Dealing with Asynchrony in Technology Transfer

Gio Wiederhold

Stanford University
December 1995

An earlier version appears in:
in P.Apers, M.Bouzeghoub, and G.Gardarin: Advances in Database Technology -- EBDT'95; Lecture Notes in Computer Science, vol.1057, Springer Verlag 1996, pages 631-634.

A major problem discussed throughout industry and government is Technology Transition (TT). New computing concepts originate in industrial, governmental, and academic laboratories must to be disseminated and, if found effective, adopted. The receptors are industrial development, organizations devoted to improvement of existing, and enterprises doing systems integration. The use of computing is broad, an most of the receptors are not primarily focused on computing but have a broader palette. There are many problems that hinder technology transfer, this note focuses on the problem of temporal and terminological gaps. We believe that these gaps encompass a large and critical fraction of TT problems. We will deal here with innovations and developments that are beneficial and should be marketable at some point, and not with the issues of irrelevant research, nor with the filtering needed to determine relevancy.

The Problem

Industry is rarely ready to accept an innovation when it first presented. There are many reasons for lack of acceptance: the two major ones are:
  1. The innovation is not understood by industry because, as a byproduct of the innovation new terms have been defined.
  2. The innovation is understood or at least understandable, but there are no resources at that time to try to develop and market the innovation

These conditions are so common that it appears that direct TT from academic or industrial laboratories to industry must be the unusual case. To address these problems we will define a third party, namely a 'transition agent' (TTA) interposed between researchers and industry. The role of a TTA is to be a holder and developer for research results in preparation for industrial requirements.

Types of Innovation

For technology transfer we can consider three major types of innovation:
  1. The bright idea that solves a problem one encounters
  2. The generalization that formalizes the problem and enables publication and dissemination
  3. The specialization that makes an engineering solution tractable
I consider the second type to be the principal province of academia.

  1. Bright ideas
  2. The seeds of innovation typically occur in an industrial setting, where the practioner is faced with a specific problem. In computer science we had instances in industry of compilers, schemas, databases, object technology, optimistic and locking concurrency controls, recovery mechanisms, and transaction processing many years before there were publications about them. An exception is time-sharing, where a practical needs to allow many students to use single computers effectively caused academia to initiate innovation.

    However, these pragmatically motivated inventions tend to be ad hoc and require analysis and synthesis before they can be disseminated and reused.

  3. Generalization
  4. There is a crucial role for researchers in academia and industry for follow-up. There is no expolitatin of a bright idea without understanding, generalizing and disseminating the inventions and discovering their essential principles. Since academics teach they need to perform this function naturally. One does not teach effectively by only citing and describing examples.

    Important generalization that have opened up problem to scientific solutions are language and parsing theories, algebras for query transformations, proofs of correctness of concurrency protocols, and generalizations of recovery mechanisms. Once academics have the principles and ideas under control, they are good at disseminating them through teaching and writing. However, this traditional process is slow (15 years is typical). Means of speedup to exploit the understanding gained are a concern here.

  5. Specialization
  6. Since many general solution are hard and costly, effective exploitation of the principles typically requires some engineering insights and further specialization. Here academia and industry can interact effectively. A good general model will also elucidate the effect of constraints, narrowing of the domain, and scaling down of the number of instances encountered in a problem. A common type of a thesis will describe solutions that can be achieved given some assumptions about the problem. If those assumptions are realistic such a thesis can be of great value to industry. However, industry may not be ready. We now come to the crux of the technology transfer problem.

Paths for Technology Transfer

In an academic research setting most innovation is associated with a student and a thesis. When the thesis is complete the student leaves, but publication and dissemination often takes years. Even if new electronic services for rapid review and self-publishing speed up this phase, time still elapses before new concepts are appreciated and industrial interest ensues. If a thesis produces an isolatable result the student may become an entrepreneur, but to be successful many new concepts have to be integrated into larger settings. It is generally wise to plan academic thesis research so that it is not competitive with industry. Targeting one's research with an precise trajectory for industrial implementation is also risky. First of all, focusing on a specific recognized industrial need is likely to compete with convenient narrow specific solutions and reduces the intellectual and educational content of the thesis. Secondly, especially if a broader topic is addressed it is easy to fall short of the goal either in scope or in time and be bypassed before the thesis work is completed.

