Trademark Examiners Cleared a Backlog in March: Non-Final Actions Jumped 57% to End Q1 2026

Quarterly USPTO Prosecution Report | USPTO data through March 31, 2026

The most interesting number in the first quarter is not the 92,404 non-final office actions the USPTO issued — it's how lumpily they arrived. Examiners issued fewer non-final actions in February than in January, then put out 39,032 in March, a 56.8% jump in a single month. A quarter that looked sleepy at its midpoint finished with the heaviest examination month of the three. If you manage a docket, March did not feel like an average month, and the quarterly total hides that.

A quarter that dips in February and then issues four out of every ten of its non-final actions in March isn't steady-state examination — it's a pipeline catching up.

— Howard Katzenberg, Founder, GleanMark

We think the February-to-March swing most likely reflects examination capacity and queue timing rather than a surge in new applications reaching examiners — office-action output tends to move in batches as examiners clear assigned dockets, not in smooth monthly increments. That's a hypothesis, not a finding; the volume data alone can't separate staffing effects from filing-cohort timing.

Executive Summary

• March carried the quarter. Non-final actions ran 28,478 in January, dipped to 24,894 in February, then jumped to 39,032 in March — 42.2% of the quarter's non-final volume in one month.
• Section 2(d) is essentially the whole refusal story. In our analyzed Q1 response subset, likelihood of confusion was the primary refusal in 6,143 of 6,277 responses — 97.9%. Everything else is a rounding error by comparison.
• Descriptiveness arguments punch above their weight. Section 2(e) refusal-type arguments posted the highest directional effectiveness score, 0.951, but on only 277 instances. Section 2(d) scored 0.814 across 4,421 instances — a lower score on a vastly larger base, which is the number that should actually shape your strategy.
• Specimen Support and Supplemental Register lead the argument-category table. Specimen Support scored 0.956 (55 instances); Supplemental Register scored 0.936 across 709 instances — the rare combination of a high score and a large sample.
• This is a subset, not the whole register. The refusal and argument analysis covers 6,659 analyzed office-action records and 6,277 analyzed responses against 92,404 total USPTO non-final actions. Read the percentages as describing the analyzed subset, not the entire USPTO docket.¹

1. Prosecution Volume

Period	Non-Final Actions	Final Actions	Non-Final and Final Actions	Non-Final MoM Change
January 2026	28,478	3,595	32,073	—
February 2026	24,894	2,764	27,658	-12.6%
March 2026	39,032	3,307	42,339	+56.8%
Q1 2026 Total	92,404	9,666	102,070	—

The headline is 92,404 non-final actions and 9,666 final actions — together 102,070 office-action documents. Across all recorded prosecution document types, the quarter generated 650,319 documents.

The shape matters more than the total. February's non-final output fell 12.6% from January; March then rose 56.8%. That isn't a gentle recovery — it's a batch. Final actions traced the same down-then-up path (3,595, then 2,764, then 3,307) but on a smaller scale, and March's final-action count never recovered to January's level. The non-final surge did all the work of lifting the quarter.

2. Office Action Document Types

Document Code	Document Type	Documents in Q1 2026
NFIN	Non-Final Action	92,404
XSS	XSearch Search Summary	71,546
NFO	Notice of Non-Final Office Action	64,085
AMC	Amendment and Mail Process Complete	59,153
ROA	Response to Office Action	55,914
ABN	Abandonment — Notice	28,650
PB3	OG Publication Confirmation	27,073
PB1	Notice of Publication	26,399
EB4	Notification of Notice of Publication	26,399
PNR	Public Note	24,792
PST	TRAM Snapshot of App at Pub for Opposition	24,773
EDS	Notice of Design Search Code Emailed	14,125
ORC	Registration Certificate	13,627
ERO	Notice of Registration Email	13,452
SPE	Specimen	11,269
FREF	Final Action	9,666

One trap in this table is worth flagging because people fall into it: NFIN (the action itself, 92,404) and NFO (the notice of that action, 64,085) are distinct document codes for related events. Adding them together double-counts. The headline office-action figure is 92,404, full stop. The long tail of XSearch summaries, publication confirmations, and registration emails is the administrative exhaust of examination, useful for reconciling counts but not a measure of legal activity.

3. Refusal Landscape

Primary Refusal Type	Cases	Share of Analyzed Responses	Registered Cases
Section 2(d) — Likelihood of Confusion	6,143	97.9%	3,380
Section 2(e) — Descriptiveness / Surname	126	2.0%	8
Specimen Refusal	3	0.05%	2
Identification of Goods/Services	3	0.05%	2
Other	2	0.03%	0

Refusal Type Appearing Anywhere in Analyzed Case	Cases with Tag	Registered Cases with Tag	Avg. Days to Registration
Section 2(d) — Likelihood of Confusion	6,143	3,380	121
Section 2(e) — Descriptiveness / Surname	146	12	86
Disclaimer Requirement	46	7	—
Identification of Goods/Services	26	3	—
Specimen Refusal	15	3	—
Other	3	0	—
Ornamental Refusal	1	0	—

The primary-refusal view counts each response once, and on that basis there is exactly one story: Section 2(d) is the primary refusal 97.9% of the time. Descriptiveness and surname refusals are a distant second at 126 cases. Everything else is in the single digits.

