Protection Matrix
Portfolio tools score the patents. None of them asks whether the product on your roadmap is read on by a claim you hold. The Protection Matrix puts a client’s own features down one axis and their patents’ claim chains across the other, and answers that question element by element — with the deciding claim language quoted on every finding.
Get in touchThe problem
A portfolio is built from invention disclosures over years. A product is built from a roadmap over months. The two drift, and nothing in the normal reporting cycle brings them back together — because the normal reporting cycle only ever looks at one of them.
A citation-derived index measures how the field responded to your patents; it has no idea what you sell. A weighted scorecard ranks your patents against each other and never asks about the thing on the other side of the comparison. Both can report a healthy portfolio while your best feature goes unclaimed.
Core capabilities
The matrix is only useful if a gap in it is real. Every structural choice below exists to stop the grid manufacturing one — or hiding one.
A whole product is never covered by a claim; a feature can be. So the rows are the features a claim could read on, and they come from the client — what they told you, a datasheet, a specification, a page they published. They are never derived from the portfolio: a product list inferred from the patents is covered by construction, and a matrix built on one would report no gaps at all. The gaps are the deliverable, so the row axis has to be independent of the column axis.
Only the head of a claim chain is a column. A dependent is narrower than the claim it narrows, so it can never supply coverage its parent lacks — it travels into the same reading instead, and the cell records which claims were actually read. A detached dependent is kept and counted, not asserted independent.
Every column carries where its claim stands — granted, under challenge, allowed, in examination, or lapsed. Coverage by a lapsed patent is not coverage today, and a claim that has not issued is not coverage yet. Green is the one signal read without checking, so the lapse is resolved before the colour is chosen.
Each matter names the office its right belongs to and how that was decided — stated by an attorney, derived from the identifier, or not established at all. A feature covered only in EP is exposed everywhere else, so one grid cannot report a single protection position unless every column says where it applies.
A proposed gap is one reading of one claim against one record. It is drawn hollow and held for review; it becomes a finding when an attorney accepts it, and only then is it filled in. A rejection needs a reason and keeps its row, so a later pass cannot quietly re-propose what somebody already judged.
An empty cell means nobody has looked, which is not the same as no coverage — and it must never be rendered like either answer. Every unassessed cell carries why: no technical record on file for that product, a claim whose wording is not on record, a matter outside what the portfolio is technically about, or a pass that ran out of budget. Each reason comes with the thing that would resolve it, because the fix for “we could not look” is different from the fix for “we looked and the record is silent”.
The deliverable
Four sections in a fixed order — Exposure, Awaiting Your Decision, Protected, Not Assessed. A document that opens with “Protected” reads as reassurance, so it does not.
How many product-and-claim pairs carry an assessment, and how many were never reached. With a hole in the grid, an empty Exposure section says so — it is not a clean bill of health.
The deciding claim element and its verbatim quote, the claims read alongside it, the standing of that claim, who accepted the finding, and when.
The title, the payload and the filename all name what the report covers — one client, or one portfolio. An unlabelled file is the one somebody attaches believing it covers everything.
What it needs
The matrix asks an element-level question, and only something that says how a product works can answer it.
Paste what the client told you on the call, upload a datasheet, manual or specification, or link a page they published. A marketing page states what a product is for rather than how it works, so the matrix marks what it reaches and what it does not — an element a product page never addresses is a question for the client, not a gap.
Nothing is fetched on a client's behalf without their consent on file, and only within the domains they approved. “Never asked” and “said no” are kept apart, because they are fixed by opposite actions, and a revocation keeps a record of who had granted it.
The patents across the top are the same matters the portfolio tracks, with the same claim state resolved from the same file wrappers. The matrix is where that record is finally pointed at the thing the client actually sells.