Skip to main content

The Specialists

In 2020 a USPTO examining attorney said applications are assigned at random, except for a short list. We tested the list against every office action since 2014. Three of five hold. One had already ended when she named it. The mark name cannot test the fifth. A desk forms on a date, and the rule reads meaning inside a class.

September 16, 20269 min readFirst published August 6, 2026 · rewritten on a re-measurement
Class 33 words by the share their busiest examiner holdsTwo hundred fifty-seven words sorted from highest to lowest share. TEQUILA 97.4 percent, NAPA 92.1 percent, MEZCAL 80.9 percent, then three words near 26 percent, five between 10 and 17 percent, and a long floor under 10 percent that ends at 0.9. Median 2.9 percent.0%50%100%BUSIEST EXAMINER'S SHARETYPICAL WORD 2.9%TEQUILA 97%NAPA 92%MEZCAL 81%257 WORDS OF CLASS 33, SORTED
Each bar is one word that appears on 100 or more Class 33 (wines and spirits) applications filed since 2014. Its height is the share of those applications whose first office action came from a single examiner. Three words have an address; the typical word's busiest examiner holds 2.9%. Public USPTO records, measured September 4, 2026.
The three words above 30%, and the distribution of the other 254.
WordApplicationsTo the busiest examinerShare
TEQUILA1034100797.4%
NAPA16415192.1%
MEZCAL28322980.9%
All 257 wordsMedian busiest-examiner share 2.9%, 90th percentile 6.7%, three at or above 30%.

The USPTO has no specialist examining units on paper. The record shows a handful of real ones, each tied to what a word means inside one class. The rest of the register is a draw.

§ 1 · The list

Three of five hold.

The claim comes from an examining attorney ten years into the job, interviewed in 2020. She said applications are assigned “entirely randomly, barring specific special circumstances,” and she named them: sound marks, color marks, political marks, Olympic marks, some categories such as cheese, tequila and cannabis, and marks related to major events such as 9/11 and Kobe Bryant. That is the only insider account of routing on the public record, and it can be tested.

We took every application filed since 2014 that drew an office action and attached one examiner to each, the one who wrote its first office action. We matched whole words in the mark name and named the class. We split the record at 2020, the year of the interview.

What she namedWhat we countedFiled before 2020Filed 2020 or laterVerdict
TequilaTEQUILA in the mark, Class 33321 of 332 to one examiner686 of 702 to the same examinerHolds
Olympic marksOLYMPIC in the mark, any class31 of 31 to one examiner43 of 46 to the same examinerHolds
Kobe Bryant marksKOBE in the mark, any class3 of 26, spread across 23 examiners22 of 44 to one examinerHolds, from 2020
CannabisCANNABIS in the mark, any classTop three examiners held 45%Top three hold 13%Real, then ended
CheeseCHEESE in the mark, Class 29Top examiner 2.5% of 276Top examiner 3.5% of 286No sign on the mark name

Applications filed 2014 or later that drew an office action. One examiner per application, the one who wrote its first office action. Whole-word match on the mark name. Sound marks, color marks and political marks are on her list but not in this table. The public record we hold carries no drawing type, and we do not describe political desks. Her other event cell, 9/11, has 19 owners and fails the owner gate.

Tequila, Olympic marks and Kobe Bryant marks each route to one examining attorney, and two of the three did so before and after the interview. Olympic marks are protected by statute, and a desk there surprised us least. Cheese shows nothing on the mark name: 191 examiners in Class 29 before 2020 and 179 since, the busiest at 2.5% and 3.5%. That is a weaker test than the others, since most cheese marks do not say cheese. The cannabis desk was real, and it had already ended when she said it.

§ 2 · How a desk forms

Tequila, year by year.

Tequila is the one cell with enough filings every year to watch the routing change. In 2005, 45 tequila marks in Class 33 drew office actions from 31 different examiners. In 2010, 44 marks from 37 different owners drew them from one.

Filed inTequila marks, Class 33Different ownersExaminers who got oneThe busiest examiner's share
20054532316.7%
200652463328.8%
200745412733.3%
20085740982.5%
20094335395.3%
201044371100.0%
201144371100.0%

Applications with TEQUILA in the mark, in Class 33, by filing year, counting the examiner who wrote each one's first office action. Single years carry no owner gate; the point is the shape.

The volume did not move. The owners multiplied. The examiners went from 31 in 2005 to one in 2010. That examiner had a long record on ordinary marks before the first tequila mark reached him. Since 2014 he has written the first office action on 1,007 of the 1,034 tequila marks in Class 33. Those marks come from 815 owners and cover every one of thirteen filing years. A desk has a start date. Nothing on the public record says whether the work was sent to him or asked for; the record shows only where the files went.

§ 3 · How a desk ends

Cannabis, 2012 to 2025.