The overlapping-tag view is where the nuance lives, and it's worth reading the two tables together rather than averaging them. Section 2(e) appears in 146 tagged cases versus 126 as a primary refusal — the gap is the set of cases where descriptiveness rode along as a secondary issue behind a 2(d) refusal. The tag table is your issue-spotting tool; the primary table is your share-of-response tool. They answer different questions, so don't merge them.

One number to sit with: Section 2(d) cases that registered did so in an average of 121 days, against 86 days for Section 2(e). Likelihood-of-confusion fights take longer to win, which tracks — they usually require argument and evidence rather than a curable amendment.

4. Argument Effectiveness

Refusal Type	Directional Score (0–1)	Argument Instances	Verified-Win Instances
Section 2(e) — Descriptiveness / Surname	0.951	277	16
Identification of Goods/Services	0.890	2	0
Section 2(d) — Likelihood of Confusion	0.814	4,421	1,443
Specimen Refusal	0.683	3	3
Disclaimer Requirement	0.950	1	1
Other	0.900	1	0

Argument Category	Directional Score (0–1)	Outcome Multiplier	Argument Instances
Specimen Support	0.956	1.189	55
Design Distinctiveness	0.940	1.200	4
Supplemental Register	0.936	1.161	709
Disclaimer	0.927	1.182	79
Class Deletion	0.903	1.102	313
Consent Agreement	0.903	1.48	113
Suggestive Not Descriptive	0.866	1.200	25
Incongruity	0.860	1.200	2
Commercial Impression	0.822	1.84	409
Cited Mark Abandoned	0.821	1.21	164
Goods Amendment	0.812	1.86	765
Third Party Registrations	0.785	1.32	244
Weakness Of Cited Mark	0.783	1.46	335
Goods Coexistence	0.766	1.32	450
Market Channel Distinction	0.763	1.21	219

The directional score is a 0–1 analytical measure, not a win rate and not a percentage. Treat it as a ranking signal, and always read it next to the instance count.

Here is the discipline the table demands. Section 2(e) tops the refusal-type list at 0.951 — but on 277 instances and 16 verified wins. Section 2(d) scores 0.814 on 4,421 instances and 1,443 verified wins. The higher score is the better headline; the lower score is the better evidence base, and it's the one that governs almost every refusal a practitioner will actually face this quarter.

In the argument-category table, the standout is Supplemental Register: a 0.936 score across 709 instances is the rare case where a high score and a deep sample coincide, which is exactly the combination you want before you bet a response on a tactic. Specimen Support edges it on score (0.956) but on 55 instances. Be skeptical of Design Distinctiveness (four instances) and Incongruity (two) no matter how good the number looks — a 0.94 on four cases tells you almost nothing.

Note the lane separation: Supplemental Register, Disclaimer, and Suggestive-Not-Descriptive arguments address descriptiveness and surname refusals. They do nothing for likelihood of confusion. For the 2(d) refusal that dominates this dataset, the relevant tools are consent agreements, cited-mark abandonment, goods amendments and coexistence, third-party registration evidence, weakness of the cited mark, and market-channel distinctions — most of which cluster in the 0.76–0.82 range. There is no high-scoring silver bullet for 2(d), which is itself the point.

Practitioner Takeaways

1. Build for a back-loaded quarter. Output dipped in February and spiked in March. Quarterly totals smooth that over; your response deadlines won't.
2. Assume Section 2(d) until proven otherwise. It's the primary refusal in 97.9% of analyzed responses. Front-load clearance and cited-registration review accordingly.
3. Match the tool to the refusal, and don't cross lanes. Supplemental Register and disclaimer arguments cure descriptiveness, not confusion. Consent and coexistence belong to 2(d).
4. Weigh the sample, not just the score. A 0.94 on four instances is noise; a 0.936 on 709 is a plan. Supplemental Register and Goods Amendment carry both score and depth.
5. Budget time for 2(d). Confusion refusals that registered took 121 days on average versus 86 for descriptiveness — they are won with argument, not amendment.

Explore more USPTO data analysis on the GleanMark Insights blog.

Footnotes

1. Coverage and methodology, stated once: the refusal and argument analysis covers an analyzed subset of 6,659 office-action records and 6,277 responses, against 92,404 total USPTO non-final actions in the quarter. NICE class references describe classes within applications, not application counts. Share-of-response percentages describe the analyzed subset only and should not be read as total USPTO refusal rates. Data sourced from USPTO TSDR and public prosecution records, reflecting activity recorded through the report date. Informational only; not legal advice. ↩