Patent portfolio pruning is the decision to stop paying for patents that are no longer worth their cost. It sounds like failure and it is not — only 41.4% of US utility patents reach full term, so releasing the majority is simply what the market does.
The failure is doing it by accident. A portfolio that shrinks into abandoned patents because deadlines were missed has lost the same patents as one that shrinks by decision, except nobody chose which.
Why pruning is normal
Computed from every maintenance fee event the USPTO has recorded — 27,273,654 rows covering 8,262,336 US utility patents granted since 1981.
| Stage | Portfolio remaining | Released at this stage | Fee |
|---|---|---|---|
| At grant | 100% | — | — |
| After first fee | 85.8% | 14.2% | $2,150 |
| After second fee | 64.6% | 21.2 points | $4,040 |
| After third fee | 41.4% | 23.2 points | $8,280 |
Grant cohort 2014, large entity fee amounts.
Nearly six patents in ten are released before full term, and the proportion is rising — patents granted in 2000 reached full term 51.1% of the time.
The economics explain it. A hundred-patent US portfolio maintained to full term costs roughly $1.4 million in government fees for a large entity, before foreign annuities or service charges. Most patents in most portfolios do not generate returns approaching their share of that.
Full figures by grant year are on the Ipiry Patent Survival Curve.
When the decision actually happens
At each maintenance fee window, and nowhere else. This is the useful thing about the US structure — the fee schedule imposes a review cadence whether or not anyone plans one.
| Fee | Timing | Typical question at this point |
|---|---|---|
| First | 3.5 years after grant | Is the product still being built? |
| Second | 7.5 years after grant | Did the technology find a market? |
| Third | 11.5 years after grant | Is four more years worth $8,280? |
Each window has a different character. At the first fee the technology is usually still current and the question is whether the business went the direction that patent assumed. At the third, the question is nearly always whether a four-year tail justifies the largest single payment in the schedule.
Review twelve months ahead, not at the deadline. A review conducted at the window leaves no time to sell what you decide to release, which forfeits whatever those patents were worth to someone else.
The four questions
For each patent facing a fee, four questions decide it.
Is anyone practising the claims?
The single most useful discriminator. A patent that reads on a shipping product has enforcement value. One that does not has option value, and option value decays as term runs down.
Establishing this properly means examining what competitors actually ship — product literature, technical specifications, their own patent filings — and comparing that against your claims element by element. Not a general impression that the field is active.
A patent with four years left and no identified infringer is worth considerably less than the same patent with twelve.
Is the technology still current?
Eleven years is a long time in most fields. A patent arriving at its third fee covering an approach the industry has moved past is usually a release, regardless of how good the claims are.
The test is whether anyone would build this today, not whether it was clever when filed.
Does it block anyone?
Defensive value is real but frequently overstated. A patent that genuinely constrains a competitor's roadmap is worth keeping even with no revenue attached. One that could theoretically be asserted against someone, someday, is the category most portfolios are stuffed with.
The question is specific: whose product would this stop, and are they shipping it?
What would it sell for?
The question most owners never ask. A patent worth less than its fee to you may still be worth something to somebody else — an operating company in the space, a defensive aggregator whose members are exposed, an assertion entity that can find the infringers you could not.
Both lapse and sale end the fee obligation. Only one pays you.
A worked portfolio review
A company with twelve US patents facing fees across the coming year.
| Patent | Fee due | Amount | Practised? | Technology | Decision |
|---|---|---|---|---|---|
| A | 1st | $2,150 | Yes, own product | Current | Keep |
| B | 1st | $2,150 | No | Current | Keep — early, option value |
| C | 2nd | $4,040 | Yes, competitor | Current | Keep — enforcement value |
| D | 2nd | $4,040 | No | Superseded | Sell or release |
| E | 2nd | $4,040 | No | Superseded | Sell or release |
| F | 3rd | $8,280 | Yes, own product | Current | Keep |
| G | 3rd | $8,280 | No | Superseded | Release |
| H | 3rd | $8,280 | No | Superseded | Release |
Total if everything is renewed: $41,230. After the review: $16,620 on four patents, with two more offered for sale before their windows close.
