Two Published Accounts of University Licensing: WARF and Stanford

Universities generally keep secret their licensing metrics. Yes, they report the number of inventions, patents, licenses, and startups in a given year–but they don’t report how those numbers relate to one another. The inventions reported in a given year have something to do with patents–but only in later years, since it takes about three years for a patent to issue, and a single invention can result in patents issuing for years thereafter through clever use of divisionals, continuations, and continuations in part, not to mention foreign counterparts of any and all of these. Similarly, licenses granted in a given year are generally for inventions disclosed in previous years. One invention might involve scores of licenses, and that might obscure the fact that many other inventions have not been licensed at all.

Given all that, here are two accounts of licensing practice that have made it into public view, along with some commentary that might help you understand what it means and doesn’t mean.


From Rima D. Apple, “Patenting University Research: Harry Steenbock and the Wisconsin Alumni Research Foundation.” At the time of the article, Professor Apple was a fellow in the University of Wisconsin department of History of Medicine. She is now professor emerita:

By 1985, the latest year for which statistics are available, WARF had seen the granting of 448 patents, of which 203 had been licensed. These have not all been financially successful: only 100 have produced income greater than expenses, and 10 alone have generated 90 percent of WARF’s royalty earnings. Yet patents, which by 1986 had brought WARF more than $30 million, account for only 20 percent of the foundation’s income. Of even more significance in the development of WARF’s substantial endowment was the skill of its trustees in investment. Under an early policy decision, WARF did not use the royalties themselves to fund research; instead, monies paid WARF were invested, and research was funded from the interest.

Dr. Apple does not provide the number of inventions WARF reviewed for management or inventions under management for which patents did not issue. According to early WARF documents–and consistent with University of Wisconsin policies until recently–inventors submitted inventions to WARF voluntarily. Instead of total inventions under management, we get the number of patents–without indication for how those patents relate to inventions. For instance, Steenbock received four patents on his Vitamin D invention(s). A similar situation involves the compounds behind warfarin, also handled by WARF.

If we look at the numbers presented for WARF, we could claim a 45% “licensing rate.” Of the inventions for which patents had been obtained, 45% were “licensed.” But keep in mind, this licensing rate involves voluntary submission of inventions. If a university unilaterally claims all inventions, then the pool of inventions that are claimed is often much greater than the number for which the university obtains one or more issued patents. The 45% licensing rate we might claim tosses all the inventions acquired by WARF but not patented. Including those inventions would likely reduce the “licensing rate” by half. And with licensing, we are not talking about achieving “practical application” of any invention–only that one or more licenses have been granted at some point in a given patent’s life (and those licenses may have later terminated–just because a patent has been recorded as “licensed” does not mean it is still licensed). That is, a “license” represents a transaction involving rights. The goal of technology transfer is not a rights transaction but the beneficial use of the invention to which the rights attach.

Next, however, we are told that only 100 produced income greater than expenses. That is, of the 203 patents licensed, only half of those licenses paid more than the costs of patenting. We could argue that those licenses never resulted in any commercial products–but we could also argue that those licenses were royalty-free licenses, and no revenue was expected, or even possible. In other words, the revenue return from any given license is a meaningless metric. Only when we know that all licenses anticipate commercial sales and payments are based on sales can we start to understand the relationship between licenses and income. In the case of Steenbock’s patents, for cereal products WARF charged an annual fee that scaled at milestones, rather than a running royalty on sales. With such a licensing practice in house, then, a company might take a license to a WARF patent, pay $15,000 (say) over three years for the right to hold the license, fail to develop any product for sale, and WARF would still have made more money than its patenting expenses on the deal. We still would not have any idea whether there ever was a product.

A further point. Ten patents account for 90% of WARF’s income. Four of these no doubt are the Steenbock patents, and probably more than one of the remainder are Link patents tied to the compounds behind warfarin. So two patent families account for a great deal of WARF’s patent income over the years from 1925 to 1985, despite activity involving nearly 450 patents and who knows how many inventions for which patent applications were filed but no patents issued.

Income, as well, becomes a problem, given WARF made its money two ways–from royalties on licenses and from investing in stocks and bonds. In fact, it appears that WARF used royalties as the input for an investment engine that relied on stock valuations rather than on, say, a share of the commercial value of each patent licensed. By 1985, 10 patents contributed 90% of royalty income but only 20% of WARF’s total income. WARF was an engine that didn’t actually need royalty income to generate further income–it had an investment portfolio for that.

In this WARF’s model is generally misunderstood. Universities starting their licensing programs with “big hit” inventions tend not to gather the revenues into an investment portfolio and then distribute income based on the success of investing royalties. University royalty sharing schedules make this misunderstanding clear–instead of investing royalties and handing out dividends from those investments, universities propose to divide up the royalties for expenditure each year. It’s like receiving a bar of gold, grinding it up into dust, and scattering it around the floors of various academic and administrative departments. Notably, UCLA, when it received its big payout for the patents underlying Xtandi, announced that it would invest the income. But here UCLA runs into the requirements of Bayh-Dole–which Bremer at WARF apparently helped to draft. Under Bayh-Dole–and the inventions underlying Xtandi are subject inventions under Bayh-Dole–a nonprofit must use royalties and any income earned with respect to subject inventions for the “support of scientific research or education.” Nothing about investing the wad and handing out interest if there ever is any. Thus, oddly, WARF officials set up Bayh-Dole to prevent universities from using WARF’s own founding model of pushing royalties into stocks and living off the income from the stocks. Of course, no university complies with Bayh-Dole, so it doesn’t really matter.


