CrystalMark Research · No. 02
The Count Rush
USPTO examiners do a quarter more work in the four months that close a fiscal quarter, and what they do with your application changes with it. A represented US filing that gets its first look in July opens with an office action 51.9% of the time. In June, 45.9%. Same corps, same kinds of filings, a different month on the production calendar.
Trademark examination runs on the USPTO's quarterly production clock. The swing holds under every composition control we applied. Neither the filing date nor the response date is a lever you can pull.
§ 1 · The sprint
Four peaks a year, every year, on the fiscal calendar.
USPTO examining attorneys work against production goals measured on the agency's fiscal calendar: the year ends September 30, and the quarters close with December, March, June, and September. Output follows the goalposts. In every year from 2022 through 2025, the four quarter-closing months together ran 21 to 32% above the other months of that same year. The drop on the far side is just as sharp. March 2023 produced 33,611 first actions; April 2023 produced 20,911. That is a 38% fall across a single month boundary, and 28% per working day. March, the mid-year checkpoint, is the tallest peak in every year of the window. December is the smallest of the four peaks. In 2024 and 2025 it barely rose above the months around it.
Deadline-shaped output is how production systems behave everywhere. What matters for the applicant is that the sprint doesn't just change how much gets done. It changes what gets done.
§ 2 · The mix
The close approves. The month after refuses.
For every application there is a first look: the moment the examiner either approves the mark for publication on sight or opens with an office action. The controlled series restricts to represented domestic filings, the composition control. Pro se domestic filings are read separately and tell the same story, and Madrid filings are excluded. Pooled across the window, the share of first looks that open with an office action is not flat across the year:
At the close · approval tilt
Quarter-end months approve more
Within represented domestic filings, the office-action rate at first look averages 1.4 points lower in quarter-end months than in the rest of the year (2.2 points lower for pro se filings). A first-action approval is the fastest count there is: nothing to write. When the clock is short, the approvable files get pulled first.
After the close · fewer approvals
July opens on the leftovers
June clears the shelf: represented-domestic first-action approvals run 120,438 in the pooled Junes and 60,840 in the pooled Julys, half the June figure. First office actions fall less, from 102,282 to 65,531. What remains in July is the work that couldn't be approved on sight. October, after the fiscal-year close, shows the same signature at 49.7%.
§ 3 · The Final that waits for the month after
At every quarter close, fewer resolved refusals go Final. The month after, more do.
The same calendar shows up on applications already carrying a refusal. Here the file itself does not change from one month to the next. What the calendar moves is which pending refusals the examiner resolves in a given month, and how. An examiner resolves a pending refusal by issuing a Final or by approving the application after the office action. The split between the two depends on the month. In the four closing months, pooled across the window, 17.4 to 18.5% of those resolutions were a Final refusal. In the other eight months, 19.4 to 21.2%. At the fiscal-year end: of refusals resolved in the pooled Septembers, 17.5% went Final; in the pooled Octobers, 20.2%. October has issued at least a third more Finals than September in every year of the window. In 2024 the corps issued 3,471 Finals in September and 5,501 in October. A Final at the close creates future work instead of booking a disposal, so it waits.
§ 4 · Five ways this could be nothing
Four ruled out. One partly true, and priced in.
Hypothesis 01
“It's the applicant mix, not the examiner.”
Ruled out
The swing holds inside narrow strata. Within represented domestic filings alone, July runs 6 points above June. Within pro se domestic filings, 5 points. Madrid §66(a) filings are excluded from the controlled series entirely. They essentially always open with an office action.
Hypothesis 02
“It's calendar mechanics: long months, short months.”
Ruled out
Working days vary by about 5% from one month to the next. The shortest month and the longest differ by about a fifth. Per working day, the four closing months still run about 24% above the rest of the year. Per working day, March out-produces April by about a third and February by about a sixth.
Hypothesis 03
“It's the pandemic backlog era.”
Ruled out
The window starts in 2022. The backlog signature is first actions landing more than two years after filing. Excluding it moves the monthly rates by less than half a point. The pattern repeats in 2025 and the first half of 2026, and faintly in 2024. July sits above June in all four complete years.
Hypothesis 04
“It's a data artifact.”
Ruled out
Every date here is the USPTO's own: action dates are the examiner-signed dates in the agency's daily XML, and approvals are the date of the examiner's approved-for-publication event. Nothing is timestamped by our pipeline.
Hypothesis 05
“Examiners pick different files, not different answers.”
Partly true, and it doesn't change the read
Some of the swing is selection. An examiner working to the count at the close reaches for the applications that can be approved without writing anything. Your mark's merits don't change. What changes is when the application gets picked up. One thing bounds how far selection goes: the swing holds inside narrow, homogeneous strata. And the selection is the examiner's, not the filer's. On pending refusals (§ 3) the file itself does not change from one month to the next. The calendar moves which pending files get resolved, and whether a Final or an approval resolves them. We also tested the one date a filer controls. By the month a response was filed, the share of resolved refusals that went Final is flat, 45.2% to 46.4%, in every month of the year. Either way, the practical effect on your docket is the same.
§ 5 · What this means for your prosecution
Two non-levers and one reading rule.
Don't re-time your filings over this.
The queue from filing to first look runs about six months, and it varies. You cannot reliably aim a filing date at a friendly examination month. If the filing calendar matters to your client for other reasons, follow those.
Don't re-time your response either.
We tested the lever this brief once offered. Across first responses to a refusal filed from 2022 to mid-2026, the share resolved by a Final refusal is the same in every filing month, 45.2% to 46.4%. Examiners resolve more pending files at a quarter close, and more of those resolutions are approvals, but which files they pick is not something a filing date buys. File the response when it is ready.
Read an early-quarter office action in context.
A first action dated July or October is the seasonal norm, not necessarily a signal about your examiner. Whether the examiner runs tough is a per-examiner question, and so is what overcomes them. That's what the profile is for.
Counting: window January 2022 through June 2026, roughly 2.8M first looks. A first look is the earlier of the round-1 non-final action date and the approved-for-publication date. Represented domestic means an attorney bar number or firm name on file, a filing basis other than §66(a), and not a 79-series serial number. First actions and Finals are counted on the examiner-signed action dates in the USPTO's daily XML; first-look approvals on the approved-for-publication event date. Figures frozen at publication, July 14, 2026. A spot revalidation against the production corpus August 21, 2026 found Final and first-action month counts within ±4. The Final-refusal section was re-measured and corrected September 4, 2026. September is not the year's low for Finals in every year. Every quarter close carries a lower Final share than the months around it, and October exceeds September every year. The first-look series was re-measured September 16, 2026 on the definition printed here, and the response-month test in § 4 and § 5 was run the same day on 331,985 first responses to a refusal.
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
The calendar is corps-wide. Your examiner isn't.
Seasonality moves every examiner a few points. The examiner assigned to your application moves the odds far more: their refusal rate, what they refuse on, how often a refusal sticks once contested, and what has overcome them.
Try free for 7 days