLD domain names (without the TM Claims notices), I agree we don't have that figure to compare.
> > But, I think it's safe to infer that any dot-com registrar that had a 93.7% abandonment rate would likely go out of business pretty quickly.
> >
> > Sincerely,
> >
> > George Kirikos
> > 416-588-0269
> > http://www.leap.com/
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> > On Wed, Mar 27, 2019 at 11:17 AM Corwin, Philip <pcorwin at verisign.com> wrote:
> > >
> > > Responding in a purely personal capacity, I find this mathematical analysis problematic for two reasons.
> > >
> > > First, it posits that there are "2 separate rates, namely the abandonment rate for cybersquatters, and the abandonment rate for good faith registrants". But there was likely a third group that might well have constituted a majority of non-completed registrations -- parties who had no intention of ever registering a domain but simply wished to test whether a particular mark had been recorded in the TMCH.
> > >
> > > Second, we have no baseline abandonment rate for domain registrations that do not generate a Claims Notice.
> > >
> > > Philip S. Corwin
> > > Policy Counsel
> > > VeriSign, Inc.
> > > 12061 Bluemont Way
> > > Reston, VA 20190
> > > 703-948-4648/Direct
> > > 571-342-7489/Cell
> > >
> > > "Luck is the residue of design" -- Branch Rickey
> > >
> > > -----Original Message-----
> > > From: Gnso-rpm-trademark <gnso-rpm-trademark-bounces at icann.org> On
> > > Behalf Of George Kirikos
> > > Sent: Wednesday, March 27, 2019 8:25 AM
> > > To: gnso-rpm-trademark at icann.org
> > > Subject: [EXTERNAL] Re: [Gnso-rpm-trademark] REMINDER - Proposed
> > > Agenda - RPM PDP WG TM Claims Sub Team - 27 March 2019
> > >
> > > Just to build on Rebecca's comment, we can create a mathematical model to understand things more easily, and try to transform the "descriptive" analysis into a more quantitative analysis.
> > >
> > > The overall abandonment rate (93.7% actual data from the Analysis Group report) really represents a blended rate consisting of 2 separate rates, namely the abandonment rate for cybersquatters, and the abandonment rate for good faith registrants (unproblematic domains like "cloud" or "hotel" or "one"). It's a weighted average of those 2 different abandonment rates, and we're not sure what the true "weights" should be.
> > >
> > > So, let's define some variables. Let:
> > >
> > > C = percentage of cybersquatting registration attempts G = good
> > > faith / unproblematic registration attempts P = abandonment rate for
> > > cybersquatting registration attempts presented with the TM Claims
> > > notices Q = abandonment rate for good faith registration attempts
> > > presented with the TM Claims notices R = overall abandonment rate
> > > (blended rate, a weighted average)
> > >
> > > It should be clear that:
> > >
> > > CP + GQ = R
> > >
> > > It's also true that  C + G = 100% = 1.000 = 1, so let's simplify the
> > > above even further
> > >
> > > C + G = 1 ----> G = 1 - C
> > >
> > > CP + (1-C)Q = R
> > >
> > > Let's go even further, and isolate our attention on "Q", which is essentially collateral damage of the TM Claims system. A high value of Q means a high abandonment rate for good faith registration attempts.
> > > Rearranging the above equation:
> > >
> > > Q = (R - CP) / (1-C)
> > >
> > > So, as "P" increases, then Q decreases, which should be obvious, since we're doing a weighted average of 2 different rates. Let's imagine a "best case" scenario, where P is the highest possible value, namely 100% ( = 1.0000), which would minimize the value of Q. That means 100% of cybersquatting attempts are thwarted by the TM Claims notices! The above equation becomes:
> > >
> > > Q = (R - C) / (1-C)
> > >
> > > Of course we know what R is empirically, the 93.7% figure from the Analysis Group report (0.937). Thus, we can simplify even futher:
> > >
> > > Q = (0.937 - C) / (1 - C)
> > >
> > > Let's try different values of "C" (percentage of cybersquatting
> > > attempts) to see what that implies about the value of Q in this best case scenario:
> > >
> > > C = 10% = 0.10   -----> Q = 0.93 = 93%
> > > C = 20% = 0.20   -----> Q = 0.92125 = 92.125%
> > > C = 30% = 0.30   -----> Q = 0.91 = 91%
> > > C = 40% = 0.40   -----> Q = 0.895 = 89.5%
> > > C = 50% = 0.50   -----> Q = 0.874 = 87.4%
> > > C = 60% = 0.60   -----> Q = 0.8425 = 84.25%
> > > C = 70% = 0.70   -----> Q = 0.79 = 79%
> > > C = 80% = 0.80   -----> Q = 0.685 = 68.5%
> > > C = 90% = 0.90   -----> Q = 0.37 = 37%
> > >
> > > The above figures should be startling. And remember, these are the *best case* scenarios. If it turns out that P, the abandonment rate for cyberquatters, is less than 100%, then the values calculated for Q (abandonment rate for good faith registration attempts) would become even higher. Indeed, because it's a weighted average, if P turns out to actually be lower than 93.7% (i.e. hardcore cybersquatters are ignoring the TM Claims notices, and proceeding to registration), then that means *all* the values above for Q would have to be *higher* than 93.7% (and that would put constraints/bounds on the value of "C", otherwise one would generate impossible values of Q that exceed 100%).
