The two main ways of holding an invention work by opposite mechanisms, and confusing them is the most consequential mistake available to an independent inventor. A patent is a bargain with the state: publish a full enabling description, and receive a time-limited right to stop others working it. Secrecy is the refusal of that bargain: disclose nothing, and rely on the fact that nobody else knows. The first requires publication. The second is destroyed by it. There is no arrangement that quietly gives both.
Two Mechanisms
One Publishes, One Hides
A patent runs for twenty years from the filing date of the substantive application, not from grant — so years spent in examination are consumed from the term, and a grant arriving four years after filing leaves sixteen. The application publishes at eighteen months from the priority date, whether or not it is ever granted. That is worth stating plainly: an applicant who files, publishes, and then abandons the application has handed the invention to the public and received nothing. The right is also territorial. It exists only where it was filed and maintained, and lapses everywhere that renewal fees go unpaid. What the instrument covers, and the conditions on which it is granted, are set out plainly in most official summaries of what a patent actually is — and the disclosure requirement is the part first-time applicants tend to underweight.
Secrecy has no term, no filing, no fee and no examination. Protection continues as long as the information stays secret and the holder takes reasonable steps to keep it that way — which in practice means restricted access, written confidentiality terms with anyone who sees it, exit procedures for departing collaborators, and documentation showing that these measures existed. That last condition is not decorative. A holder who cannot demonstrate reasonable protective measures generally finds the protection unavailable when needed, because the information was, by the court's assessment, not actually being kept secret.
A published application that is later abandoned is the worst outcome in the system: full disclosure, no right, no way back.
The failure mode nobody plans forCoverage
What Each Route Will Actually Cover
Patents cover inventions that are novel, involve an inventive step over what is known, and are capable of industrial application. They do not cover abstract ideas, mathematical methods as such, or discoveries of natural phenomena, and business methods and software occupy contested ground that shifts between jurisdictions and decades.
Secrecy has almost no subject-matter limits. Anything with commercial value that is not generally known qualifies: a formulation, a supplier list, a process temperature profile, a set of tolerances, an internal algorithm, the accumulated knowledge of which of forty variables actually matters. Much of the most valuable industrial knowledge in existence is held this way because it could never have been patented at all — it is not a single inventive step but a thousand small calibrations.
The decisive question is usually simpler than either analysis: can a competent competitor work it out from the product? A mechanism visible in a teardown, a geometry measurable with callipers, a circuit traceable from a board — these are self-disclosing. Secrecy over a self-disclosing feature lasts until the first unit is sold, because reverse engineering a lawfully acquired product is legitimate almost everywhere and defeats the protection completely. Conversely, a process step that leaves no trace in the finished article, or a parameter set that would take years of experimentation to rediscover, can be held quietly for decades.
Economics
Cost, Term and the Arithmetic of Choosing
Cost profiles differ in shape as well as size. Patenting is front-loaded and continuing: official filing fees in the low hundreds for an individual filer, drafting typically several thousand, examination responses adding more, translations and agents for each additional territory, then renewal fees that escalate through the term. A single well-drafted patent held in a handful of territories for its full life is a five-figure commitment at minimum, and the money is spent years before any revenue arrives.
Secrecy costs almost nothing to start and something permanently thereafter: access controls, agreements, and the operational drag of compartmentalising work so that no single collaborator holds the whole picture. It also carries an invisible cost that arrives at the worst moment — it is a weak asset in a transaction. A licensee can be shown a granted patent number and a claim set. Being shown an unpatented process and asked to take its exclusivity on trust is a much harder conversation, and it is priced accordingly.
Against that, the term comparison is stark. Twenty years is the ceiling for a patent, and it begins before the product exists. Secrecy has no ceiling at all. Where a formulation's advantage is durable and its composition is genuinely hard to determine, filing means volunteering to hand it over on a fixed date in exchange for an enforcement right that may never be exercised.
Ask one question first: could a competent competitor work this out from the product? Everything else follows from the answer.
The test that decides most casesEnforcement
What Happens the Day Somebody Copies You
Here the two diverge most sharply. A patent is infringed by anyone who works the claimed invention, regardless of how they arrived at it. Independent invention is no defence. Good faith is no defence. Ignorance of the patent affects the remedy in some systems but not the liability.
Secrecy protects only against improper acquisition or use — theft, breach of a confidentiality obligation, industrial espionage, a departing collaborator taking files. Against a competitor who developed the same thing independently, or who bought a unit and measured it, there is no remedy whatsoever. Two firms can hold identical secrets lawfully and forever, each unable to touch the other.
Proceedings differ too. Infringement is argued over a public document: the claims exist, their meaning is construed, and the comparison is technical. A misappropriation claim requires proving what the information was, that it had value, that reasonable measures protected it, and that the defendant obtained it improperly — all of which is evidentiary, and all of which depends on records created long before anyone contemplated a dispute. Dated access logs, signed agreements and versioned development history are the whole case, which is why the discipline described in the inventor's evidence file matters more for a secrecy strategy than for a patenting one, not less.
In Practice
Hybrid Positions and the Point of No Return
Most sophisticated positions are not a choice between the two but a division of the invention between them. Patent the self-disclosing part — the mechanism, the visible geometry, the architecture a competitor could read off the product. Hold as trade secrets the part that leaves no trace: the process window, the tooling adjustments, the parameter set that took two years of iteration to find. A granted claim then blocks the obvious route, while the undisclosed know-how makes copying the claimed thing well considerably harder than copying it badly.
The division has to be made before drafting, because a specification must enable the claimed invention. Deliberately omitting the information needed to make the claimed thing work risks an unenforceable patent; including it publishes it at eighteen months. Drawing the line — which embodiment is described, at what level of detail, and what is genuinely not required to practise the claim — is the substantive decision, and it is made once.
Two practical notes. First, secrecy is fragile in exactly the situations independent inventors most need: pitching, prototyping with third parties, crowdfunding, retail launch. A strategy that requires nobody outside the room to understand the product is difficult to reconcile with selling it. Second, filing forecloses secrecy permanently at the eighteen-month publication date, while secrecy forecloses filing the moment the information becomes public — and if a competitor publishes first, both routes close at once.
Which is why the choice deserves the same seriousness as the technical work. Popular accounts of invention tend to skip it entirely: the influence of speculative fiction on what people try to build shapes ambition rather than strategy, and profiles of long careers spent supporting inventors describe the same recurring gap between having a good idea and holding it properly. The documented cases are more instructive than the narratives — the development history of one household device brought to market by its inventors reads as a sequence of decisions about what to file and what to keep quiet, taken in order, on paper. Trade secrets and patents are not competing philosophies. They are two tools with opposite failure modes, and the work is knowing which part of an invention belongs in which.
Publish what can be seen. Keep what cannot. Decide before the drafting starts.