Trying to invent a new product works better backwards than forwards.

Not: what could I invent? That produces ideas nobody needed.

But: what problem do I understand that nobody has solved properly? That produces inventions, because you already know why the existing answers fail.

And the second step costs nothing — searching what has already been tried, which eliminates most ideas and shows where the real gap sits.

What you are trying to produce

Not But
A clever idea A specific implementation
Something nobody thought of Something non-obvious
A concept A described, enabled invention
Inspiration A solution to a costly problem

The sequence

Step Cost What it does
1 Find a problem you understand Free
2 Search the prior art Free
3 Read the close references Free
4 Design around the gap Time
5 Test the smallest version Varies
6 File before disclosing ~$60 + drafting
7 Develop, or license, or sell

Steps one to four cost nothing but time, and they decide whether anything after them is worth doing.

Why outsiders struggle

Outsider sees Insider knows
An obvious improvement It was tried in 2009 and failed
A simple fix Why the constraint exists
Inefficiency The regulation behind it
A gap Why nobody fills it

Domain knowledge tells you which gaps are real. That is the whole advantage.

The one rule

Search before you spend. It is free and it changes what you do next.

Problems, not ideas

Source Quality
A problem you encounter in your own work Highest
Something you repeatedly work around High
A failure you have seen many times High
A cost everyone in the field accepts High
Brainstorming sessions Low
"Wouldn't it be cool if" Low

Domain knowledge is the asset. Someone who services equipment sees failures the designers never encounter, and that gap is where inventions live.

The best problems are ones people have stopped noticing because they have worked around them for years.

Write it down as you go

Record Why
The idea, dated Evidence of conception
Search results What you found and when
Alternatives considered Cannot be added to a filing later
Test results Supports non-obviousness

Ask this about the problem

Question Why
Does it cost someone real money? Determines whether anyone will pay
How often does it occur? Frequency drives value
What do people do now instead? The workaround is your competition
Why has nobody fixed it? Sometimes there is a good reason
Who has the budget to fix it? Your buyer

"Why has nobody fixed it" deserves honesty. Occasionally the answer is that nobody tried. More often it is that the obvious fixes have been tried and failed, and finding out which is the point of the search.

Tool Covers Cost
Google Patents Full text worldwide, translations Free
USPTO Patent Public Search US records, classification Free
Espacenet Foreign families Free
Product listings and marketplaces A product on sale is prior art Free
Trade publications Where examiners search least Often free
Academic and conference papers Frequently missed Often free

Search by function in several vocabularies. What you call a flow stabiliser, the prior art calls a damping baffle, and one vocabulary finds one slice.

Search products as well as patents. Something sold but never patented is still prior art, and no patent database contains it.

Use CPC classification browsing to find the neighbourhood keywords miss. See patent an idea free.

Search terms to try

Approach Example variation
Your term What you call it
The industry term What suppliers call it
The function What it does
The problem What it prevents
The component What it is made of

Reading the close references is the valuable part

What to look for Why
What was tried Saves you repeating it
Why it fell short Stated in the background section
What the claims actually cover Usually narrower than the title
Whether it is in force Expired art blocks patenting, not selling
Who filed it Who else works on this

The reformulation usually comes from here. Understanding exactly what the closest reference discloses shows you where the remaining space is.

Most people skip this step and treat the search as pass or fail. It is neither — it is research.

Novel is not enough

Requirement Test
Novelty (§102) Not disclosed in a single prior art reference
Non-obviousness (§103) Not a predictable step for a skilled person
Eligibility (§101) Not an abstract idea or natural law
Enablement (§112) Can be built from your description

§103 rejects more applications than §102. Combining known elements predictably is obvious even where nobody has done it before.

That is the bar most inventions fail, and the prior art search gives you a strong indication of it before you spend anything. See what can be patented.

Worked example: the reformulation

An equipment technician notices filters clogging early on one machine type.

Step Action Finding
1 Problem from own work Filters fail at 40% of rated life
2 Free search, CPC browsing 60 close documents
3 Read the 6 closest All use a fixed-aperture baffle
4 Why do they fall short? Efficiency drops outside design velocity
5 Reformulate Aperture that varies with flow velocity
6 Search again on the new concept 2 close references — both need a fixed reference standard
7 Reformulate again Self-referencing, no external standard
8 Novel over both Filed

Step three produced the invention, not step one. The problem was known; the gap was visible only after reading what had been tried.

Step six is the one people skip. Searching again after reformulating is what turned an anticipated idea into a filable one.

Total cost of steps one to seven: nothing.

Adjacent gaps

Where incumbents focus Where gaps persist
Capability and performance Servicing and maintenance
Headline specifications Installation and setup
The main use case Edge cases and failure modes
New products Replacement and repair
Features Cost and simplicity

Large companies optimise for capability because that is what sells at launch. Servicing, repair and simplicity are frequently under-filed relative to how much they cost users.

