The Inventor's Evidence File

An invention exists twice. Once as a thing that works, and once as a stack of paper that says when it was thought of, who thought of it, what it consists of and how anyone can be sure. The first version is the one inventors care about. The second is the one every institution downstream will actually read, and the gap between them is where most independent projects quietly come apart.

What follows is about the second version: the evidence file. Not a keepsake, not a scrapbook, but a working archive assembled with the assumption that somebody sceptical will one day go through it line by line — an examiner querying whether the specification supports a claim, an attorney reconstructing who contributed what, a licensee's counsel checking that the chain of title has no gaps, or, in the worst case, an opposing party looking for the one loose thread that unravels the rest.

The Premise

A Patent File Is Built From Paper, Not Memory

Almost every substantive question in patent practice reduces to a question about a document. Was the invention described in enough detail that a competent person in the field could build it without inventing anything themselves? That is enablement, and it is answered by reading the specification as filed — not by asking the inventor what they meant. Did the applicant have possession of the full scope they are now claiming? That is written description, and it is answered the same way. Who is properly named as an inventor? Answered by reconstructing conception, which means reconstructing dated records.

The system has moved steadily in the direction of paper. Most major granting offices now award priority to the first applicant to file rather than the first person to invent, which removes the old contests over who conceived a mechanism earliest and replaces them with a single hard timestamp. That sounds like it makes records less important. It does the opposite. Filing dates are now decisive, and everything an applicant wants to rely on must already be inside the document that carries that date. Material added later — a better explanation, an extra embodiment, a numerical range the inventor knew about but did not write down — gets its own later date, and in a contested case that later date is often fatal.

So the file has two distinct jobs. It supplies the raw material for a specification that is complete on the day it is filed. And it establishes the surrounding facts that the specification itself cannot carry: who did what, when, under what confidentiality terms, and with whose money. The first job ends at filing. The second runs for the full twenty-year term and, in a dispute, well beyond it.

Contents

What Actually Belongs in the File

A good file is narrower than a hoard and wider than a notebook. Six categories cover nearly everything that turns out to matter.

1. Conception records. Dated entries describing the problem and the specific mechanism that solves it, in ordinary technical language, with dimensioned sketches however rough. One entry per working session, never overwritten.
2. Reduction-to-practice evidence. Test logs with dates, conditions, measurements and outcomes — including the failures. A log showing three unsuccessful geometries and a fourth that worked is far stronger evidence of real development than a single triumphant photograph.
3. Search results. What was searched, where, with which terms and classification codes, and what came back. Expired patents belong here as well as live ones; a lapsed teaching defeats novelty exactly as thoroughly as a competitor's granted claim.
4. Disclosure ledger. Every person outside the inventor who has heard the idea, the date, the medium, and whether a confidentiality agreement was signed before the conversation or after it. Only the first of those is worth anything.
5. Contribution notes. Who suggested what, and when. Inventorship is a legal determination about who conceived the claimed subject matter, not a courtesy list of collaborators, and correcting it later is possible but expensive.
6. Filing and office correspondence. Receipts, application numbers, priority dates, examination reports, every amendment and every argument submitted in response.

That last category is the one inventors under-value most, because it looks like administration. It is not. The arguments made during prosecution become part of the public record and are read back against the patent for its entire life: a scope surrendered in a response to narrow around prior art cannot generally be recaptured later by reinterpretation. An inventor who cannot reconstruct why a claim was narrowed in year three is negotiating blind in year nine.

Integrity

Why Loose Records Fail and Ordered Ones Hold

The weakness of most inventor records is not their content but their mutability. A document is only evidence to the extent that its date can be trusted, and a folder of files on a laptop offers a date that anyone can change in four seconds. Three properties make a record durable, and they are worth engineering deliberately.

Fixity. Entries should be difficult to alter after the fact without the alteration being obvious. Bound and sequentially numbered pages, entries in ink, corrections struck through rather than erased and initialled with a date. In digital form, the equivalent is an append-only history: a version-controlled repository where each commit carries a cryptographic hash of everything before it, or a periodic cryptographic timestamp lodged with a third party. The mechanism matters less than the property — that a change made later leaves a trace.

