Invention ideas are not property. Nothing is owned at the idea stage, nothing can be sold, and nothing can be infringed.
That is not a technicality about paperwork. It reflects what a patent actually is: a description specific enough that a skilled person could build the thing from it. An idea is a statement of a desired result. A patent claim is a statement of how the result is achieved.
The distance between those two things is the whole subject, and closing it is real technical work rather than a filing exercise.
What an idea is worth at each stage
| Stage | What you own | Can you sell it? | Typical cost |
|---|---|---|---|
| Idea | Nothing | No | — |
| Documented conception | Evidence of who conceived it and when | No | Your time |
| Enabling description | A draftable application | Not really | Your time |
| Provisional filed | A priority date, for 12 months | In principle, rarely in practice | ~$60–$300 in fees |
| Non-provisional filed | A pending application, publishable | Yes — applications are assignable | $8,000–$20,000 all in |
| Granted patent | An enforceable right to exclude | Yes | Plus $2,894–$14,470 in maintenance fees |
The first sellable point is the filed application. Before that there is no defined asset — no claims, no filing date, nothing an assignment document could identify. Buyers of patent assets need something with a serial number.
Documented conception still matters, even though the United States moved to first-inventor-to-file in 2013 and conception records no longer establish priority the way they once did. They remain useful for resolving inventorship disputes and for derivation proceedings where someone else files on your work.
Where invention ideas actually come from
Not from brainstorming. Ideas generated by trying to have ideas tend to be solutions in search of problems, and they fail prior art searches at a high rate because obvious needs attract obvious attempts.
Domain knowledge is the input that matters. The people who generate patentable inventions consistently are people deeply embedded in a specific practice, who notice failures nobody outside the field would see.
| Method | How it works | What it produces |
|---|---|---|
| Friction logging | Record every task in your own work that wastes time, breaks or requires a workaround, for two weeks | Problems you can verify exist, because you have the evidence |
| Adjacency transfer | Take a technique standard in one field and apply it where it is unknown | Often non-obvious, because the fields do not read each other's literature |
| Constraint removal | Ask what a product would look like if its most expensive component were free, or its heaviest were weightless | Reframes design around what the constraint was hiding |
| Failure archaeology | Study products recalled, discontinued or abandoned, and identify what defeated them | The problem is documented and the market is proven |
| Expired patent mining | Read patents that lapsed for non-payment and identify what has changed since | Working technology, freely usable, with a written record of what was tried |
Expired patent mining deserves particular attention because the raw material is enormous and nobody competes for it.
| 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% |
| Lost at the first maintenance fee | 14.2% |
Computed from 27,273,654 USPTO maintenance fee records covering 8,262,336 US utility patents. Nearly three in five granted US patents are abandoned before their term ends, and everything in them is in the public domain.
What that gives an inventor. A vast library of technically detailed, professionally drafted descriptions of things that were built or designed, freely usable, each one recording an attempt and often explaining what the alternatives were. A patent abandoned in 2011 for want of a $4,040 fee may describe something that only became viable when sensors got cheap. See the patent survival curve for the full dataset.
The record also warns you. If four patents in a field all lapsed at the first maintenance fee, that is evidence about the market, not just about the owners.
Fields where problems are documented
Prefer fields where the unsolved problem is written down somewhere, because you can verify the need rather than assume it.
| Area | Why room exists | Where the problem is documented |
|---|---|---|
| Repairability and serviceability | Right-to-repair legislation is forcing design changes on products never designed for disassembly | Regulatory filings, repair guides, teardown reports |
| Small-scale energy storage | Grid-scale chemistry does not transfer cleanly to household and tool scale | Standards committee minutes, safety recall notices |
| Water handling in agriculture | Efficiency requirements rising while equipment cost must stay low | Extension service reports, irrigation district rules |
| Accessibility hardware | Small markets deter large manufacturers, leaving mechanical problems unsolved | Clinical literature, user community forums |
| Industrial sensing in harsh conditions | Sensors fail in heat, vibration and contamination; failure modes are specific and documented | Maintenance logs, failure analysis papers |
| Materials separation and recycling | Mixed-material products were designed without recovery in mind | Waste processing studies, municipal contracts |
None of these is a shortcut. They are places where a person with relevant experience is likely to find a problem that is real, specific and not already solved — which is the only useful definition of a promising area.
