A note on the middle nobody counts

Every technology transfer office I have worked in reported two things well. What came in, and what went out.

The US figures for fiscal 2025 give the shape of it. 28,278 invention disclosures. 13,729 new patent applications filed. 10,277 licenses and options executed. 1,095 startups formed. Those are published and comparable across hundreds of institutions.

One caution before anyone does arithmetic on them. These are separate annual flows, not one batch followed from disclosure to signature. The ratio of 10,277 to 28,278 is not a conversion rate, and reading it as one is how a field talks itself into the wrong diagnosis.

Both ends count events. A disclosure is a form submitted. A license is a document signed. Events are easy to count because they happen on a date. Between them sits work that is not an event.

The three things in the middle

Evidence someone outside the lab will trust. Not a result that satisfies a reviewer, but one that survives a skeptical engineer who has to put their own budget behind it. These are rarely the same experiment.

Rights clean enough to sign. Inventorship settled, funder obligations mapped, background and foreground separated, co-ownership resolved. None of this produces a publication or a patent count. All of it stops a deal.

One user willing to go first. Not a letter of support. A named operator who will take delivery, carry the disruption of being first, and say afterwards whether it worked.

A project can hold a granted patent and a strong paper and still be unlicensable because one of these is missing. In my experience it is the second or the third, almost never the science.

Why it goes unfunded

Because the layer is not reported, it rarely gets funded on purpose. It gets done in the gaps, by people hired to do something else, so it gets done inconsistently and nobody sees the pattern.

A budget line follows a reported number. We report disclosures and licenses, so we resource intake and deal closing, and the stretch between them goes to whoever has time.

The same shape has now turned up in these notes three times. We inherited a founding statistic nobody audited and kept its habit of counting what we hold and what we signed. A patent does not buy the delivery capability that decides who gets paid. Industry's absence when the research question is chosen is the binding constraint. The middle layer is where all three meet.

Two objections worth taking seriously

We already track this. We run stage gates and readiness levels. That is the closest instrument we have. The test I would apply is what happens when a project sits at the same readiness level for two years. Usually the label is accurate. What is missing is a budget line that owns the move.

That is what proof-of-concept funding is for. Partly. That money buys experiments well. It rarely pays to clear a right or secure a first user, the two parts most often blocking.

One change that costs nothing

At your next portfolio review, write one line per project: the next piece of evidence it needs, and who would have to believe it. No new system, nothing to report upward. It will tell you more than another patent count.

What I could not establish

Whether any national survey reports this layer. I looked for a reporting line covering evidence generation, rights clearance or first-user access and found none. Absence from my search is not proof of absence.

Whether a GCC or regional equivalent of the US series exists on a matching definition. I did not find one, so the argument here is about practice, not a comparison between regions.

What the 2025 movement means. Disclosures rose nearly 8%, licenses with startups 15%, and new US patent applications fell 4%. The published data does not separate more invention from better reporting, and I will not guess which it is.

One request

If you work in this layer, I would value one example. A project where the blocker was not the science but a missing piece of evidence, an unclear right, or no first user. One line is enough, and the organization stays unnamed.

---

Sources

• AUTM, Driving the Innovation Economy: U.S. Academic Technology Transfer in Numbers, FY2025 infographic. Disclosures, patent applications, licenses and options, and startups formed. Checked 1 October 2026.
• AUTM, 2025 Licensing Survey, for the year-on-year movement: disclosures up nearly 8%, licenses up more than 7%, licenses with startups up 15%, new US patent applications down 4%. Checked 1 October 2026.