The portfolio that forgot it was one.

01 · the dormant-portfolio problem

Start with the measured facts. The PatVal survey of more than nine thousand European patent inventors found that about 36 percent of patents are unused: neither practiced by their owners nor licensed to anyone.¹ Follow-up work across seven thousand patents put numbers on the split, with roughly 17 percent held as pure blocking assets and a similar share simply sleeping, protecting nothing and earning nothing.² Universities run worse: most academic patents are never licensed at all. Meanwhile the intangible share of corporate value has climbed for four decades, with commonly cited estimates putting it near ninety percent of the S&P 500’s market value.³ The asset class grew; the reading of it did not.

And sleeping is not free. Patents carry maintenance fees that escalate with age, which produces a quietly absurd arrangement: companies paying an annuity to preserve options they have never priced and do not remember holding. When the fee notice arrives, the decision defaults to renewal, because nobody can say what the asset is worth, and the safest answer to an unpriced question is always yes. Multiply that by a few hundred assets and a few decades, and the drawer becomes a line item.

02 · why nobody read the drawer

The economics of reading were prohibitive, and it is worth being precise about why. A patent is written in claims language, which is designed for lawyers and examiners rather than operators, so the first task is translation: what capability does this actually describe, in words a businessperson can act on? The second task is collision: holding that capability against the living market, who is funding what, which job postings reveal which build-outs, where products are missing features that customers complain about, which standards are emerging that this asset touches. Translation times collision, across every asset and every adjacent market, is a matrix no advisory firm could staff, so the traditional IP audit sampled the twenty “most important” patents, which is to say the twenty everyone already knew about. The sampling was the problem: sleeping value is, by definition, in the part nobody reads.

This is the same inversion as the rest of this series. The bottleneck was never judgment; it was reading capacity. Fable 5 reads the whole drawer, restates every claim in operator language, and runs the collision scan against external sources continuously rather than once a decade. The judgment, what to do with a collision once found, stays human, and the four-outcome sort below is where it gets applied.

03 · the method, and the four honest outcomes

The scan runs in three passes. The inventory pass restates every asset in plain language: what problem does this solve, for whom, and what would a product built on it look like? No asset is exempt, because exemption is how sampling sneaks back in. The collision pass holds each restated capability against external demand signals: funding rounds in adjacent categories, hiring patterns that reveal what competitors are building, customer complaints about products that lack what the asset enables, and standards activity that could make an old claim suddenly load-bearing. A collision is a specific, named overlap between something you own and something the market is currently paying for, and most assets will not produce one, which is fine, because the scan’s job is triage rather than flattery.

Every collision then lands in one of four outcomes, and the honesty lives in there being four rather than two. License it, when the demand is real but the business is not yours to build; licensing is found revenue on an asset that was pure cost yesterday, and the markets-for-technology literature has documented for decades that firms systematically under-license out of inertia rather than strategy.^4,5 Build it, when the capability belongs inside your own product line and the collision is your own roadmap arriving late. Spin it out, when the demand is real, the business is not core, and the asset can seed an independent company, which is the Foundry’s second harvest running on IP instead of an operating problem, with the same structure: recruited founder, design partners, the client as founding minority holder. And lapse it, when the honest answer is that the asset protects nothing and collides with nothing, in which case stopping the maintenance payments is the return, and it is the only outcome available on day one with no further work.

ip_scan_rules:
  plain_language_first: >
    Every claim is restated as: capability, beneficiary, and
    the product it implies. An asset that cannot be restated
    plainly is flagged for counsel, not skipped.
  collision_requires_evidence: >
    A commercialization opportunity is a named overlap with a
    live demand signal (funding, hiring, complaints, standards),
    dated and sourced. "This could be big" is not a collision.
  four_outcomes_only: >
    Every scanned asset ends as license / build / spin / lapse,
    with the case stated in two sentences. "Keep and revisit"
    is a decision to keep paying; it must be argued, not defaulted.

04 · what it returns, for owners of sleeping drawers

The returns stack in order of certainty, and I present them that way deliberately, because it is the same discipline as the Foundry’s two harvests. The certain return is subtraction: lapsing true deadwood stops real payments, and it is the rare cost cut that requires no operational change at all. The probable return is licensing: even one collision converted to a royalty on a two-decade-old asset re-prices the whole drawer from expense to portfolio. The optional return is the spinout, and it is the most interesting one for mid-market owners and PE firms holding industrial companies with long R&D histories, because a spinout seeded by existing IP starts with its hardest asset, the defensible core, already built and already owned. The added-value test from Field Note 03 and the canvas loop from Field Note 05 apply unchanged; the only difference is that the second harvest was sitting in a filing cabinet the whole time.

For a PE operating partner, the portfolio version is worth stating plainly: an IP scan across eight industrial portfolio companies is the cheapest diligence-grade look at value creation the firm has not already priced, and it occasionally finds the one asset that changes an exit narrative. For a family owner, it is often simpler than that. Thirty years of solving your own problems leaves traces, and some of those traces are worth more outside your walls than they ever were inside them.

05 · the honest limits

Four this time, because the legal one deserves its own sentence. A patent is not a product: the commercialization gap between a claim and a paying customer is real work, and the scan finds candidates rather than companies. AI narrows legal questions but does not answer them: freedom-to-operate, claim validity, and infringement exposure are opinions rendered by counsel, and the scan’s job is to make sure counsel reads the right ten assets instead of a random twenty. Collision evidence ages: a demand signal from last quarter is a hypothesis, not an annuity, so the scan runs continuously or it decays into another one-time audit. And valuation discipline holds here as everywhere in this series: a sleeping patent with one collision is an option with a story attached, and it goes in the option column, never the headline, until someone pays for it.