Avoid categories that attract volume without differentiation. Anything where thousands of near-identical applications exist means both a crowded prior art field and a difficult obviousness argument.
Testing an invention idea before spending anything
Three questions, in this order, because they cost radically different amounts to answer.
1. Has it been done? Free. This eliminates the majority of ideas and is the single highest-value hour an inventor can spend.
| Tool | Cost | Best for |
|---|---|---|
| Google Patents | Free | Full-text search, easy to start, machine translations |
| USPTO Patent Public Search | Free | Authoritative US records, classification searching |
| Espacenet | Free | Worldwide coverage, family data |
| Ordinary product search | Free | Products nobody patented but that are still prior art |
| Professional search | $500–$3,000 | Before committing to a non-provisional |
Search by function, not by your name for it. You call it a "flow stabiliser"; the prior art calls it a "damping baffle." Write down five different ways a stranger might describe what it does, and search all five.
Search non-patent literature too. Journal papers, conference proceedings, product manuals, standards documents and technical forums are all prior art, and non-patent literature is where the damaging references usually hide.
2. Would it have been obvious? Harder, because hindsight makes everything look obvious once you have seen the answer. The useful test is whether the combination required a reason to combine, and whether anyone tried and failed.
3. Can you describe it well enough to build it? If an engineer in the field would have to solve significant problems your description does not address, you have an idea rather than an invention, and filing is premature.
What can be patented works through the statutory conditions those questions correspond to.
Worked example: an idea through the funnel
The starting sentence. "A bike lock that tells you on your phone when someone is tampering with it."
Where it stands as written
| Test | Assessment |
|---|---|
| Is it an idea or an invention? | Idea. It recites a desired result, not a mechanism |
| §101 risk | High as phrased — "detect and notify" is data collection and display |
| What is missing | How tampering is distinguished from ordinary movement |
The claim as stated would fail. Detecting an event and sending a notification is exactly the pattern that struggles under subject-matter analysis, and generic phone-plus-sensor implementation adds nothing.
The prior art search, one hour
| Search | Result |
|---|---|
| "bicycle lock accelerometer alert" | 40+ patents and applications; concept thoroughly covered |
| Retail search | Multiple products already sold with this exact function |
| Conclusion on the idea as stated | Dead. Both anticipated and commercially occupied |
Cost of learning this: one hour. Cost of learning it after filing a non-provisional: $8,000 upwards, plus the twelve months spent waiting.
The reformulation
What the search also revealed. The existing products share a documented failure — false alarms from wind, passing traffic and bumping, severe enough that reviews consistently report users disabling alerts entirely. The real unsolved problem is not detection. It is discrimination.
Reformulated: a method of distinguishing tampering from ambient disturbance using the specific frequency signature that tool contact with a hardened shackle produces, by comparing energy in a defined band against a rolling ambient baseline, and triggering only when the ratio exceeds a threshold sustained across a minimum number of cycles.
| Test | Assessment |
|---|---|
| Idea or invention? | Invention. Specific mechanism, specific parameters |
| §101 | A specific technical solution to a technical problem, not a result |
| §102 | Requires a fresh search on the signal processing, not on bike locks |
| §103 | Turns on whether this analysis was obvious to apply here |
| Enablement | Requires real measurement data — this is the remaining work |
The problem was found by the search, not before it. This is the ordinary pattern and the reason searching first is not merely a risk check — it is a source of inventions.
The spend decision
| Path | Cost | When it makes sense |
|---|---|---|
| Stop now | $0 | If the reformulation also proves anticipated |
| Provisional, prepared yourself | ~$60–$300 in fees | You have measurements and can describe them fully |
| Provisional, attorney-drafted | $1,500–$4,000 | You want the description to actually support later claims |
| Straight to non-provisional | $8,000–$20,000 | The market is moving and 12 months of deferral costs more than it saves |
A thin provisional is worse than none. Priority extends only to what the provisional actually enabled, so a two-page sketch filed to feel safe gives a priority date the eventual claims cannot rely on. See provisional patent application for what the filing does and does not preserve.