Here’s Kathy Ku, “Is Technology Transfer a Winning Proposition?” (Look Inside and search for “6,400”). Kathy Ku recently retired as the director of the Stanford Office of Technology Licensing and is one of the top university licensing officers in the country:

A brief summary of the high points of the Stanford technology transfer efforts is revealing. The good news is that during our thirty-six-year history (fiscal years 1970-2006), we received over 6,400 invention-technology disclosures and we generated over $1 billion in royalty income. Excluding our top three greatest royalty generators as of the end of the 2006 fiscal year (Google, with over $335 million to date; the Cohen-Boiyer patents with $255 million for Stanford and the University of California; and Functional Antibody with $125 million), $345 million as been received for “all other inventions.” It is sobering to note that three inventions generated 67 percent of the cumulative income.

Here we have an accounting based on inventions, not patents. Some of Stanford’s early income involved software that wasn’t patented, for instance, so “invention” might mean, for this account, more along the lines of “whatever was disclosed to Stanford’s OTL that the OTL chose to take under management.” Given how universities now define “invention,” it is not possible, in general, to expect “invention” to mean “patentable invention” or “invention that is or may be patentable.” An invention could mean pretty much whatever a university administrator or attorney says it is–software, know-how, improvements, tangible materials–whatever.

Furthermore, from 1970 until 1994, Stanford’s patent policy was similar to that of the University of Wisconsin–Stanford left invention ownership with the inventor “whenever possible”–that is, inventors could continue to own their inventions unless Stanford was legally required to take ownership. Thus, inventions might be reported to the OTL, but there would be no policy requirement for the OTL to require assignment of an invention to Stanford–there would have to be some other legal constraint on ownership. That in good part is how the Stanford v Roche litigation came to head with Stanford arguing that federal law compelled Stanford to have ownership of inventions made in research receiving federal support. Stanford’s revised policy asserts that a wide range of inventions “belong” to Stanford, and thus invention counting takes on a different significance, tracking not just the inventions the TLO finds it must manage plus those it is asked by inventors to manage–but rather all inventions that Stanford claims it owns. There’s a difference there that matters to practice involving selectivity and the judgment about what inventions should benefit (if that’s the right term here) from institutional administration.

So we will not be able to compare figures with WARF. Three inventions generate two thirds of the income. For Google, the income came by way of the value of stock in the company, and so reflects not simply the patented page rank algorithm but the overall future value of Google’s business activities when the company went public. Google may well have been valued as it was for the IPO regardless of whether it held a patent on its page rank algorithm, and regardless, too, of whether Google was actually still using the algorithm developed at Stanford and licensed to the company. For Cohen-Boyer, the income is shared with the University of California, and so Stanford reports the total income rather than the amount retained by Stanford, though Stanford managed the patent licensing effort. More so, Stanford did not attempt to extract maximum value (or anything like that) from its Cohen-Boyer patents. Instead, it offered licenses at very low cost, permitting pretty much anyone to obtain a license. Thus, the $255 million represents not so much the “value” of the gene-splicing patents to industry but rather represents the degree to which the inventions were broadly adopted by a rapidly expanding biotech industry.

The number of licenses granted might be an even better indicator, as Stanford could have achieved roughly the same result with an even lower licensing price. At some point, however, one enters a kind of “uncanny valley” in which the bother to pay the licensing fee is greater than ignoring the license altogether. Licensing rates don’t necessarily go up as the price of a license approaches zero. People just use the invention and ignore the licensing process. Thus, advocates of Bayh-Dole could argue that the federal government did a lousy job with licensing inventions–but only because the federal government practice did not depend on interposing a licensing step between citizens and access to government-supported inventions.

Similarly, there’s an odd behavior in which something that’s free may not be as attractive for commercial use than something that has a price attached. Why? One part of the behavior appears to be that something that’s free may not have a future–no further developments, no one available to provide assistance, no one training people in how the technology works, no place that demonstrates the technology. Thus, if there’s a price, and the money paid goes to maintain a going concern, whether nonprofit or for-profit, then the new technology appears more attractive than if it were free. It has nothing to do, really, with the “merits” of the technology in the abstract, but rather to do with the technology as intertwined with those that care enough about it to keep working with it and assisting others to do so as well. The pricing of technology–or services related to a technology–can be managed independently of patents or copyrights. This is an important point. A university can “make money” transferring “technology” and have no patent positions at all–income can come in other forms, such as research grants, consortium fees, affiliate program fees, workshop fees, sale of software or reference materials or data sets, donations of money or equipment. A technology transfer program can thrive without patents. I know. I ran such a program at a research university for decade.

