Fighting Back, Part 1: What the Cure-Rate Data Actually Says About Beating a Refusal

A refusal from the U.S. Patent and Trademark Office reads like a door slamming. It is not one — but it is heavier than the pep talks admit.

Pull the USPTO's own published prosecution-event records — the government's log of every action taken on an application — and follow a full month of filings from the refusal letter to their final fate. We did exactly that: every U.S. trademark application with a January 2022 filing date, 46,443 of them, setting aside international filings that arrive through the Madrid system, each one tracked to where it stands today. Of the applications that drew a refusal letter and have since finished, 46.9% registered anyway. And of the ones that took the examiner's harshest letter — a final refusal — 43.7% still walked away with a registration.

Roughly half. Not a death sentence, not a formality. A fight you win about as often as you lose, which means the response you file is where the outcome actually gets decided.

What we counted, and how you could count it too

The cohort is simple enough to rebuild. Take every U.S. application filed in January 2022, excluding Madrid Protocol filings (serial numbers beginning 79, which run on a different international track). Sort each file by its prosecution-history events — examiner's amendment, non-final action, final refusal — and follow it to its current outcome. "Registered" means a registration number issued, even if the registration later lapsed. "Died" means the file ended without one. About 4.5% of the cohort is still pending four and a half years later; those files are excluded from every rate below.

An office action, for the newcomer, is the examiner's written objection: a letter explaining why the mark can't register as filed. "Non-final" means you get a full response. "Final" means the examiner has dug in.

Here is the January 2022 picture, measured against the register on August 31, 2026:

• 25,439 applications — 54.8% of the cohort — drew at least one refusal letter. More than half. A refusal is not a freak event; it is closer to the default experience of filing.
• Of the refusal recipients that have finished, 46.9% registered (11,411 of 24,311).
• Files that never drew a refusal letter registered 73.7% of the time. That's the gap a clean application buys you.
• The luckiest group never got a letter at all: about 12% of the cohort was handled entirely by examiner's amendment — the examiner phones with a proposed fix, you agree, and no refusal is ever written. Those files registered just over 75% of the time.

No survey, no estimate, no internal guesswork. If you ran the same query against the public register — refusal events are logged under codes like "non-final action written/e-mailed" (CNRT/GNRT) and "final refusal written/e-mailed" (CNFR/GNFR) — you'd get the same numbers, give or take the tenths that move as the register updates daily.

The finding that should make you sit up: "final" barely changes the odds

The word final does a lot of intimidating work. In practice it's a procedural label, not a verdict — and the register proves it.

In the January 2022 cohort, 3,172 applications received a final refusal. Of the ones that have finished, 1,300 — 43.7% — still reached registration. Compare that to the 46.9% cure rate for refusals overall: getting the examiner's most emphatic "no" moved the odds by about three points. Thirteen hundred applicants took the letter that reads like the end of the road, responded anyway — with new arguments, amended goods, a request for reconsideration, or an appeal to the Trademark Trial and Appeal Board (the TTAB, the in-house tribunal that hears these disputes) — and won.

Applications whose worst event was a non-final action registered 47.4% of the time. A final refusal, in other words, is not a wall. It's a fork. Some applicants treat it as the end and walk; the ones who stay have very nearly the same odds they had before the word "final" appeared.

Here is one real file, so this is not abstract. CLEANOMIC — a cleaning-products brand — was filed January 12, 2022. A non-final action arrived February 1. The applicant responded March 15; the examiner answered with a final refusal on March 22. Five months later, on August 31, 2022, the examiner approved the mark for publication anyway, and it registered on August 1, 2023. The file that took the "final" letter in month two was a registration by the following summer.

Two ways to measure a "cure rate" — and why the honest answer is both

One caution before you take these numbers shopping. What we measured here is events to outcomes: a refusal letter of any kind, followed to the file's fate. It says nothing about which objection was inside the letter — and that is where the odds genuinely diverge.

Some office actions are housekeeping. The classic example is the identification requirement: the examiner telling you your description of goods or services is too broad, too vague, or in the wrong class (classes are the 45 categories the trademark system sorts everything into — software in one bucket, clothing in another, restaurant services in a third). Fixing that is often a matter of adopting the office's suggested wording; it's exactly the kind of problem examiner's amendments exist to resolve by phone. At the other end sits the likelihood-of-confusion refusal under Section 2(d) of the Trademark Act — the examiner has found a registered mark close enough to yours, on related goods, that consumers might confuse the two. The USPTO itself calls likelihood of confusion the most common reason for refusing registration, and it is the objection where the outcome turns hardest on the quality of the response. Measured by refusal ground rather than by event, the rates spread far apart — and confusion refusals sit at the bottom of that ladder. Part 1 of our Refusals 101 series walks through how examiners decide what goes in the letter.

What a founder should actually do with this

Three things follow from the numbers, and none of them is "panic."

First, budget for the response before you file. A refusal letter reached 54.8% of this cohort — the majority. The applicants who register are the ones who planned to answer, not the ones who treated the office action as a surprise expense.

Second, read the refusal to find out which kind it is before you react. A requirement about your description of goods is not the same event as a Section 2(d) confusion refusal, even though both arrive on the same letterhead. One is closer to paperwork; the other is closer to litigation — and your response, and your budget, should diverge accordingly. Our data analysis of what actually wins office actions breaks the strategies down by refusal type.

Third, don't read the word final as over. Thirteen hundred marks in one month's filings overcame a final refusal — at odds barely worse than the first letter carried. The people who abandon at that stage are, statistically, handing wins to the people who don't. Final refusals have also been getting rarer: they fell every month of Q2 2026 even as examiners issued more non-finals.

The through-line of the whole system is that its rules quietly decide who keeps their name — and the rule hiding inside this data is that persistence, applied intelligently to the right kind of refusal, is itself a strategy. For an early-career practitioner, the same numbers are a client-management tool: they let you tell a nervous founder, truthfully, that a refusal is a stage roughly half the applicants who see it through come out of with a registration — not the end of the road.

Part 2 in this series looks at how the winners fought — the specific moves that show up in the prosecution histories of marks that beat a Section 2(d) refusal. Preview: the best argument is often no argument at all.

This analysis is based on public USPTO records: every U.S. application with a January 2022 filing date (46,443 files, excluding Madrid Protocol serials beginning 79), classified by prosecution-history events and followed to its register outcome as of August 31, 2026, with still-pending files excluded from the rates. It is not legal advice. Want to see how a specific mark's prosecution actually unfolded before you respond? That's what a prosecution-history search surfaces in minutes.