In industrial laboratories less research is performed than in earlier days, and there is much stress on development. It would be good if industrial researchers and developers would have good access to all types of innovative research results, but even though nearly all results are published and available on on-line networks they are very hard to find. Reasons for difficulty of access are both the volume and the marginal refereeing with respect to industrial value of what is being published, and the use of new and often excessively innovative terms to distinguish one's work from that of predecessors. Most initial contacts leading to TT are in fact personal, made at workshops and meetings, typically during informal discussions.

Many meetings that do not have TT as their objective do serve as TT venues, for instance review meetings where academic and industrial participants assess new proposals, and in the process learn from each other about what is needed and what is already available. Much of the success that MITI can claim in TT is due to their frequent and lengthy meetings to evaluate proposals that will anyhow be funded. It to industry's credit that in Japan industrial line managers participate in these discussions [Ref JTEC study 1991].

Today the pressures and time schedules many government labs are similar to the pressures found in industry. However, the remoteness from industrial delivery pressures makes performing in a product role difficult. In addition, there is much oversight and a great deal of time is spent in writing reports to justify ones' existence.

Asynchrony of Innovation Availability and Exploitation

The industrial need for innovation is controlled primarily by schedules and market forces which are unrelated to research schedules. When the need for innovation in development becomes clear, the sources to be exploited are hard to access. The student and the advisor are already involved in other enterprises, and the industrial researcher has been pulled off to another project, or else left in frustration. Although we can assume that relevant information will be available on some digital library somewhere, it is unlikely that the terms used to express a need will be the same as the terms used to identify the research result, for instance research results supporting 'multi-attribute search for information' was published as 'partial match retrieval' and ignored for some time. Especially in fast-moving fields these terminological gaps are prevalent. Without people able to span those gaps adoption of the innovation, even if recognized, will be awkward.

A Need for Transition Agents

There is hence a need for intermediate organizations to be the initial receptors, normalizers, and maintainers of research results. Vic Reis, when director at ARPA, recognized that many research results will not be of immediate use and will rest on shelves until needed However, no overt provision was being made in Vic Reis' model to establish the shelves, making delayed use unlikely.

Not only passive shelves are needed, a potential adopter needs people to talk to, get explanations, assess the status, and feedback on transition potential. Most research results also warrant some level of maintenance. For instance, demonstrations age rapidly as equipment changes. As standards become established the utility of many prototypes can be enhanced by adaptation to standard interfaces, often simplifying the product in that process. As the terminology of a field normalizes, descriptions and keywords used for search may be updated to enhance access.

Within the ARPA setting for instance, the Intelligent Integration of Information (I3) program funds a base level TTA effort at ISX Corp., a small ARPA-oriented contractor. The role assigned to ISX is to help assemble research results from the various participants in the program, adapt them to the interoperation conventions being developed, support interactions with emerging standards efforts, demonstrate capabilities to potential user organizations, and prepare business plans for technology insertion if requested. The expense, at about 1 person-year/year is well worth it in terms of reduced confusion and rapid availability of resources. A major benefit is, of course, having someone to talk to when I3 related problems arise.

A person who is involved both with researchers and industry can bridge the terminological gaps better than individuals on either side of the fence. All other program participants are (or should be) aware of the role that ISX plays. The ISX company, of course, also benefits by having an early handle on opportunities for further work, including tasks outside of DoD. An increasing fraction of informative interactions leading to TT can take place on the Internet, although initial resolution of terminological differences will require human mediation. The people at the TTA are to be quite capable, willing to gain insight into the prototypes and products they are supporting, maintain awareness of the changing infrastructure of computing platforms, networks, services, and interface standards, and at the same time be able to understand the needs of the recipients and honestly point out which of their wishes and expectations will be satisfied and which are best deferred into yet another timeframe.

Helping in Setting Research Directions

The structure established for TTAs should also support TT in the opposite direction. If the TTA personnel can abstract the needs voiced by industry into terms and concepts that are understandable by researchers, they can help focus research on topics of eventual interest. When industry rejects research results because they duplicate products already available or in advanced stages of development the TTA gains high-value knowledge, that normally would never be transmitted to researchers. The TTA will also, in time, understand the end-users needs in depth. Without deep understanding their is a danger to solve problems by applying instant 'hack' solutions, leading to worse problems later. Without guidance from industry many researchers are left to wallow in problem spaces of their own imagination. This direction is not meant to disparage 'curiosity-driven research'. If a researcher is truly curious to gain some new insight, that is marvelous. But many researchers would just as soon work on foundational research on which substantial industrial structures can eventually be built.

