Every claim in every granted patent began as something considerably vaguer: a sentence said out loud, a sketch with arrows on it, a description that made complete sense to the person giving it and almost none to anyone else. Most of the work of patenting is the work of closing that gap.
01 / Raw Material
What an Idea Looks Like Before Anyone Has Drafted It
Ideas arrive as outcomes. "A bag whose straps never tangle." "A dispenser that cannot be over-filled." "A sensor that tells you before the pipe freezes." Each of these is a result, and results are the one thing a patent system will not let you own. If a claim could be written to cover any means of preventing straps from tangling, the first person to notice the problem would own every future solution to it, including the ones they never thought of and could not have built.
So the claim has to describe a thing, not an ambition: an arrangement of parts, or a sequence of steps, defined in terms concrete enough that a competent person in the field could reproduce it from the words alone. That requirement is why the first serious conversation about an invention almost always feels like an interrogation. What is the part that does the work? What is it attached to? What moves, and against what? At what point does the advantage actually appear?
It is worth knowing how little of a patent document is doing the legally binding work. A granted specification commonly runs twenty to sixty pages of background, drawings and description. The claims occupy the last page or two. Only those final paragraphs can be enforced against anybody. Everything else exists to support them, explain them, and set the boundaries of what they are later allowed to mean. Public accounts of invention rarely dwell on this — coverage that examines how popular storytelling frames the moment of invention tends to focus on the flash of insight rather than the eighteen months of definition that follow it.
02 / Extraction
Separating the Mechanism From the Product
An inventor is usually attached to a product: a complete object with a housing, a colour, a price point and a name. A patent is not interested in the product. It is interested in the single arrangement within it that nobody has described before. Separating those two is the most useful hour anyone spends before drafting begins.
The standard exercise is subtraction. List every element of the thing. Then remove them one at a time and ask whether the advantage survives. If the device still solves the problem without the aluminium bracket, the bracket is not essential and has no business appearing in the main claim. If it collapses without the spring-loaded catch, the catch is essential and must appear. What remains after the exercise — the shortest list of elements that still delivers the effect — is the skeleton of the broadest claim the invention can support.
Two failure modes show up here repeatedly. The first is including everything, producing a claim so specific that a competitor need only change the bracket to avoid it. The second is including too little, producing a claim that reads onto something already sold in the 1980s. Both are recoverable during examination, but only one of them is cheap to fix.
- EssentialRemove it and the advantage disappears. It belongs in the independent claim.
- OptionalImproves the product but is not required for the effect. It belongs in a dependent claim.
- CommercialColour, finish, packaging, branding. It belongs nowhere in the claim set at all.
A claim is not a description of your product. It is a fence. Every word you add moves the fence inward.
The rule that governs every drafting decision
03 / First Draft
Writing the Broadest Claim the Disclosure Will Carry
A claim is a single sentence, however long it runs. It has three parts. The preamble names what the thing is and, sometimes, what it is for. The transitional phrase joins the preamble to the body and does far more work than its length suggests: "comprising" leaves the claim open, so a competing device that includes every listed element plus twenty more still falls inside it, while "consisting of" closes the claim to exactly what is listed and is almost always a mistake outside chemistry. The body then sets out the elements and, critically, how they relate to one another.
Drafting conventions exist because ambiguity gets punished. Each element is introduced once with an indefinite article — "a housing", "a valve" — and referred to afterwards with a definite one — "the housing". A claim that mentions "the valve" before any valve has been introduced lacks antecedent basis and will draw a formal objection on its own, before the examiner has even reached the question of novelty. Relative terms invite the same trouble: "substantially rigid", "approximately vertical" and "a suitable adhesive" all read as clear until somebody with an interest in reading them differently arrives.
The draft that results is not one claim but a ladder. The broadest version sits at the top with the fewest elements. Beneath it sit narrower versions, each adding a further restriction, so that if the top rung is knocked out by prior art the ones below survive on their own terms. That architecture is the reason a patent with a rejected main claim is not a dead patent, and it is worth understanding how the layers of a claim set are built to fail gracefully before any of them are written.