The saving is $24,610 in fees. The upside is whatever D and E fetch, which may be nothing and may be considerably more than the fees saved.
Note patent B. No infringer identified, but it is only at the first fee, the technology is live, and $2,150 buys four more years to find out. Option value is worth most early and least late, which is why the same reasoning that keeps B would release it at the third fee.
Selling rather than lapsing
This is the part most pruning exercises skip, and it is where the money is.
| Let it lapse | Sell before the window | |
|---|---|---|
| Fee obligation | Ends | Ends |
| Cash | $0 | Sale proceeds |
| Time required | None | Weeks to months |
| Needs a valuation | No | Yes |
The constraint is timing. A brokered sale takes six to eighteen months; a fee window is six. A review conducted twelve months before the deadline can run a sale process. One conducted at the deadline cannot.
Which patents are worth offering? Generally those where the technology is still live but the owner has moved on — patent D and E in the example above. Patents covering genuinely obsolete approaches usually have no buyer either, and the honest answer is to release them.
Bundle where you can. A single patent is a hard sale; a set of five covering one technology area is a portfolio, and portfolios attract buyers that individual patents do not.
Pruning a foreign family
Patents are national rights, so a family can be maintained in some jurisdictions and released in others. This is frequently the largest saving available and the one most often overlooked.
The questions are different from the US ones:
- Do you sell in this market, or plan to?
- Does a competitor manufacture here?
- Is this jurisdiction a realistic enforcement venue?
- What does the annuity cost, and how steeply does it escalate?
European annuities escalate sharply in later years, and a family validated in several European states carries a separate fee in each. Reducing validation to the states where you actually sell is standard practice.
Keep at least one jurisdiction where a competitor manufactures, even if you do not sell there. A patent in a manufacturing country can be worth more than one in a sales country.
What not to prune
Patents covering the current product. Obvious, and still done — usually when a cost review is conducted by someone without visibility of what the patents cover.
Patents in a live family. Releasing a parent while continuations are pending can create prior art problems and complicate the family's structure.
Patents ahead of a funding round or acquisition. Portfolio size and coverage are examined at diligence, and a portfolio pruned aggressively in the preceding year invites questions about why. Prune after, not before.
Patents with unresolved assertion potential. If a patent is genuinely being considered for enforcement, letting it lapse mid-consideration forecloses the option permanently.
Anything nobody has actually looked at. The most common pruning error is releasing a patent that nobody reviewed because it was low-numbered on a spreadsheet.
Running the review
Start from the fee calendar, not the portfolio list. The patents facing fees in the next twelve months are the ones that need decisions; the rest can wait.
Get technical input. Whether a patent reads on a current product is an engineering question, not a legal one, and portfolio reviews conducted without technical input systematically over-value old patents and under-value ones whose claims turned out to cover something unexpected.
Value the borderline cases. For patents where the decision is genuinely unclear, a patent valuation costs less than the fee and answers the question.
Record the reasoning. A note explaining why each patent was released takes five minutes and saves an argument two years later when circumstances change and somebody asks why the portfolio does not cover something.
Decide in one sitting. Portfolio reviews that drift produce defaults, and the default is renewal — which is how portfolios accumulate patents nobody can justify.
Before your next fee window
- Build the fee calendar for the next twelve months across the whole portfolio, US and foreign.
- Get technical input on which patents read on current products, yours or anyone else's.
- Identify the release candidates and check whether any have a plausible buyer.
- Offer those for sale now, not at the deadline, because a sale takes months.
- Value the borderline cases rather than guessing.
- Review the foreign family separately, country by country.
- Record the reasoning for every release.
- Confirm the payments posted on everything you keep.