One of the key factors in a likelihood of confusion analysis is a comparison of the similarity or dissimilarity of the goods or services offered under the two marks – the second du Pont factor.
One way to determine if consumers would understand two products or services to have a common source is to use the information already on trademark registry – aside from occasional dead wood, it’s a great compendium of information about what marks are in use in the US for what products and services. The Trademark Office has long permitted limited examples of this sort of evidence, and set out the current standards for this evidence in the Mucky Duck case.
As the Office summarizes more succinctly than I can, in TMEP 1207.01(d)(iii):
Third-party registrations that cover a number of different goods or services may have some probative value to the extent that they may serve to suggest that goods or services are of a type that may emanate from a single source, if the registrations are based on use in commerce. In re Mucky Duck Mustard Co., 6 USPQ2d 1467, 1470 n.6 (TTAB), aff’d per curiam, 864 F.2d 149 (Fed. Cir. 1988).
More recent case law suggests that the propounding party also needs to provide evidence of current use for the relevant goods/services. The sweet spot seems to be between 3-6 examples of registrations that cover both Party A’s goods/services and Party B’s goods/services. See, e.g., In re Albert Trostel & Sons Co., 29 USPQ2d 1783, 1785 (TTAB 1993). More than this can easily be considered “duplicative.”
What’s missing is context. Are these five registrations the only registrations that share the goods, out of thousands that include one but not the other, or are they shared a significant percentage of the time? It’s extraordinarily easy to get at least a rough idea of how common the overlap between two goods or services is – but the current legal standard does not allow Examining Attorneys or Applicants even attempt to do this. TESS makes this sort of comparison pretty simple:
Good A: Shaving cream: 1189
Good B: Razor: 1153
Good A + B: Shaving cream and razor: 61
Good A + B / Good A: 5.13%
Good A + B / Good B: 5.29%
Unfortunately, Office rules expressly prohibit this type of use of TESS. USPTO online records are fine for certain types of uses that can be accessed via TESS, like current status and title information from TSDR, but not use of TESS results as such, which would be required in order to present statistical results in a way that is not crushingly paper-intensive.
Updating the Mucky Duck principals to allow Examining Attorneys and Applicants to incorporate basic statistical information about the extent of overlap between relevant goods and services could significantly improve the quality of examination and TTAB decisions. That data may not have been widely available and easily accessible in 1988, but it is now, and the Office and Board should take advantage of it.
 This does not account for the same company owning registrations for the same mark for Goods A and Goods B in different registrations rather than in the same registration.
 37 CFR § 2.122(d)
 See, e.g., The Crazy Horse Memorial Foundation v. Spencer, Opp. No. 91204980 at *41 (TTAB Feb. 26, 2014) (“Applicant also relied upon a summary list of a search of the USPTO’S TESS database for live registrations and pending applications that include ‘Crazy Horse.’”)
 Pure stats shouldn’t be the be-all and end-all – the fact that a few gigantic keiretsu make two very different types of products doesn’t mean that customers generally consider them to be the sort of products that are generally from a common source.