Sharing an invention idea without losing it
Disclosure has consequences that filing does not undo.
| What you do | US position | Rest of world |
|---|---|---|
| Nothing public | Clear | Clear |
| Told a friend, no obligation | Arguably a disclosure | Risk |
| Under a signed NDA | Not a public disclosure | Not a public disclosure |
| Conference talk, crowdfunding page, trade show | 12-month grace period running | Rights likely lost |
| Offered for sale | 12-month grace period running | Rights likely lost |
NDAs work with counterparties who need something from you. Manufacturers, contract engineers and development partners sign them routinely.
Large companies generally will not sign one for an unsolicited idea. Most operate a submission policy requiring you to waive confidentiality, because they may already have similar work underway and cannot accept the liability of having seen yours. Read the waiver before submitting anything.
Which means the sequence is: file, then talk. A pending application converts the conversation from "please respect my idea" into a discussion about an asset with a priority date. Then see patents for sale and how to sell my patent for what happens next.
Invention promotion companies
The industry that exists to monetise people with invention ideas rather than to monetise the ideas. Treat unsolicited approaches as marketing, because that is what they are.
Federal law requires disclosure. Under 35 U.S.C. §297, an invention promoter must tell you in writing, before you contract:
- How many customers they have had in the past five years
- How many of those customers received a net financial profit as a result of their services
- How many received license agreements as a result of their services
- Any other names the promoter has traded under, and any related companies
Ask for it in writing and read the second number. A firm with thousands of customers and a single-digit profit count has told you its success rate. Failure to make the disclosure gives you a civil action for damages.
| Red flag | What it usually means |
|---|---|
| Unsolicited call or ad, then enthusiasm about your idea | The enthusiasm precedes any evaluation |
| A paid "market evaluation" that comes back positive | Positive evaluations are the product |
| Large upfront fee for submission to manufacturers | Submission is cheap; the fee is the business |
| Reluctance to give the §297 disclosure | The numbers are bad |
| Pressure to decide quickly | Filing deadlines are real; sales deadlines are not |
| No registered patent attorney or agent named | Nobody accountable is drafting your application |
Legitimate alternatives exist. Registered patent attorneys and agents are searchable on the USPTO roster. Many bar associations run reduced-fee inventor clinics, the USPTO operates a Pro Se Assistance Program, and law school IP clinics take on inventors at no charge.
Turning an idea into income
Three routes, and they are not equally available.
| Route | What it requires | Realistic for a solo inventor? |
|---|---|---|
| Build and sell it | Capital, manufacturing, distribution | Hardest, highest return |
| License it | A granted or pending patent, plus a licensee who wants it | Possible, slow, usually royalty-based |
| Sell the patent | A granted or pending patent with claims someone practises | Possible, faster, one payment |
All three require an asset. None of them work on an idea alone, which is why the filing decision comes before the monetisation question rather than after it.
Value tracks whether anyone would practise the invention, not how clever it is. A patent covering something a competitor already wants to build is worth many times its cost. One covering something nobody makes carries option value that declines each year — and the 58.6% abandonment rate above is what that decline looks like at scale.
Which is the honest framing for the whole exercise. Most invention ideas should not be patented. Establishing that quickly and cheaply, before spending five figures, is the most valuable skill an inventor can develop.
From idea to filing: the checklist
- Write the idea down as a problem, not a solution. What fails, for whom, how often, and what it costs them.
- Search the prior art before spending anything. Google Patents, USPTO Patent Public Search, Espacenet, and ordinary product listings. Search by function, in five different vocabularies.
- Search non-patent literature too. Journals, manuals, standards, forums. This is where the references that defeat you usually are.
- Read what the search reveals about failures. The reformulation that makes an idea patentable is often hiding in the complaints about existing products.
- Restate the invention as a mechanism. Specific steps, specific structure, specific parameters. If it still describes a result, keep working.
- Check what you have already disclosed. Presentations, sales, crowdfunding, trade shows. If anything is public, the US twelve-month clock is running and foreign rights are probably already gone.
- Confirm you can enable it. Could an engineer in the field build it from your description alone? If not, gather the missing data before filing.
- Get the §297 disclosure from anyone offering to commercialise it for you, and read the net-profit figure rather than the customer count.
- File before disclosing further — see how to patent an idea for the full sequence.
- Decide the monetisation route while the application is pending, because building, licensing and selling require different evidence and it is easier to gather it early than to reconstruct it later.