The word "final" is doing a lot of unearned work. It does not mean the application has been refused, and it does not mean the examiner's view has been reviewed by anyone. It means prosecution is closed: the applicant may no longer file amendments as of right, and the examiner is no longer obliged to consider new arguments in the ordinary way.
What follows is a genuine fork with four exits, and the wrong one is chosen surprisingly often — usually out of exhaustion after two years of correspondence, and usually in the direction of surrendering claim scope that the invention needed.
Fork · The Four Exits
What Closing Prosecution Actually Forecloses
Once an action is designated final, the applicant has a fixed window — commonly two months, extendable in some systems to six on escalating fees — in which to take one of four routes.
- Abandon. Sometimes correct. If the surviving claim scope is too narrow to affect a competitor, spending another three thousand on the file is worse than losing the filing fees already spent.
- Amend within the closed record. A limited amendment can still be entered after finality, but only if it places the application in condition for allowance or cancels claims. The examiner has discretion to refuse to enter it, and frequently does.
- Buy another round. Most systems sell a continued examination of some form: a fee, a fresh amendment, and the same examiner reopens the file. It is faster and cheaper than appealing, and it is a poor idea when the disagreement is about law rather than wording, because the same examiner will apply the same reasoning again.
- Appeal. The rejection is reviewed by a panel — usually three technically qualified members of an appeal tribunal sitting above the examining corps — who decide whether the rejection was correct on the record as it stands.
The fourth option has a property the others lack: it puts the question to someone whose job is not to maintain the office's earlier position. That is the entire value proposition, and it is why a patent appeal is worth understanding properly rather than treating as a last resort.
Procedure · The Paper Trail
Notice, Brief, Answer, Hearing
The sequence is rigid and the deadlines are the kind that end applications. A notice of appeal is filed first, within the response window running from the final action, accompanied by its own fee. That notice buys time but decides nothing; the substance arrives with the brief, typically due two months after the notice, again extendable on fees in most systems.
The brief is the whole case. It sets out the claims on appeal, the grounds of rejection being contested, and an argument addressing each ground separately — and it is confined to the record already built. New evidence is generally inadmissible, new amendments are not entertained, and an argument never made to the examiner may be treated as waived. Anything an applicant wishes the tribunal to see must already have been put in front of the examiner during prosecution, which is why the quality of the written responses filed two years earlier turns out to matter enormously.
The examiner then replies with a formal answer, and this is where a substantial proportion of appeals quietly end. Faced with drafting a defence for review by a panel, examiners reopen prosecution or withdraw the rejection in a meaningful share of cases — figures in the region of a third are commonly reported where offices publish them, and some systems add a short pre-appeal review specifically to catch these before the file leaves the examining group at all. An applicant whose objection rests on a plain misreading of a cited document often never reaches the tribunal.
If the file does go up, an oral hearing is optional, carries its own fee, and usually runs about twenty minutes per side before a panel that has already read everything. It is worth requesting when a single technical misunderstanding drives the whole rejection and worth skipping when the dispute is documentary.
The brief cannot introduce new evidence. Everything the tribunal will ever see was decided by how carefully you answered the examiner two years ago.
Why prosecution records matter laterCost · Fees, Hours and Years
The Three Prices, Only One of Which Is Published
Official fees are the smallest and most visible component: a notice fee plus a forwarding or brief fee, together landing in the low thousands for a standard-size applicant in most large systems, with substantial reductions — often half, sometimes more — for individual inventors and small entities where such discounts exist.
Professional time is the real expense. A properly constructed brief is twenty to forty hours of work: reading every cited reference in full, mapping each claim feature against each rejection, and building an argument that a panel can adopt without doing its own research. At ordinary specialist rates that is frequently the largest single invoice of the entire application, larger than the drafting fee. Briefs written in a hurry lose on grounds the applicant would have won on, which makes the cheap version of this the expensive one.
The third price is time. From notice of appeal to decision is commonly twelve to twenty-four months where tribunals are keeping pace, and materially longer where they are not; total elapsed time of two to three years from the final rejection to a decision is unremarkable. That is drawn from a term measured from the original filing date, so an appeal in year four of a twenty-year term is spending an eighth of the right's life to establish what the right covers. Some systems restore part of that period as term adjustment when the applicant prevails, which softens the loss without eliminating it, and the same acceleration and delay trade-offs that govern how long an application waits before it is read apply to the review stage as well.
Odds · What the Statistics Say
Reversal Rates, and How to Read Them Honestly
Where outcomes are published, three results are recorded: the rejection is affirmed, affirmed in part, or reversed. Full reversal typically runs somewhere between a quarter and a third of decided appeals; partial reversal accounts for another tenth or so; the remainder are affirmed. Read cold, that looks like poor odds.
Read properly, it is better than it appears, for two reasons. The first is selection: the appeals that reach a decision are the residue left after the winnable ones were conceded by examiners at the answer stage. Counting those withdrawals as wins, which they are, moves the practical success rate for a well-chosen appeal considerably above the headline reversal figure. The second is that a partial reversal is often a complete commercial victory — one independent claim restored to its filed scope can be the whole point of the exercise, whatever happened to the eleven dependents.
Two further asymmetries favour the applicant. A reversal is not a grant, but it removes the ground of rejection and returns the file to examination with the tribunal's reasoning attached, which is difficult for the same examiner to work around. And a rejection overcome by argument leaves a cleaner record than one overcome by amendment: nothing was surrendered, so nothing narrows the claim in a later dispute. That distinction is invisible at the time and decisive years later.
Winning by argument costs money. Winning by amendment costs scope. Only one of those two bills arrives after the patent is granted.
The comparison that settles most casesJudgement · When It Pays
The Cases That Justify It, and the Cases That Do Not
The decision resolves into a single comparison: the cost of the appeal against the value of the scope that would be surrendered by amending instead. Everything else is detail.
Appealing is usually right where the disagreement is legal rather than factual — the examiner has combined references with no articulated reason to combine them, has read a feature into a document that is not there, or has applied a claim-construction position the specification contradicts. It is right where the contested claim is the one a competitor would have to infringe, and where the available amendment would move the claim to a feature that any rival could design around in a fortnight. It is also right, less obviously, where the field is crowded enough that a narrow patent is worth close to nothing: in those cases there is no cheap fallback, so the choice is a patent appeal or nothing.
It is usually wrong where a narrowing amendment costs little commercially, where the objection rests on a reference the applicant has not read in full, where the budget cannot absorb both the brief and the possibility of losing, or where a fresh continued examination with genuinely different claims would reach a better right sooner. It is always wrong as an expression of annoyance. Panels decide on the record, not on how unreasonable the correspondence felt.
Where a decision goes against the applicant, most systems provide a further route into the courts, on a longer timescale and a much larger budget — and at that point the question has stopped being about the patent and started being about the business. Long-running inventor programmes tend to be candid about this. Accounts of four decades of work supporting independent inventors, of inventor-led design in architectural innovation, and of the MixAid product's route from concept to market all describe the same discipline: persistence spent on the claims that matter and abandonment accepted quickly on the ones that do not. Even the cultural fascination surveyed in this piece on how popular culture romanticises invention tends to skip this stage entirely, which is roughly where it stops being useful.
Two related decisions belong in the same conversation. If the objection is that the application covers more than one distinct invention, appealing is the wrong instrument and dividing is the right one. And the whole sequence — first action, amendment, finality, review — is easier to judge with the shape of the process that follows a filing in view rather than one letter at a time.
Final means final for that examiner. It does not mean final for the application.