How good has TMView become?

TMView, the free trade mark register search platform maintained by OHIM, has increased its coverage tremendously. Today, it covers all 28 member states of the EU, as well as Russia, Turkey and the US (federal register). Has it reached a stage where it can replace commercial trade mark databases?

I am going to compare TMView with the online search tool of one of the leading commercial trade
mark search providers, which shall remain unnamed for reasons "I don't want to be sued".

Coverage

While the coverage of TMView is broad, it is limited essentially to Europe and a few select - albeit in the case of the US important - countries elswhere in the world (Canada just joined mid-September). One notably missing register is the Swiss one. The commercial database has a far wider coverage (including US state marks) extending to Asia Pacific, Africa and Latin America. However, if you are only interested in the European market, that is pretty irrelevant. So how do the search capabilities for the European registers compare?
For all the following searches, I limited the registers to those of the 28 EU member states and the USA (including International Marks extending to those countries).

Completeness of searches

To get a feel for the completeness of the results of basic searches, I am going to do some simple searches (the TMView input for all searches can be found in the footnotes).

A search for LUKAS, exact match, in Nice class 21, all status of marks results in 8 hits in TMView and 12 in the commercial database. The difference results from the fact that the commercial database also returns partial hits, namely LUKA, which TMView does not with exact match.

A search for all 3-dimensional marks in Nice Class 5 with the Vienna Code "02" (human beings), limited to registered marks, results in 44 hits on TMView. The help file for TMView says that you cannot search for top level Vienna Codes, but that is not true (note that all Vienna Codes need a leading zero on TMView).
The commercial database does not allow the use of Vienna Codes. Without the Vienna Code, the search returns 153,336 hits. Something clearly went wrong - there are not that many registered 3-D marks in class 5. TMView returns "only" 1,053 for the same search. The problem seems to be that the commercial database classifies figurative marks without a word element as "3 dimensional" marks.

The next search is for registered colour marks in class 35. TMView returns 118 hits, among them figurative marks with a colour claim (not only abstract colour marks). The commercial provider does not allow a limitation to colour marks. They fall into the "other" category, but this search returns 329,898 hits, so it's useless (the problem seems to be the French register, where all marks are classified as "other").

Wildcard searches

In TMView, the wildcard character * stands for none or any number of characters. It is not possible to limit the wildcard to an exact number of characters. The commercial provider allows different wildcards that stand for any number of characters or exactly one character, or even just a vowel or a consonant.

An exact search for VITA in class 28, registered marks only, in TMView returns 167 hits. The same search in the commercial register returns 117 hits. The difference stems from the way the exact search is implemented: while both databases are supposed to only return hits where VITA is not joined with other characters (i.e., DOLCE VITA, but not VITAMIN), TMView in fact returns some

This cat is really excited about wildcard searches

hits where VITA is part of a word (e.g., IR 658 301 VitaLIFE, which is not returned by the commercial search).

A search for VITA* in TMView with otherwise the same criteria returns 472 hits, the same search in the commercial database 435 hits. The wildcard really makes a difference here - there are a lot of trade marks containing VITAMIN that are not found with VITA, but turn up when searching for VITA*.

Similarity search versus "fuzzy" search

The advanced search option of TMView provides a fuzzy search option. This allows searching for similar marks. The similarity is assessed by the the Levenshtein distance, a string metric for measuring the difference between two sequences. Informally, the Levenshtein distance between two words is the minimum number of single-character edits (i.e. insertions, deletions or substitutions) required to change one word into the other. The practitioner may raise his or her eyebrow - this has little to do with similarity as it is assessed in trade mark law. The commercial search provider, on the other hand, provides a similarity search that recursively replaces the letters in the searched for string with similar sounding letters, which is closer to how humans assess the similarity of marks.

PRO TIP: TMView allows setting the degree of similarity under the fuzzy search option. Somewhat counter-intuitively, the higher the percentage of "fuzziness", the more similar the mark. If you want to cast a wide net, select a low percentage.

