This chapter deals with the organizational and economic questions raised as the second objective of the TULIP project (chapter I 2). As many organizational, and to some extent economic (cost), issues have been discussed as part of the previous chapters, specifically chapter II, this chapter reflects the suggestions/conclusions/“lessons learned” contributed by participants in interviews and reports. Some 35 interviews were held in October/November 1995 with project managers, project liaisons, administrators, head librarians, (reference) librarians, and technical management and staff at the nine participating universities.

V 1. Critical success factors for the implementation of TULIP

From the interviews conducted with key players at the TULIP universities, we have distilled four major factors which can make or break a project like TULIP: the (project) organization, understanding your user community’s possibilities and needs, infrastructure and promotion-related factors.

V 1.1. Organization and project management

First of all, dedicated project management, cooperation among the parties involved and having the right resources available at the right time seem to be crucial to getting a large scale project off the ground. For instance, at the University of Michigan there were several groups that had interest in the information infrastructure, and a common goal to move ahead. Certain areas had expertise that would contribute to the whole, but no one area had all expertise. So they had to work together, and by working together they moved the agenda forward quickly. For an implementation like TULIP both technical and user expertise are necessary. Working together with both technical and “contents” specialists is easier if common goals are set. In some cases, the systems staff and the librarians report to one person, facilitating setting priorities.

Almost all participants implemented TULIP in addition to their ongoing other tasks. To make this work, especially when priority projects come along, what is needed are very capable people who creatively manage their time. Like at MIT for example:“We couldn’t afford a large staff so we had to focus on getting the right people at the right time.” A member of the BULB group, the persons who worked on TULIP at MIT, said: “None of us have a lot of margin or excess staff or money lying around. So we had to think carefully about, in building the system, what were things that already existed, so that we didn’t reinvent the wheel.” During the first stage of the project they met weekly during lunch, then biweekly and during the last period once a month.

Another participant described it as follows: “Spelling out a common set of values, common visions of the future and the divisions of responsibility and the role of each organization”. Politics, lack of priority and lack of responsibilities can cause long delays. Problems of this kind have occurred at many of the TULIP universities, and all but killed the project in a few of them. Internal organizational/political problems can turn out to be very real stumbling blocks.

V 1.2. Comprehend end users’ possibilities and needs

Knowing your users is the next key element for a successful implementation. What happens if there is an interface, up and running in the library, accessible on the desktop of users, and it appears not to be what users really want? One of the main lessons learned at the University of Washington was that you have to make sure that the product you implement is appropriate for your user community. The task force for library services is now working to acquire more information and to work more closely with major users to find out what their needs are.

A few universities actually stated, that they felt in hindsight that the first step of the project should have been to consult and involve their end users. This was especially true for those who attended the focus groups, which they found very useful for their assessments of user needs.

“Users said it (TULIP) was useful, but that it was only a piece of what they need. They want it (electronic information) on their desktop; they want electronic access to the full text (not just to citations); local, fast printing; and they want it now.”

Not all users like electronic information, though. A remark heard many times was, that in general, the graduate and undergraduate students are computer literate and were the heavy TULIP users. Faculty are coming along, but it varies per discipline and per person: “the students are way ahead of them”. But, if users are to be provided with the advantages of electronic access, one must understand what is important to them.

V 1.3. Resources and infrastructure

If there is no adequate infrastructure in place, a project like TULIP can not be implemented, even with commited team members who know what their users want. There are three aspects to infrastructure here:

• 1.  systems on which TULIP runs;
• 2.  networking and (non-local) printing infrastructure;
• 3.  user desktop systems.

Re 1.

Universities where TULIP was linked to their existing systems had less of a daunting challenge than those where the system was developed from scratch, but most universities have stated they had underestimated the amount of time, effort, and to a lesser extent money, it would take them to get their TULIP system up and running. These experiences are well documented in chapter II. Some additional comments: “Libraries with less resources are advised to think twice before starting to make their own implementation: It won’t do, in particular for small libraries, to be responsible in a way the TULIP participants were. They don’t have the human resources and technology to do it. So they will have to turn to a vendor who can supply them with the right kind of clients to get at this information. They also won’t have the capability of storing the data locally.”

Re 2.

