Patentable means an invention meets every legal condition for a patent to be granted and to survive challenge.
Seven questions answer it, and they are worth asking in a specific order — cheapest and most decisive first.
Most inventions fail on question three, which is also the one you can answer yourself in an afternoon for nothing.
Work down the list and stop at the first failure. There is no point assessing enablement on something already described in a 2015 patent.
Why order matters
Running the questions out of order wastes money and occasionally destroys rights.
| Doing this first | Costs you |
|---|---|
| Drafting claims | Effort on something already described |
| Commissioning a professional search | Money the free search would have saved |
| Showing it to a manufacturer | Foreign rights, if unfiled |
| Building a prototype | Time, if the concept is anticipated |
| Assessing commercial value | Analysis of something unpatentable |
| The free prior art search | Nothing — and it answers the most likely failure |
The disclosure question is the only one with no remedy, which is an argument for checking it before any conversation with a third party rather than at question six.
Everything else can be reordered without loss. Only the search-first rule and the disclose-nothing rule are strict.
The test in order
| # | Question | Cost to answer | Failure rate |
|---|---|---|---|
| 1 | Is it in a statutory category? | Minutes | Very low |
| 2 | Is a judicial exception in play? | Hours | High for software and diagnostics |
| 3 | Has anything disclosed it already? | Free — an afternoon | Highest |
| 4 | Would it have been obvious? | Harder | High |
| 5 | Can you describe it well enough to build? | Your own judgement | Moderate |
| 6 | Have you disclosed it publicly? | Minutes | Fatal if outside the window |
| 7 | Is it worth the cost? | An hour | The one most often skipped |
Question three first if you want one answer. It eliminates more inventions than everything else combined and costs only time.
Question six is the one that cannot be fixed. Everything else has a remedy.
1. Is it in a statutory category?
| Category | Covers |
|---|---|
| Process | A series of steps |
| Machine | A device with interacting parts |
| Article of manufacture | A made thing |
| Composition of matter | Chemical combinations |
Almost everything physical passes. Improvements count too — a new and useful improvement to an existing machine is squarely within §101.
If it is a physical thing or a series of steps, move on. This question rarely stops anything.
2. Is a judicial exception in play?
| Exception | Fields most affected |
|---|---|
| Abstract ideas | Software, business methods, fintech |
| Laws of nature | Diagnostics, personalised medicine |
| Natural phenomena | Biotechnology, isolated compounds |
Only assess this seriously if you are in one of those fields. For mechanical and most chemical inventions it is not the battleground.
The drafting test that predicts the outcome: does the claim recite what is achieved or how it is achieved?
| Claim shape | Prospects |
|---|---|
| "A system that predicts availability" | Poor — recites a result |
| "Partitioning the dataset by time window, computing X using [specified technique]" | Better — recites a mechanism |
If your description is a result, rewrite it as a mechanism before going further. See what can be patented.
3. Has anything disclosed it already?
The decisive question, and free to answer.
| Search | Where |
|---|---|
| Full text, worldwide | Google Patents |
| Authoritative US records | USPTO Patent Public Search |
| Families and foreign equivalents | Espacenet |
| Products on sale | Ordinary shopping and trade searches |
| Non-patent literature | Journals, conferences, manuals |
Search by function, in five different vocabularies. You call it a flow stabiliser; the reference calls it a damping baffle. Write down five ways a stranger would describe what it does and search all of them.
A product on sale is prior art whether or not anyone patented it, so market searching matters as much as database searching.
Prior art is anything public, anywhere, in any language, whether or not anyone read it. See prior art definition.
4. Would it have been obvious?
Harder, because hindsight makes everything look obvious once you know the answer.
| Ask | Suggests non-obvious |
|---|---|
| Did others try and fail? | Yes |
| Was there a long-felt need? | Yes |
| Were the results unexpected? | Yes |
| Did the combination require a reason to combine? | Yes |
| Is it just two known things put together? | No — likely obvious |
Document your evidence now. Failed competitor attempts, unexpected results and long-felt need are §103 evidence, and they are far easier to record today than to reconstruct in year six.
