Inventing is a process with five repeatable stages: find a problem worth solving, map the constraints the solution has to live inside, generate far more concepts than you need, kill most of them within an hour each, and reduce the survivor to a described and tested mechanism. Working inventors do not wait for an insight. They pick a problem domain they know well, run the loop on it, and produce a usable concept every few weeks. The failure rate inside the loop is high by design. Killing 30 concepts to keep one is a good week, not a bad one.
What follows is the method, with the screening criteria and the time each stage takes.
Start With the Problem, Not the Product
The most common mistake at the start is beginning with a product concept and searching for a problem it fits. That order produces solutions nobody asked for, and it makes the market question unanswerable, because you cannot find the buyer for a solution to a problem nobody named.
A problem worth working on has four properties:
- Somebody is already spending time or money on a workaround. Duct tape, a second tool, a manual step, a service call. The workaround is proof the problem exists and it also tells you the price ceiling.
- The frequency is high enough to notice. A problem that occurs once every three years does not get remembered at the point of purchase.
- The person with the problem is the person with the budget. These come apart more often than expected, and when they do the sale is hard regardless of how good the solution is.
- The existing solutions fail for a reason you can name mechanically. “It is annoying” is not a reason. “The seal fails because the compression is uneven across the flange” is a reason, and it points at a solution.
Where Problems Come From
Five sources produce most independent-inventor patents.
Your own occupation. The largest single source. A nurse, an electrician, a farmer, a hairdresser sees the same friction 200 times a year and knows the tolerances the fix has to hold. Domain knowledge you already have is worth more than novelty instinct you do not.
Watching people work around something. People stop reporting problems they have adapted to. Watch what they physically do rather than what they say. A shim under a machine leg, a rubber band on a handle, a step performed twice because the first one does not hold.
Transfer from an adjacent industry. A mechanism standard in aerospace fastening that nobody has applied to consumer furniture. A dosing method routine in agriculture that nothing in the kitchen category uses. Cross-domain transfer is the most reliable source of non-obvious inventions, because the reference art sits in a classification code the examiner in your field does not search first.
A change in what is possible. A component that dropped in price, a material that became available, a regulation that changed. Every one of those opens a set of products that were not buildable last year.
Complaint mining. Product reviews at one and two stars, warranty claim categories, returns data, support forums. People describe failure modes in detail when they are angry. Read 200 of them in a category and the same three complaints will appear over and over.
Write the Problem Statement
Before generating anything, write the problem in a fixed format. This takes twenty minutes and it prevents weeks of work aimed at the wrong target.
Who: [specific user, not “people”]
Doing what: [the task, at the level of physical action]
Fails because: [the mechanical or functional reason]
Currently works around it by: [what they do now]
Would pay: [the observed price of the workaround]
Filled in: “A residential electrician pulling wire through an occupied attic cannot see the far end of the run, loses the fish tape into insulation, and currently works around it by having a second person hold a flashlight, costing an hour of a second body per job.”
That statement tells you what to solve, who buys, and what the ceiling is. A concept that does not attack the “fails because” line is off target no matter how clever it is.
Map the Constraints First
Concepts generated without constraints are entertainment. Write the constraint set before the first sketch.
| Constraint | Question to answer | Typical effect |
|---|---|---|
| Landed cost ceiling | What does the workaround cost the user today? | Sets the manufacturing budget at roughly 20% to 25% of retail |
| Size and weight | Where is it carried, stored, mounted? | Kills whole mechanism families early |
| Use environment | Wet, hot, dirty, outdoors, food contact, sterile? | Drives material and sealing choices |
| Power | Corded, battery, manual, none? | Battery adds cost, weight, certification, and a charging story |
| Regulatory | UL, FCC, FDA, CPSC, food contact, child safety? | Adds $5,000 to $150,000 and months, and it is not optional |
| User capability | One hand? Gloved? Untrained? Impaired? | Constrains actuation forces and interface |
| Distribution | Shelf, box, direct ship, contractor supply? | Sets package size and margin structure |
Constraint mapping takes a day. It eliminates half the concept space before you have drawn anything, which is the point.
The cost ceiling and the distribution line are the two most inventors guess at. Both can be worked through at no charge with a Small Business Development Center advisor funded through the Small Business Administration, and an hour there beats a month of estimating.
Generate More Concepts Than You Want
Quantity beats quality at this stage because early concepts are anchored to whatever you saw last. Target 25 to 40 distinct mechanisms per problem. Distinct means a different operating principle, not a different housing color.
Practical generators that produce volume:
- Change the energy source. If it is manual, make it spring, gravity, air, magnetic, electrical, thermal. Each swap produces a different mechanism family.
- Invert the motion. If the part moves, hold it still and move the frame.
- Move the function to a different part. Put the intelligence in the consumable rather than the tool, or the reverse.
- Remove a component and force the rest to compensate. Part-count reduction generates real inventions, and it also lowers cost, which is why it shows up throughout design for manufacturability.
