Patent Strategy for Startups

The most expensive patent mistake a startup makes is not a bad filing — it is spending scarce capital and founder attention on patents that do not matter for the business. Before anything else, answer honestly: is a patent actually part of your moat?

File when: your core differentiator is a functional innovation a competitor could copy or reverse-engineer; you are in an IP-heavy field (hardware, biotech, medical devices, deep tech, materials); patents materially strengthen your fundraising or acquisition story; or you need to deter a larger incumbent.

Defer or skip when: your moat is execution speed, brand, data, or network effects; the technology will be obsolete before a patent grants (often 2 to 4 years); you could not afford to enforce a patent anyway; or disclosing how it works (which a patent requires) helps competitors more than the monopoly helps you. Plenty of great software companies never patent anything.

For most startups, the right opening move is a strong provisional application. It locks in a priority date cheaply and buys you 12 months of "patent pending" while you validate the market and raise money — before committing to the far larger cost of a full non-provisional.

The discipline that matters: the provisional must FULLY describe the invention. A provisional gives you a priority date only for what it actually and enably discloses. A rushed two-page provisional that leaves out the real details is a trap — the details you add later get the later date and lose to anything published in between. Spend the time (or money) to make the provisional complete.

Then, within 12 months: if the invention has proven valuable, convert to a non-provisional (and consider a PCT if you need international protection). If it has not, let it lapse — your sunk cost was minimal, and you learned something.

Rough orders of magnitude: a DIY provisional can cost just the USPTO fee (a few hundred dollars at micro-entity rates), while an attorney-drafted provisional runs a few thousand. A full non-provisional with professional drafting and prosecution typically totals $8,000 to $20,000+ over its life, plus maintenance fees later.

Claim your discounts. Most startups qualify for small entity (50% off USPTO fees) or even micro entity (75% off) status — confirm which applies and use it.

Where NOT to cheap out: the specification. A thin spec limits claim breadth, invites Section 112 challenges, and forecloses valuable continuations. If you economize anywhere, economize on the number of applications, not on the quality of the one that protects your core technology.

Patent the things competitors can see and copy. If your differentiator is visible in the product or reverse-engineerable once sold, a patent is the right protection — a trade secret offers nothing against reverse engineering.

Consider a trade secret for things you can genuinely keep hidden — a manufacturing process, a training methodology, a backend algorithm that never ships to customers. A trade secret lasts as long as it stays secret and costs nothing to "file," but vanishes the moment it leaks or is independently discovered.

The decision is per-invention, not company-wide. A hardware startup might patent the device, keep the manufacturing process a trade secret, and copyright the firmware — all at once.

One strong patent on your core technology beats five weak ones on peripheral features. Concentrate resources on the inventions that actually protect the business.

Use continuations to stay flexible. Keep at least one application pending in your core family so you can add claims later that target how competitors actually implement around you. Once a patent issues and the family closes, that flexibility is gone forever.

Sequence with funding. File the core provisional pre-seed or seed; convert and expand the portfolio as you raise larger rounds and the technology (and its commercial value) becomes clearer. Do not front-load a big portfolio before you know what is worth protecting.

Ownership first. Investors check whether the company actually owns its IP: is there a clean chain of title, with every inventor (founders, employees, contractors) having assigned their rights to the company in writing? Missing assignments are a classic deal-killer.

The employer-overhang question. Did any founder develop the core technology while employed elsewhere, on a previous employer's time or equipment? If so, that employer may have a claim to the IP — an existential risk a sharp investor will probe.

Freedom to operate. Can you actually sell your product without infringing someone else's patents? A patent on your own invention does not guarantee you can practice it if others hold blocking patents.

Real moat vs decoration. Investors increasingly distinguish patents that cover a genuine competitive barrier from patents filed for show. A single well-drafted patent on the core technology, with clean ownership, beats a pile of weak filings.

Public disclosure before filing. Demoing at a conference, publishing, or launching before you file can destroy your patent rights — immediately and permanently outside the US (most countries have no grace period). File before you disclose.

Naming the wrong inventors. Inventorship is a precise legal concept; adding a co-founder who did not contribute to a claim, or omitting one who did, can render the patent unenforceable. Get it right.

Using employer resources. Building your startup's core tech on a day-job laptop or cloud account can hand your former employer a claim to it. Keep everything on personally-owned equipment.

Thin provisionals. Filing a vague provisional creates false confidence in a priority date that does not actually cover your real invention. Describe it fully.

Patenting the wrong things. Spending on patents for features with no commercial value, while leaving the actual moat unprotected, is pure waste.