> > >
> > > But wait, there's more! What's a realistic value for "C"? That's a critical value....do we have any evidence as to what percentage of registration attempts are problematic//cybersquatting, vs. those that are made in good faith and non-problematic? We do! We have the top 10 most frequently requested strings as per the analysis report, which Rebecca referenced, ALL of which were common terms like "ONE" or "HOTEL" or "CLOUD" which have multiple legitimate and non-infringing uses. (We also know that the number of domains disputed via a URS or UDRP is also small) Thus, we can infer from that empirical data that the value of "C" is not close to 100%, but is closer to 0%. That implies that the collateral damage, even in the best case scenario with 100% effectiveness of TM claims notices on cybersquatting registration attempts, swamps the possible benefits of the TM Claims notices.
> > >
> > > Sincerely,
> > >
> > > George Kirikos
> > > 416-588-0269
> > > http://www.leap.com/
> > >
> > > On Tue, Mar 26, 2019 at 1:31 PM Tushnet, Rebecca <rtushnet at law.harvard.edu> wrote:
> > > >
> > > > I was unable to attend the Kobe meeting, so I wanted to post my view on the first tentative answer:
> > > >
> > > > QUESTION 1 Is the Trademark Claims service having its intended effect?
> > > >
> > > >
> > > > (a) Is the Trademark Claims service having its intended effect of
> > > > deterring bad-faith registrations and providing Claims Notice to
> > > > domain name applicants?
> > > >
> > > > Tentative answer of Subteam in Kobe session:  Probably
> > > >
> > > > RT: We don't have enough information to make any kind of judgment about this.  It might be a deterrent, if you believe that cybersquatters can be deterred by a notice, but we have collected zero data that indicate that this is true.
> > > >
> > > > Where data are absent: (1) what percentage of registration attempts that received a notice turned back because of the notice, with a huge potential range; (2) what percentage of registration attempts that received a notice were “cybersquatting” or even “inadvertent trademark conflict” as opposed to completely unproblematic uses of strings like cloud and hotel; (3) whether the cohort that turned back was (a) proportionate to the problem intended to be solved, (b) disproportionately made of cybersquatters or other potential infringers, or (c) disproportionately made of legitimate potential registrants (since we have no data on what deters cybersquatters nor any other information about how people who intend to cybersquat respond to notices); (4) whether the absolute number or relative percentage of cybersquatting attempts changed when there was no longer a notice in place (as a change in behavior post-notice period, or its absence, could have provided relevant evidence); (5) we have no data about what the general incidence of cybersquatting in the new gTLDs is in the first place, whether in absolute terms or as compared to legacy gTLDs without the notice system.  In sum, there is evidence that notice may increase costs and deter registrations, but no evidence about what those registrations would have been in the absence of notice.
> > > >
> > > >
> > > > At best, it is "possible" that the Claims service is having its intended effect. There is insufficient evidence that this effect is "probably" happening.
> > > >
> > > >
> > > >
> > > >
> > > >
> > > >
> > > > Rebecca Tushnet
> > > > Frank Stanton Professor of First Amendment Law, Harvard Law School
> > > > 703 593 6759
> > > _______________________________________________
> > > Gnso-rpm-trademark mailing list
> > > Gnso-rpm-trademark at icann.org
> > > https://mm.icann.org/mailman/listinfo/gnso-rpm-trademark
> > _______________________________________________
> > Gnso-rpm-trademark mailing list
> > Gnso-rpm-trademark at icann.org
> > https://mm.icann.org/mailman/listinfo/gnso-rpm-trademark
> _______________________________________________
> Gnso-rpm-trademark mailing list
> Gnso-rpm-trademark at icann.org
> https://mm.icann.org/mailman/listinfo/gnso-rpm-trademark