The ability of the (inter-)campus networks to deal with bandwidth-consuming graphical data is crucial. If you want to provide users with a system which enables them to access graphical information from their desktops, you need a high speed network able to transmit graphic information with acceptable response times. Although “acceptable” is a highly relative term here, dependent on the users’ expectations, there is a certain delay at which users will get frustrated and stop using the system. The printing infrastructure should also be ubiquitous, fast and easy.

In some cases network speeds and slow or difficult printing have been limiting factors. At Georgia Tech, this has been circumvented by implementing a service that did not offer desktop display but rather off-line printing. At the University of California however, where viewing was very slow and printing almost impossible for most campuses, this has been a major factor hindering success. Additionally, it led to an unwillingness on the part of some of the campus liaisons to promote a system which they knew did not work satisfactorily. The University of California will, on the basis of their experiences with TULIP, upgrade its inter-campus network to be able to deal with the increased requirements of image based information.

Re 3.

“For a successful implementation, make sure the product is appropriate to your user community.” Regarding the equipment on the desk of your users, there are two possibilities: either to provide high-end workstations (required for image viewing) to the users, like the Athena Project at MIT, or adapt your interface to a standard that can be used at the workstations your users have. However, hardware purchases are highly decentralized in many places, resulting in an amalgam of hardware configurations on users’ desks. Several universities now require students to have high-end PCs. However, this is still rare and it may take a while before, as someone said: “We’ll end up with computers being in the same class as text books and calculators, something you can build into a financial aid package, averaging out over four years with a mid-point upgrade in two years”. And even then the required hardware may vary from department to department. Furthermore, faculty and graduate students still have widely varying equipment.

The lowest common denominator then is a terminal emulation, which severely limits the possibilities of the application (no graphics at all), and therefore result in non-user friendly and uninteresting applications. With the advent of the Web, with clients for every platform, this situation has changed dramatically for the better, as you can now build a Web-based system which can be used by almost everyone, without having to write clients for all the platforms they use.

V 1.4. Introduction, promotion and training

At most universities, promotion was not viewed as a major issue at the beginning of the project. However, over the course of time this attitude changed. Some quotes:

• “Products are not used unless you promote them. Promotion is a key issue and easily overlooked”;
• “Promotion is absolutely necessary for electronic products. When a new database was installed, I put balloons above it to attract attention...” and
• “There is so much out there, people get confused. You have to show them that there are many more things besides books to help them get answers to their questions”.

Nevertheless, some librarians feel “a perfect product will sell itself. But if no one knows a service exists, no one will use it. Initial promotion is a minimal requirement”. At sites which did more extensive promotion we see a significantly higher use of TULIP. So, promotion does seem to be a condition to get your users interested and started. Continued promotion is obviously a necessity to attract new user groups. Additionally, the TULIP implementations were often far from perfect in the beginning. Therefore, continued promotional activities are needed to make users aware of improvements.

While general computer training, that is teaching the use of the basic tools needed for research, is seen as a requirement, many feel that separate training for a specific implementation should not be necessary. Users should be able to start using the implementation without specific knowledge, and discover the more advanced features as they start to need them. “We begin to observe that people will use systems to a point of satisfaction. No matter how whiz-bang a system is, they are probably not going beyond the point where it satisfies their needs. Only when they become dissatisfied with their current capabilities, are they going to move forward and learn more about the system’s possibilities and that is where the functionality begins to pay off.” The kind of people who use scientific journals are said to generally prefer to experiment themselves instead of attending training. They want to be able to do what they intuitively think is right and prefer not to read the documentation.

V 2. Economic issues

Of the objectives of the TULIP project (technical, organizational/economic, user behavior studies), the “economic issues” part of the second objective is the least conclusive. This is in part due to the fact that it simply proved impossible to test some of the economic models as we had set out to do, and partly due to the fact that the project was still so much in an experimental phase, struggling with technical and user behavior issues, that the discussions about economic issues did not really start until close to the end of the project.

However, although we have not been able to directly test many economic models, TULIP still helped us a great deal in the development of ideas about theses issues, specifically about cost. Below is a summary of the views of the participants on economic issues, both from interviews conducted with them and the university reports (appendices I-IX), and of initiatives which were taken to test economic models.