A nexus is required. The evidence must tie to the claimed feature, not to marketing or price.
5. Can you describe it well enough to build?
| Test | Standard |
|---|---|
| Could an engineer in the field build it from your description? | Enablement |
| Without solving problems you have not addressed? | Without undue experimentation |
| Do you describe alternatives, not just your version? | Supports broader claims |
| Is every claim term defined or clear? | Definiteness |
No prototype is needed. A sufficiently detailed written description counts as constructive reduction to practice, and many patents issue on things never built.
Describing alternatives is where breadth comes from. New matter cannot be added after filing, so a variation not written on day one is permanently unavailable.
6. Have you disclosed it publicly?
The question that cannot be un-answered.
| What you did | US position | Rest of world |
|---|---|---|
| Nothing public | Clear | Clear |
| Told someone under an NDA | Clear | Clear |
| Conference talk, demo, crowdfunding page | 12-month clock running | Rights likely lost |
| Offered for sale, even confidentially | 12-month clock running | Rights likely lost |
| Public more than 12 months ago | US rights gone too | Gone |
The grace period is personal and US-only. It removes your own disclosure from the art against your own application, and almost nowhere else provides an equivalent.
A confidential sale still counts. Following Helsinn, a commercial offer for sale triggers the bar even where the invention details stayed secret.
If anything is public, file now and assume foreign rights are gone.
The alternatives if it is not patentable
| Route | Gives you |
|---|---|
| Trade secret | No expiry, no disclosure — while it stays secret |
| Defensive publication | Stops anyone else patenting it, very cheaply |
| Trademark on the name | Indefinite, and separate from the invention |
| Copyright in software or content | Automatic, long term |
| Design patent on appearance | Cheaper and faster than utility |
| First-mover execution | No legal protection, real commercial value |
Trade secret is the genuine alternative for process inventions. If a competitor could not tell from the product that you use the technique, secrecy can outlast any patent.
Filing forecloses secrecy permanently, because publication at eighteen months destroys it whether or not the patent grants. That is the one IP decision that cannot be revisited. See can you patent something and make it free.
7. Is it worth the cost?
Patentability is a legal test with no commercial component, which is why this question is separate and last.
| Cost | Amount |
|---|---|
| Provisional application | ~$60–$300 in fees, plus preparation |
| Non-provisional to grant | $8,000–$20,000 |
| Maintenance fees, large entity | $14,470 |
| Maintenance fees, micro entity | $2,894 |
| Ipiry Patent Survival Curve v1.0 | Rate |
|---|---|
| Survive the 3.5-year fee (2022 cohort) | 85.8% |
| Survive the 7.5-year fee (2018 cohort) | 64.6% |
| Reach full term (2014 cohort) | 41.4% |
| Abandoned before full term | 58.6% |
Computed from 27,273,654 USPTO maintenance fee records covering 8,262,336 US utility patents — see the patent survival curve.
Every one of those abandoned patents passed the patentability test. Nearly three in five were later judged not worth the fee, which is what the gap between patentable and worthwhile looks like at scale.
Ask whether a competitor would practise the invention. That question predicts value better than any assessment of cleverness.
The four statutory conditions behind the test
The seven questions map onto four conditions in the statute.
| Question | Condition | Statute |
|---|---|---|
| 1, 2 | Eligible subject matter | §101 |
| 3 | Novelty | §102 |
| 4 | Non-obviousness | §103 |
| 5 | Adequate disclosure | §112 |
| 6 | Timing of your own disclosure | §102(b) grace period |
| 7 | — | Not a legal condition at all |
All four legal conditions must hold simultaneously. Strength in three does not compensate for failure in the fourth.
Question seven is not in the statute. Commercial worth has no bearing on patentability, which is exactly why it has to be asked separately. See the meaning of patentable.