From 2012 through 2017, three examining attorneys wrote the first office action on 63% to 90% of the marks carrying a cannabis word, every year. In 2018 the category grew, and in 2019 it was more than triple 2017. That year it reached 185 examiners, and that year's busiest three held 14%.

Filed inCannabis-family marksExaminers who got oneShare held by that year's busiest three
201222686%
2013531570%
20141843679%
20152232390%
20162362070%
20173525263%
201858611122%
20191,28818514%
202061819811%
202152715919%
202235713214%
20232328426%
20241786225%
20251506423%

Marks carrying CANNABIS, MARIJUANA, MARIHUANA, CBD or THC, any class, by filing year. The last column is the combined share of the three busiest examiners in that filing year. From 2012 through 2017 they are the same three people every year; from 2018 on they are not. Filings for 2026 are still being examined and are left out.

The three did not leave. Their cannabis counts fell from 223 of 352 in 2017 to 119 of 586 in 2018, then 145 of 1,288 in 2019. Over the same years their total office actions held or grew. The work went elsewhere. Since 2023 the busiest three hold about a quarter, and they are different people. That is a real tilt on a category that now spreads across dozens of examiners a year, 64 of them in 2025.

A second example is smaller and falls below this page's gates. In Class 25, one examiner wrote the first office action on 24 of the 27 STRAIGHT OUTTA marks filed from 2014 through mid-2017. He wrote none of the 31 filed from 2018 on, while his own docket grew. Identical marks went to him in 2015 and to others in 2019.

This section corrects our August version. That version described a cannabis desk fifteen examiners strong, holding 40% of cannabis-named filings. That figure read the examiner from a field that the record fills on under a third of filings. It then picked the twenty examiners with the most cannabis and reported that they had a lot of cannabis. The series above replaces it.

§ 4 · What the rule reads

Meaning, inside a class.

Before 2020, KOBE on a trademark meant beef, a steak sauce, a tire company: 26 marks across 23 examiners, no pattern. Since 2020 the word means something else, and the routing knows it. The estate's own KOBE and MAMBA filings went to one examiner, 27 of 28. So did 33 more from 31 unrelated owners, WHAT WOULD KOBE DO and MAMBA MENTALITY among them. The Kobe leather, Kobe hot pot and green mamba marks, filed in the same years, did not.

The rule reads the class along with the meaning. TEQUILA outside Class 33, on apparel and restaurants and everything else, spreads across 243 examiners, the busiest at 14%. OLYMPIA, the city, correctly does not go to the Olympic examiner: 131 marks across 98 examiners. NAPA in Class 33 goes 151 of 164 to one examiner. SONOMA, next door, spreads 80 marks across 69 examiners. Champagne spreads across 101 examiners and sake across 98. The word cognac reaches 41, its busiest at 15%. Appellations are not a category. Mezcal has its own examiner, 150 of 153 since 2020. The tequila and mezcal desks share no application in either direction.

The Law Office is not the unit either. The other 17 examining attorneys in the tequila examiner's own office caught none of the 1,034. Whatever routes these marks, it routes to a person.

§ 5 · What it changes

The examiner and the ground are on the record before you file.

One registration, the certification mark TEQUILA, has been cited as a Section 2(d) bar against 351 applications since 2017. 340 of those citations came from the tequila examiner. 11 examiners have cited it at all, and the other ten account for eleven citations between them. The applications behind those citations come from 279 owners.

From inside the office it is a corner of the work. Wines and spirits, all of Class 33, are about one eighth of that examiner's docket on filings since 2022. Tequila marks are a slice of that eighth. From the applicant's side two things are close to certain. File a mark carrying TEQUILA in Class 33 and the examiner is known before the filing receipt arrives, 97 times in 100 since 2014. A likelihood-of-confusion refusal follows on 94 of those 100. The certification mark itself is the cited registration on about one in four. The citation issues under Section 2(d) like any other, on the same likelihood-of-confusion factors, with the certifier standing where a brand owner usually stands. In practice the answer lies in the identification of goods more than in argument about the marks. It can be written into the application as filed.

§ 6 · How rare

Six cells in fifteen thousand.

The chart at the top of this piece is Class 33: 257 words that appear on 100 or more applications, sorted by the share their busiest examiner holds. The typical word's busiest examiner has 2.9% of it. Three words sit above 30%, all three above 80%, and nothing sits between.

The same count across every class shows how rare this is: every word that appears on 100 or more applications of a single class, every filing since 2014. That comes to 14,935word-and-class cells. The median cell's busiest examiner holds 2.4%; the 90th percentile is 4.2%. 6 cells reach 30% and clear four times their class baseline. Three are the Class 33 desks on this page. Two are single filers' house accounts and fail the single-filer gate. One is a political word we do not describe. That is the complete list at that floor. Cells too small to reach that floor are not in the count, the Olympic and Kobe cells in § 1 among them.