A fuzzy search for WIMETU in TMView, with fuzzy set to 50%, all classes, all status, returns 85 hits. Cranking the similarity down to 10% does not return more hits. At 90%, on the other hand, 0 hits are returned. The fuzzy factor is important. Generally, you want to put it lower than the default 70%.

In the commercial database, a phonetic search for WIMETU returns 0 hits. The similarity search returns 316 hits (and looking at those hits costs a cool EUR 3,500).

Interestingly, neither of the searches returns any hits for marks comprising WINNETOU (and there are several of those), although I would consider WINNETOU and WIMETU to be similar.

Finally, I searched TMView for CHANNEL in class 3, registered marks only, fuzzy set first at 50% and then at 10%. TMView finds 15 hits for both searches. The similarity search of the commercial provider, on the other hand, returns 3,720 hits. Among them also the (numerous) marks containing CHANEL registered in class 3 (where cosmetics and perfumes are at home). TMView misses the CHANEL marks even with the broadest possible search, which is somewhat surprising since CHANEL and CHANNEL only differ in one character and are undoubtedly highly similar.

Alerts

TMView allows to set alerts when the status of a specific mark changes. It does not allow setting an alert for when the result of a specific search changes. It is not a suitable tool for checking the register for the registration of conflicting marks. The commercial provider, on the other hand, allows to set alerts when certain strings are registered, allowing a collision search.

Conclusion

As should be evident, there is no clear "winner" here. If you like wildcard searches, the commercial provider's offering is far more comprehensive than TMView (but then so few people use wildcard searches that Google has disabled the feature by default on its main search page). TMView cannot be used for collision searches, although technically, it should not be difficult to implement this.

For clearance searches, TMView's fuzzy search is too limited. A similarity search by a commercial provider - while returning many more hits - also provides no certainty that there are no conflicting rights, as the example with the search term WIMETU has shown. But currently the fuzzy search of TMView cannot replace even an online similarity search by a leading commercial provider. If you do use it, set the "fuzzy factor" to 10% and pray that you haven't missed anything.
____________________________________________________
Input for TMView searches (if you want to replicate the searches, just paste the part starting with oc: into the TMView search field):
LUKAS: oc:AT,BG,BX,CY,CZ,DE,DK,EE,ES,FI,FR,GB,GR,HR,HU,IE,IT,LT,LV,MT,PL,PT,RO,SE,SI,SK,US,EM,WO AND tm:LUKAS AND nc:21
3D: oc:AT,BG,BX,CY,CZ,DE,DK,EE,ES,FI,FR,GB,GR,HR,HU,IE,IT,LT,LV,MT,PL,PT,RO,SE,SI,SK,US,EM,WO AND ty:3-D AND nc:5 AND sc:Registered AND vc:02
Colour: oc:AT,BG,BX,CY,CZ,DE,DK,EE,ES,FI,FR,GB,GR,HR,HU,IE,IT,LT,LV,MT,PL,PT,RO,SE,SI,SK,US,EM,WO AND ty:Colour AND nc:35 AND sc:Registered
VITA*: oc:AT,BG,BX,CY,CZ,DE,DK,EE,ES,FI,FR,GB,GR,HR,HU,IE,IT,LT,LV,MT,PL,PT,RO,SE,SI,SK,US,EM,WO AND tm:VITA* AND sc:Registered AND nc:28
WIMETU: oc:AT,BG,BX,CY,CZ,DE,DK,EE,ES,FI,FR,GB,GR,HR,HU,IE,IT,LT,LV,MT,PL,PT,RO,SE,SI,SK,US,EM,WO AND tm:WIMETU~0.5 AND sc:Filed,Registered,Expired,Ended
CHANNEL: oc:AT,BG,BX,CY,CZ,DE,DK,EE,ES,FI,FR,GB,GR,HR,HU,IE,IT,LT,LV,MT,PL,PT,RO,SE,SI,SK,US,EM,WO AND tm:CHANNEL~0.1 AND nc:3 AND sc:Registered