What to do at each failure
| Failed at | Remedy |
|---|---|
| 1 — Category | Rare; reframe as a process or machine |
| 2 — Judicial exception | Rewrite the claim as a mechanism, not a result |
| 3 — Anticipated | Reformulate around what the art did not do |
| 4 — Obvious | Gather secondary-consideration evidence, or narrow |
| 5 — Enablement | Gather the missing data before filing |
| 6 — Public disclosure | US: file within 12 months. Foreign: gone |
| 7 — Not worth it | Stop, and consider trade secret or publication |
Failing question three is usually a reformulation, not an ending. The close prior art shows what has been tried and where it fell short, and the unsolved problem is frequently visible in the failures documented there. See invention ideas.
Failing question seven has good alternatives. A trade secret lasts as long as secrecy does and requires no disclosure; defensive publication stops anyone else patenting it for very little money.
Worked example: running the test
A modification to a bicycle lock that alerts on tampering.
| # | Question | Result |
|---|---|---|
| 1 | Statutory category | Pass — a machine |
| 2 | Judicial exception | Concern — "detect and notify" is a result |
| 3 | Prior art search, 1 hour | Fail — 40+ patents, products on sale |
Stopped at question three, for nothing but an hour.
The reformulation
The search also showed what the existing products get wrong — false alarms from wind and passing traffic, severe enough that reviews report users disabling alerts.
| # | Question | Result on the reformulation |
|---|---|---|
| 1 | Category | Pass |
| 2 | Exception | Pass — a specific signal-processing mechanism |
| 3 | Prior art | Requires a fresh search on the signal processing |
| 4 | Obviousness | Turns on whether applying it here was obvious |
| 5 | Enablement | Needs real measurement data |
| 6 | Disclosure | Nothing public — clear |
| 7 | Worth it | Depends on whether anyone would practise it |
The invention came out of the search, not before it. That is the ordinary pattern and the reason question three is not merely a filter.
Who can answer which question
| Question | You can answer | Needs help |
|---|---|---|
| 1 — Category | Yes | — |
| 2 — Judicial exception | Partly | Software and diagnostics |
| 3 — Prior art | Yes — free search | Professional search before filing |
| 4 — Obviousness | Partly | Hindsight makes it hard |
| 5 — Enablement | Yes | — |
| 6 — Disclosure timing | Yes | — |
| 7 — Worth it | Yes | — |
Five of seven you can answer yourself. Obviousness is the hardest, because hindsight makes everything look obvious once you know the answer, and assessing your own invention is where that bias is strongest.
Subject-matter eligibility is worth professional input in affected fields. The Alice framework is applied in ways that are difficult to predict from the outside.
What changes the answer over time
| Change | Effect |
|---|---|
| Someone else publishes | Novelty may be lost |
| You disclose publicly | Grace period starts; foreign rights go |
| An unpublished application publishes | Becomes prior art from its filing date |
| Case law shifts on eligibility | §101 answer can move |
| You develop the invention further | May become patentable |
| The market changes | Question seven changes |
Patentability is assessed at a point in time and the ground moves. An assessment from two years ago says nothing about today, because applications publish continuously and each becomes prior art from its own filing date.
Which argues for filing once the answer is yes. Waiting to be more certain allows the certainty to expire.
Is it patentable: the checklist
- Search the prior art first, before assessing anything else. It is free and it eliminates most inventions.
- Search by function in five vocabularies, not by your name for the thing.
- Search products on sale too, since they are prior art whether or not patented.
- Assess subject matter only if you are in software, business methods or diagnostics.
- Check whether your description recites a mechanism or a result. Rewrite if it is a result.
- Read the close prior art properly. The reformulation is usually in it.
- Document failed attempts and unexpected results now, as §103 evidence.
- Test enablement honestly — could an engineer build it from your description alone?
- Check every public disclosure. Outside twelve months, US rights are gone too.
- Ask last whether a competitor would practise it. Patentable and worthwhile are different questions, and 58.6% of patents fail the second one.