§ 7 · Five ways this could be nothing

Each named, each checked.

Concentration has boring explanations. We tested each one before calling any cell a desk.

  1. Hypothesis 01

    It is one company filing in bulk.

    Ruled out: hundreds of owners

    The tequila cell holds 815 distinct owners since 2014 and its largest filer is 1.5% of it. Mezcal: 233 owners. Napa: 93. The Kobe Bryant desk took 33 marks from 31 owners unrelated to the estate. Every cell on this page clears 25 owners with no filer above 25%, or it is labelled as failing that gate.

  2. Hypothesis 02

    It is the companion rule: an owner's co-pending applications go to one examiner.

    Ruled out for these cells

    The USPTO does send an owner's related applications to one examiner, and that is what the owner gate is for. The companion rule cannot build a 97% share from 815 owners whose largest filer holds 1.5%. The rule does explain the filers this page leaves out: the slot-machine makers, a postal service, a racing series. Each is one owner and a house account.

  3. Hypothesis 03

    It is the class, not the word.

    Ruled out: the class is spread, the word is not

    The tequila examiner writes about 1.5% of the first office actions in Class 33 since 2014. VODKA in the same class reaches 706 examiners and the busiest holds 1.7%; WHISKEY reaches 646. The other 17 examiners in the tequila examiner's own Law Office caught none of the 1,034 tequila marks. Outside Class 33 the same word spreads across 243 examiners.

  4. Hypothesis 04

    The August figures were wrong, so these could be too.

    Different column, checked twice

    The August version read the examiner from a field that the record fills on under a third of filings. This version reads the examiner who wrote each application's first office action. The record carries that examiner for every application that received an office action: about 87% of Class 33 filings since 2014. Every cell here was re-derived on that basis between August 31 and September 4, 2026. The Class 33 word distribution was measured twice, on September 2 and again on September 4, with the same result. The cheese cell was measured again on September 16, after a correction to the office-action record, and moved by a few applications.

  5. Hypothesis 05

    Assignment is random, so a few high shares are noise.

    Ruled out: measured against the whole register

    Across 14,935word-and-class cells, the busiest examiner's median share is 2.4% and the 90th percentile is 4.2%. Noise does not produce 97% on 1,034 applications from 815 owners over 13 years. Assignment is also not independent, application by application. On 2024 filings, two unrelated filers with consecutive serial numbers, filed within a week of each other, drew the same examiner about one time in six. Chance would give one in seven hundred. The consecutive serial numbers run in blocks of about ten, so no block can gather 1,034 applications from 815 owners across thirteen filing years.

§ 8 · What this means for your prosecution

Know whether your mark has an address.

  1. If your mark carries one of these words in its class, the examiner is known before you file.

    So is the ground. A mark carrying TEQUILA in Class 33 has drawn a likelihood-of-confusion refusal 94 times in 100 since 2014, and the certification mark itself is cited on about one in four. Clearance starts from the assumption that the refusal is coming, and the application as filed answers it.

  2. If it does not, the draw is a draw.

    Across every class, the typical word's busiest examiner holds 2.4% of it. A handful of words have an address. The rest of the register does not, and the examiner you get is the subject of Research No. 01.

  3. Cannabis is no longer a desk. Brief the examiner you drew.

    The lawful-use refusal is still the one to expect on a cannabis mark, and the case law still moves. What changed in 2018 is who writes that refusal. The category spread across the corps. Read the profile of the examining attorney on your file, not a desk that dispersed that year.

Counting: for the capture figures, applications filed January 2014 or later that received an office action; the tequila year series in § 2 runs from 2005 on the same basis. One examiner per application, the one who wrote its first office action. Whole-word matching on the mark name, the Nice class named in every capture figure, and every cell gated at 25 owners, 5 filing years and no filer above 25% unless marked otherwise. The tequila corridor and owner counts were measured on the public record August 31, 2026. The year series was measured September 1; the whole-register sweep September 2, 2026; the Class 33 word distribution September 2 and again September 4. The cheese cell, the tequila examiner's share of Class 33 first actions, and the refusal and citation rates in § 5 were measured September 16, 2026. First published August 6, 2026; rewritten September 4, 2026, with the August cannabis figures withdrawn (see § 3). No examining attorney is named on this page.

Built from the public record. CrystalMark is not affiliated with, sponsored by, or endorsed by the United States Patent and Trademark Office or any of its employees.

The per-examiner version

Your examiner's docket is knowable before you respond.

Every active examining attorney has a profile: the grounds mix, the most-cited registrations, and what has overcome each. Those are the same reads this piece is built on.

Try free for 7 days