Those are the areas an individual can realistically compete in, and they are visible to anyone who works with the equipment.

Testing before spending

Test Answers
Does the physics work? Calculation or simple rig
Smallest working version Proof of concept
Would anyone buy it? Ask people with the problem
At what price? Determines whether the economics work
Is the prior art crowded? From your search

Build the smallest thing that tests the riskiest assumption, not a finished product. A rig on a bench answers the question a polished prototype answers no better.

Test the invention against §103

Ask Bad sign
Is this a combination of known elements? Common
Would the combination be predictable? Obvious
Did others try and fail? Supports non-obviousness
Is there an unexpected result? Supports it
Did the field consider it impossible? Strong support

"Others tried and failed" is the most useful evidence you can gather, and reading close prior art is where you find it.

Disclosure destroys rights

Action Effect
Public disclosure before filing Forfeits rights in most countries
US grace period 1 year — US only
Disclosure under a signed NDA Protected
Trade show, crowdfunding, social post Public disclosure
Offer for sale, even one unit Public disclosure

File first, or disclose under NDA. A provisional costs about $60 in micro entity fees and preserves the position for twelve months.

Prototype shops need NDAs signed before the technical conversation, not after the quote. See manufacturing a prototype.

Timing the filing

Point Detail
Before any disclosure Non-negotiable for foreign rights
After the search So you know it is worth it
After reformulating So the claims are the right ones
Provisional first Defers cost 12 months

Patent or secret

Patent Trade secret
Duration 20 years from filing Indefinite while secret
Cost $8,000–$20,000 + fees Security measures
Publishes At 18 months, permanently Never
Protects against independent invention Yes No
Best for Visible products Undetectable processes

Filing forecloses secrecy permanently. The application publishes at eighteen months whether or not it grants, and abandoning does not restore it.

A process running inside your own factory is frequently undetectable, which makes trade secret the stronger route.

Combinations count

Type Patentable?
A wholly new principle Rare
An improvement to something existing Common
A combination of known elements If non-obvious
A known element used in a new field Sometimes
A predictable combination No — obvious

Most granted patents are improvements or combinations, not breakthroughs. The bar is non-obviousness, not novelty in the everyday sense.

The test is whether a skilled person would have found it a predictable step, which is why understanding why previous attempts failed matters so much.

Crowded fields

Effect Consequence
Dense prior art Narrower claims
Well-funded incumbents filing Hard to get broad coverage
More rejections, more rounds Higher prosecution cost
More potential infringers If you get a claim, it matters

Look for adjacent gaps rather than the field's main problem. Incumbents optimise for capability; servicing, cost and simplicity are frequently under-filed. See new invention ideas for the future.

Where prototypes help and do not

Helps with Does not help with
Understanding the invention Filing — enablement is written
Proving it works Novelty
Demand testing Non-obviousness
Investor conversations Priority date
Finding failure modes Anything legal

Build to learn, not to file. The specification requires description, not a working model.

What it costs

Item Micro entity
Free prior art search $0
Provisional USPTO fee ~$60
Provisional drafting $1,500–$4,000
Non-provisional USPTO fee ~$400
Non-provisional drafting $5,000–$15,000
Prosecution $2,400–$8,000
Maintenance fees, lifetime $2,894

Micro entity status cuts USPTO fees by 80% for the whole life. Check eligibility before paying anything. See small entity status.

What usually happens

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.

Nearly three in five granted patents are released early by their owners, usually because nobody practises the claims. That is the base rate any invention is competing against, and the free search is what keeps you out of it.

Common mistakes

Mistake Cost
Skipping the prior art search Filing on known art
Spending on prototypes before searching Wasted
Disclosing before filing Foreign rights, permanently
Treating the search as pass or fail Missing the reformulation
Inventing outside your domain No insight into why solutions fail
Paying an invention promoter Activity without an asset

The last one is worth naming. Federal law requires invention promoters to disclose how many customers made a net financial profit — ask for that number before paying anyone. See invention companies.

How to invent something new: the checklist

  1. Start from a problem you understand, not from a blank page.
  2. Ask whether the problem costs someone real money.
  3. Search the prior art before anything else. Free, and it eliminates most ideas.
  4. Read the close references properly. The reformulation is usually in them.
  5. Search again after reformulating.
  6. Test for non-obviousness, not just novelty. It fails more applications.
  7. Build the smallest thing that tests the riskiest assumption.
  8. File before disclosing if foreign rights matter at all.
  9. Consider trade secret for undetectable processes. Filing forecloses it.
  10. Check micro entity eligibility before paying any USPTO fee.