Who are Candidates to be Transition Agents (TTAs)?

There are quite a number of industrial research groups, both profit and non-profit, who seem well able to take on the role of a TTA. We consider here academia, industry, and goverment labs.

  1. Academia
  2. We have argued already that it is naive to expect academics to perform direct-to-industry TT. Most useful acedemic projects culminate in a PhD thesis and a demo. It is highly unlikely that industry is waiting at the end of the semester to reap the results. If the problem addressed was recognized earlier by industry, they will already have people building (pragmatic) solutions. If the problem is not yet recognized it will take time and multiple publications for the issues to excite industry. When industry is ready the student has graduated, and is working on something else somewhere else. Even if the professor ever knew how to run the demo, it will not work 3 months later because the supporting systems have been `improved'. It is awkward for academic instition to hire substantial staff for this role, since funding and research directions will fluctuate rapidly. Furthermore, the quality of staff retained in academia will vary since long term career path for non-teaching staff in universities are limited.

  3. Industry
  4. In the I3 case, a small commercial company was employed. However, the long-term viability of such a TTA has not yet been proven, and is likely that as such a company grows it may become less nimble and service oriented than it was initially.

    Large industrial organization are be defintion driven by their own needs, and must give priority to their internal projects over external TT. We have seen many examples where joint industry-academic proposals named excellent industrial staff, but during execution such staff was moved to internal projects and was replaced with equally well-pid, but non-productive participants.

  5. Government Labs
  6. Many government labs, as the National Institute for Standards and Technology (NIST), have mission statements which encompass the tasks outlined for TTAs. A governmental organization is by its very nature more stable than either academia or modern industry. Investment in advancing broad industry needs is justified, and the tasks of integration, requiring establishing and validating standards are part of NIST's perceived mission. Being a transition agent is also unlikely to be viewed by industry as undue interference, and the aspects of industrial policy are minor. Academics should appreciate any help in providing a TT path for their work.

The change in manpower supply in computing (much greater) makes the implementation of such a role for NIST's computing laboratories more feasible than it would have been, say, 5 years ago. Today many graduates would enjoy having the chance to come to a post-doc position at NIST where they could demonstrate, enhance, and package their work. Such an effort will benefit with interaction of NIST permanent staff who can convey industrial insights often lacking in academia. At the same time an influx of young rotators can enhance the staff awareness of the changing world and technology outside of the government. They are likely to put pressure on the environment that would discourage going along with old, comfortable systems. Those postdocs that eventually obtain academic positions will also bring valuable experience and awareness of terminology used from their industrial contacts to their future students.

No matter if the TTA is in industry or government, is profit or non-profit, it will be crucial that management defines the role and the criteria for success. The TTA model presented above can only work if the reward system is appropriate. If promotions and status depends on counting papers published, or on transitioning ones own research to industry, then the required functions for aiding TT will be abrogated. We believe that the need for the TTA role is sufficiently crucial that it behooves laboratories that can perform in this role to assess how to implement these functions.


The change that is required to go the model of technology transfer I am proposing here can be a wrenching change to methods that have been established elsewhere. A discussion that analyzes both Japaneses efforts, spearheaded by MITI, and European efforts, led by Esprit, could clarify these issues further. In the meantime, we have to consider the communities that may provide barriers to change in TT policy in our own environment:

  1. Academics, since they fear losing money (and power), and many actually believe they know how to do Technology Transfer.
  2. industry, who would like to get `useful', i.e., saleable stuff directly and cheaply from universities.
  3. NSF managers, who are pressured to show results from their funding of universities, without having the funds to support transfer agents in addition to researchers and their students.
  4. Current managers at governmental laboratories, who recognize weaknesses in their research portofolio, and want to upgrade their laboratories to do `better' research and gain stature.


The issue of Technology Transfer in our work is crucial, and absorbs much of our research resources, likely more than 50%. Given its importance it is useful to devote some rational discussion and analysis to this topic, since its metrics are today mainly guesses and hopes. If it were science, we would insist on hypotheses and proofs before devoting so many millions to its development and achievement. Let's at least bring the alternatives on a scientific table beyond committee meetings and coffee breaks.