Contemporaneity. Records written on the day carry weight that reconstructions do not. Retrospective entries are not worthless, but they must be honestly labelled as retrospective; a backdated entry that is discovered destroys the credibility of the whole file, including the parts that were genuine.

Independent corroboration. A record signed only by the person it benefits is the weakest form of proof. Corroboration by someone who understood the entry and had no stake in it converts a claim into a supported claim — which is why the practice of formally dating and witnessing invention records is worth the small friction it adds to each working session.

One historical practice deserves burying. Posting a description to yourself and leaving the envelope sealed — often called a poor man's patent — establishes nothing useful. It proves a postmark, not authorship, not completeness, and certainly not a filing date. Under a first-to-file framework it has no operative effect whatsoever, and the confidence it produces is actively dangerous, because it persuades people that they are protected while they talk freely to manufacturers.

Leverage

How the File Changes Every Later Conversation

The practical argument for keeping records well is not defensive. It is that a good file makes every subsequent transaction faster, cheaper and more favourable.

Start with drafting. An attorney working from a complete file — mechanism, variants, test data, search results — can draft a specification with a broad independent claim and a well-stocked set of dependent fallbacks, because the fallbacks come from the rejected variants the inventor already documented. An attorney working from a two-page summary and a phone call drafts narrowly, because narrow is all the material supports. The difference in scope is decided before anyone opens a word processor, and it is the single largest determinant of what the patent is eventually worth. It is also where practical help with patenting invention ideas earns its keep, since the questions being answered at that stage are simultaneously technical and legal and rarely go well when treated as either alone.

Then examination. First reports typically arrive somewhere between eighteen and thirty months after filing, and they are usually rejections. Responding well means finding, in the original disclosure, support for a narrower claim that still covers the commercially important version. A file with dimensioned variants and measured performance data has that support available. A thin one leaves the applicant choosing between a claim so narrow it is trivially avoided and an amendment that adds new matter and is refused.

Then licensing. Companies conducting diligence are not primarily assessing whether an invention is clever. They are pricing risk: is the chain of title clean, is inventorship correct, was anything disclosed before filing that could invalidate the grant, are there earlier confidentiality obligations to a third party? Every one of those is a documentary question, and every unanswered one becomes a discount, a warranty, or a reason to walk. Practitioners who have spent decades working alongside independent inventors tend to report the same pattern: deals rarely collapse over the technology, and often stall over paperwork nobody thought to keep.

There is a field-specific dimension too. Where an invention's value rests on measured performance rather than novel geometry — efficiency gains, emissions reductions, durability under load — the test log is the asset, and claims about improvement that cannot be evidenced are worth nothing in negotiation. Much of the recent activity in inventor-led environmental technology turns on exactly that kind of documented margin. Where value rests instead on physical form arrived at slowly, as in design-driven work in the built environment, the version history is the asset, because it shows which features were essential and which were incidental.

Maintenance

Keeping the File Alive for Twenty Years

A patent term runs twenty years from the filing date of the substantive application, not from grant, and the file has to stay usable for all of it. Three habits do most of the work.

Keep a single dated index at the front, updated whenever anything is added. A file nobody can navigate is functionally a file nobody has. Second, diarise the deadlines that carry consequences rather than the ones that carry reminders: the twelve-month anniversary of a provisional filing, after which its date is simply gone; the twelve-month convention period for filing abroad; the thirty- or thirty-one-month deadline for entering national procedures under an international application; and the renewal or maintenance fees, which recur for the life of the grant and which lapse patents with complete indifference to how valuable they were. Diarise each at nine months as well as twelve, because preparation is the part that needs the time.

Third, record the decisions, not only the documents. A one-paragraph note explaining why a claim was narrowed, why a market was abandoned, why a variant was dropped — these are the entries that make a file legible to a stranger years later, including the stranger the inventor becomes about their own project after a decade. Deciding not to pursue something is a perfectly respectable outcome. Drifting into the same position without noticing is not, and the only difference between the two, from the outside, is whether the reason was written down.

None of this is difficult and none of it is expensive. A bound book, a disciplined folder structure, a witness who signs, an index and five diarised dates cost less than a single hour of professional time. What they buy is the ability to answer, quickly and credibly, every question that the rest of the process will ask — and the difference between an invention that can be defended and one that merely happened.

Keep the record as though a stranger will read it. Eventually one will.