- Change the state. Rigid to flexible, single-piece to segmented, continuous to indexed.
- Steal the mechanism from another field. Go read patents in a different classification code that solve the same physical problem.
Sketch each one at the level of “how does the force get from here to there.” Thirty seconds to two minutes each. Do not evaluate while generating, because evaluation kills the weird ones, and the weird ones are where the non-obvious inventions live.
Kill Fast
Screening is a separate session, and it should be brutal. Run each concept against five gates in order and stop at the first failure.
Gate 1: Does it solve the “fails because” line? Not the general problem. The specific mechanical failure you named. Most concepts die here.
Gate 2: Can it be built inside the cost ceiling? Estimate part count, material volume, and process. A 14-part mechanism aimed at a $9 retail product is dead.
Gate 3: Is there prior art on the mechanism? A twenty-minute search on the classification code for the mechanism family. Free, and it kills concepts before you invest in them. The technique is in how to search Google Patents.
Gate 4: Does it clear the regulatory path? Anything touching skin, food, children, medical use, or mains power has a certification cost that has to fit the business.
Gate 5: Is the difference non-obvious? If the only difference from prior art is size, material, or quantity, the patent road is hard. If it is an interaction producing behavior neither element would produce alone, keep it.
Out of 30 concepts, expect two to four survivors. That ratio is normal and it is what makes the volume worth generating.
Reduce the Survivor to Practice
The survivor now has to become specified. This is where an idea becomes an invention.
Write the mechanism out: every component named, its function, how it connects, how it moves, with dimensions or ranges, materials, and alternatives for each element. Then build the cheapest thing that answers the biggest open question. Cardboard, foam, hardware store parts, a printed part from a library makerspace. A first prototype exists to find out what is wrong, and the way to structure that is covered in how to make an invention prototype.
Expect three to seven iterations before the mechanism holds. That is the normal band, and the reasons it lands there are in how many prototype iterations you need. Each iteration should have one question attached to it. A prototype built without a question attached teaches nothing.
Then test it against the constraint list, with users who have the problem, doing the real task. Not a demo. The test protocol matters more than the build quality.
How Long the Loop Takes
| Stage | Time | Cost |
|---|---|---|
| Problem selection and statement | 1 to 2 weeks of observation | $0 |
| Constraint mapping | 1 day | $0 |
| Concept generation, 25 to 40 | 2 to 4 sessions of 2 hours | $0 |
| Screening to 2 to 4 survivors | 1 day plus search time | $0 |
| Written reduction to practice | 1 to 3 weeks | $0 |
| First prototype | 1 to 4 weeks | $0 to $500 |
| Iterations 2 through 5 | 4 to 12 weeks | $200 to $3,000 |
| Professional patentability search | 1 to 3 weeks | $399 to $2,500 |
Ten to twenty weeks from problem to a searched, prototyped, described invention. That is the realistic clock for someone working evenings, and it is the same loop run faster when a design firm carries the engineering.
When to Bring in Help
Two points where outside help changes the outcome.
The first is when the mechanism will not close. You have run five iterations and the geometry still fights you. That is an engineering problem and it is solvable, but not by more evenings. What to do at that point is covered in when an invention idea does not work yet.
The second is before you spend money on filing. Run the professional patent search after screening and before drafting, because the search result determines what the claims should aim at. Verify what you find against the USPTO patent databases yourself as well, since reading the closest three references teaches you more about your own invention than another month of sketching will.
If the concept clears both, the rest of the road, with the decision points and the numbers attached to each, is mapped in inventor questions answered.
FAQ
Do I need a technical background to invent something?
No, and domain knowledge matters more than engineering training at the front of the process. The problems worth solving come from people who do the work. Engineering is required to close the mechanism, and that part can be hired. What cannot be hired is knowing which 200 annoyances in a trade are the ones worth attacking.
How many ideas does an inventor go through before one works?
Working inventors generate 25 to 40 concepts per problem and carry two to four past screening. Of those, one usually survives prototyping. The ratios are why the generation stage is deliberately unfiltered. Judging concepts while producing them cuts the volume and removes the unusual ones first.
Should I patent every idea I have?
No. Filing costs money at every step and most concepts fail screening for reasons that have nothing to do with patentability. Filing is worth it when the concept has cleared prior art, has a mechanism that works in a prototype, and has a buyer who has said what they would pay. Before those three, a filing protects something that has not been proven worth protecting.
Where do I find problems if my job does not have any?
Complaint mining and observation. Read 200 one-star and two-star reviews in a product category and the recurring failure modes will surface. Or spend a day watching people perform a task and note every workaround. Both produce more raw material than introspection does, and both give you the buyer’s own language for the problem.
How do I know if my invention is any good before spending money?
Run it against the five screening gates: does it attack the named failure, does it fit the cost ceiling, is the mechanism already in the prior art, does the regulatory path fit the business, and is the difference more than size or material. A concept that clears all five is worth a search and a prototype. One that fails any of them is worth a note in a file and nothing more.