This is also the stage where most independent inventors stop working alone. The vocabulary is unfamiliar, the consequences of a badly chosen word are invisible until years later, and there is no practice run. Structured help with patenting invention ideas is worth more here than at any later point, because the drafting choices made now set the ceiling on everything the application can ever become.
04 / Pressure Test
Reading Your Own Claim the Way an Examiner Will
Examination is a mapping exercise, not an aesthetic judgement. The examiner searches the published record, selects the closest documents, and attempts to map every element of the claim onto them. If a single earlier document contains all of the elements, the claim is anticipated and fails for lack of novelty. If two or three documents between them contain all of the elements, and there is a reason a skilled person would have combined those documents, the claim fails as obvious instead. Obviousness is where most arguments actually happen, because it turns on a judgement about motivation rather than a comparison of parts.
The timetable is slow enough to be easy to underestimate. Applications are typically published around eighteen months after the earliest filing date, whether or not examination has begun. A first substantive report often lands somewhere between eighteen and thirty months after filing, and in congested technical fields it can run considerably longer. Response periods are usually measured in a few months, extendable on payment of a fee. Meanwhile the twenty-year term runs from the filing date, not from grant — every year spent in examination is a year subtracted from the enforceable life of the patent.
The useful thing an inventor can do before any of that begins is to attack their own claim. Take the sentence, ignore what you meant, and ask what it literally covers. Then go looking for something old that fits the description. If a product from thirty years ago reads onto the wording, the wording is too broad, and the fix is a further element, not a further adjective. Deciding how far to push in each direction is a strategic question with real trade-offs on both sides, examined more closely in this account of how far a claim should reach before it becomes fragile.
None of this makes a first rejection a verdict. Most applications of any commercial interest receive one, and the exchange that follows is the ordinary mechanism by which scope gets settled. What separates a workable application from a stalled one is whether the original document left anywhere to retreat to. Product histories bear this out: the route behind one household mixing device that reached production looks nothing like a single decisive filing and everything like a long sequence of narrowing decisions.
05 / The Record
Why the Specification Decides What the Claim Can Become
There is one rule that quietly determines how much room an application will have later: nothing may be added after filing that was not disclosed at filing. Claims can be rewritten, split, narrowed and renumbered during examination, but every word of the new version must find support in the description already on file. Material introduced afterwards is new matter and is refused outright.
The practical consequence is that the specification has to be written for a future you cannot yet see. If the prototype uses a steel spring, the description should also record that a polymer spring, an elastomeric band or a magnetic return would work, and why. If the operating range is measured at forty degrees, it should record the wider band the mechanism tolerates and the narrower band it prefers. Each of those disclosures is a retreat position that becomes available the moment an examiner produces a document you did not know about. A specification that describes exactly one embodiment in exactly one material gives its owner exactly one argument.
That is also why narrowing later is a controlled operation rather than a free one. Restrictions taken on to overcome prior art tend to be permanent, and in many territories they limit how broadly the surviving wording can be read against a competitor who does something similar but not identical. The mechanics of that trade-off — what may be given up, what should be kept in reserve, and what a written argument costs you afterwards — are set out in this look at how narrowing is done without hollowing out the patent.
You cannot add to the disclosure after filing. Everything you might one day need to claim has to be sitting in the document on day one.
The constraint that shapes the whole specification
None of this is glamorous, and none of it resembles the way invention is usually portrayed. It is closer to drafting a contract with an opponent you have not met yet. Organisations that have spent decades working alongside independent inventors tend to describe the same pattern: the applications that hold up are rarely the cleverest ones, but the ones whose authors treated the description as an investment in options rather than a formality to be completed.
A rough idea becomes a defensible claim through subtraction, not addition — by finding the smallest arrangement that still solves the problem, describing it in terms nobody can reinterpret, and writing down every alternative before the door closes.
End of report