More from Ku:

Out of 6,400 inventions in thirty-six years, only three of them were “big winners,” defined as having generated $50 million or more in cumulative royalty revenue. Of the 6,400 inventions, 16 have generated $5 million in cumulative revenue; 53 have generated $1 million or more in cumulative revenue; and 287 have made $100,000 or more.

Now it’s not clear whether Ku’s 16 inventions earning $5 million includes the 3 that earned $50 million. Let’s say so–that the 16 have earned between $5 million and $50 million. Those inventions represent substantial commercial value–we might posit that for a company to pay those sums, surely a commercial product must have been produced. That’s a good assumption, though $5 million is close to the thresholds where companies might pay upfront, properly motivated or desperate, to obtain exclusive access to an invention just to ensure they have access and some other company doesn’t get there first. So we might posit that 0.25%–one quarter of a percent–of Stanford’s inventions have likely seen commercial products formed around them. Among these are the Stanford patents on voltage controlled oscillators that form the basis for the production of “electronic” music.

A search of the USPTO returns about 1,500 U.S. utility patents assigned to Stanford from 1976-2006. We miss, then, the years 1970 to 1976. But given it takes three years or so to obtain a patent, the missing years won’t change much the total.  We might then figure that perhaps 23% of inventions resulted in patents. But that would only be ballpark. Some inventions might result in multiple patents, and other “inventions” might be handled without patents–software, materials, data.

The most notable statistic is that 80 percent of disclosed inventions did not bring in any revenue. In general, however, 90 percent of the licensed inventions brought in enough revenue to cover patent expenses.

That is, over 5,000 inventions didn’t result in licensing revenue. But we don’t have figures for the number of patents that didn’t result in licensing revenue. An invention that’s not patented may be practiced without a license, while an invention that’s patented creates a barrier to practice. Thus, unlicensed patents is a measure of how much a university has kept from broader practice. No wonder university administrators don’t report that metric, and AUTM doesn’t ask for it in its licensing survey. The same is true for exclusively licensed patents that never result in commercial products–the exclusive license does not change the availability of the invention for public use. But exclusive licenses are reported as part of total licenses, as if licensing is a measure of the success of “technology transfer” when really, it’s not. Technology transfer may happen without licensing. Licensing may not transfer a technology. And a transferred technology may never be used or developed. Licensing means little until one starts to make assumptions that aren’t warranted.

Ku reports that 90% of licensed inventions paid for their patent expenses. That’s generally a result of the university practice of insisting that a licensee pay the university’s patenting costs or setting an upfront licensing fee high enough that the university can cover its patenting costs. We might put it another way, then–in 10% of the university’s licensed inventions, the licensee failed to pay the required fees. Perhaps the company terminated the license (it still counts as a licensed invention, just not for very long); perhaps the company disputed the patenting costs and so paid only a portion; perhaps the license was royalty free; perhaps the license only paid a royalty if there ever were product sales. It’s just not possible to derive the reasons for nonpayment from the metrics of payment!

Of the rest of the inventions–53 making $1m or more; 287 making $100k or more–we are into the territory of upfront fees, reimbursement of foreign patenting costs (which can easily run over $100k), annual license maintenance fees and milestone payments, and realized income from sale of equity positions. While such activity points to an estimate of commercial (or investment) value for these inventions, it does not clearly indicate actual use or sale of commercial products. For that, we would want to see “earned royalties” based on sales. Even then, a royalty based on sales does not have to amount to much of anything. Patents can be licensed royalty-free, as they can be in standards based licensing and are in industry cross-licensing agreements. We might say, then, about 5% of Stanford’s inventions over 36 years were sufficiently valuable for companies to pay at least $100k to gain access to them. That $100k includes patenting expenses (often $10k to $20k for a US utility application) and various licensing fees. There’s no way to tell if payments also mean commercial use or commercial sale of what was licensed.

These two accounts–WARF and Stanford–suggest mostly that published metrics are no help in ascertaining how inventions made in research settings ought to be handled with regard to ownership, patenting, or licensing. We might learn that of those patented, only a very few–2% of patents in the case of WARF, 5% of inventions in the case of Stanford–result in significant income. But what is income for a technology transfer program dedicated to royalty-free licensing? What happens if the income that is received is not in exchange for a promise not to sue for infringement but rather for the delivery of research services or for assistance or instruction or in recognition of future advances? Then what?

Of course, if one looks at either WARF or Stanford from an investment portfolio perspective, both organizations have made a pile from patent licensing. Hundreds of millions into the billions. But making money in a portfolio, no matter how “successful” that may appear, tells us very little at all about the condition of the hundreds–thousands–of inventions that have entered such portfolios. For that matter, a university can make millions suing for infringement of an otherwise unlicensed invention–the patent in such a case (here, think Caltech suing Broadcom and Apple, say, over wi-fi patents incorporated into wireless standards) clearly was not required for the infringing company to invest in development of a commercial product, and thus any suit for infringement does nothing to “protect” the public from the prospect of nonuse. Any money made is purely from the exploitation of patent positions, not from a share of the beneficial use of an invention that, but for a patent monopoly, would never benefit the public.



This entry was posted in Metrics and tagged , , , , , . Bookmark the permalink.