V 2.1. Cost

 V 2.1.1 Costs of implementing TULIP

All participants have been confronted with the harsh economic realities of building even a prototype of an electronic journals system, from Elsevier Science’s side the costs of conversion and distribution, on the universities’ side the costs of implementation of their respective systems. These experiences guide us in determining what we want to and can do in a full scale production. Some universities referred to high costs, especially of storage, as an important factor inhibiting the development of their TULIP systems, and view these costs as a major roadblock for further development. However, MIT reports:

“The cost of building the Page Image Delivery System (PIDS) involved staff costs and technical costs. We estimate that the aggregate effort to build the delivery architecture has been three FTE for two years. Because MIT’s infrastructure is already in place, hardware costs for TULIP consisted primarily of the server and storage, in total, approximately $100,000. This seems like a relatively inexpensive effort with high return on investment”.

V 2.1.2 Costs of storage

Storage costs have been a major factor in the TULIP project, as storage media had to be dedicated to the project, whereas use of people’s time, use of the server etc. could be shared with (many) other projects. When TULIP started, most university technologists minimized the storage costs. But, it seems that storage costs remain a very important cost element, even though the costs of the media keep dropping:

“....the current cost for magnetic storage is 5.3 times greater than print, and jukebox storage on CDS is 1.2 times greater. It is also worth considering that the typical book stack will last well over 30 years, but not the information technology. The $16,891 server and magnetic storage will have to be replaced at least every eight to ten years, and this is stretching the point. Based on a ten-year replacement cycle, digital storage and access will cost academic libraries 16 times as much as print to store locally. It seems unlikely that a persuasive case can be made for the added value of electronic access and retrieval without a dramatic change in the ownership concept. .......one final note on comparative capital expense - the cost of building will continue to increase at something like the CPI, and (if recent past is a guide) the cost of servers and storage will continue to fall at a more rapid rate. The problem is guessing correctly when these trend lines will cross and make digital storage cheaper than print”.

Additionally, the costs of managing large collections of electronic data must also not be underestimated. Consequently, Carnegie Mellon University argues:

“The traditional model of local ownership, which has dominated the vision of library organization and collection development for a century must change. The access model which is emerging will mean the libraries may subscribe or license access to information formerly packaged as a book or a journal, but it is not likely that they will store much of it on the local campus network. It only makes sense to share information technology resources among libraries and the cost of shared access to databases.....It also seems likely that the library vendor and publishing community will provide some forms of access......However, these are all new relationships. They mean that the nature of ownership must be carefully redefined and this will take time and it may not be very easy to accomplish. Publishers will want to know that their materials are being used appropriately. They will expect that access is for the campus community, that ILL and reserve reading conform to ’fair use’ and that authentication and authorization prevent the significant access to information for those who have not paid for it. Libraries will want to know that a subscription to a title gives them permanent access to the contents over time, that the server on which it is found will be consistently available, that the technology will be robust and stable, and that if the supplier (e.g., consortium, publisher, network) ever withdraws the service, then there is a plan for giving them the data they paid for. If such relationships are properly worked out, new types of subscribed access may be expected”.

V 2.1.3 Costs of content

While in the TULIP project access to the electronic journals was free on the basis of paper subscriptions, many universities expressed concern about the costs of the electronic journals after the experimental phase, for instance:

“Because the online versions of the TULIP journals received only light use, it is doubtful that we would have been ready to pay additional charges for the online versions at this time. The pressures on our science materials budgets are too great to justify such expenditures. We can no longer provide access to paper journals now needed. We would therefore need to consider very carefully purchasing ’added copies’ online of items also held in paper at the expense of materials needed in paper form. If the costs of the online versions were less than the paper versions, then there would be some incentive to purchase online versions as an alternative to the paper - but the current Elsevier pricing policy (i.e., online versions on their own being more expensive than the paper versions) does not make that possible”.

And:

“With regards to continuing to purchase Elsevier electronic content, we have elected to discontinue loading content in the present mode primarily because the cost is too high and comes at a time of shrinking library budgets. We are excited about electronic content, and look forward to exploring other economic models that are a better fit with our economic realities. MIT is willing to pay for high value information but is under tremendous pressure to spend fewer dollars as time goes on. For example, electronic-only distributions are particularly attractive since they eliminate postal, paper, and packaging costs”.

V 2.1.4 Costs of full scale digital libraries

“Too often, work on digital libraries, not to mention much theoretical discussion, proceeds without a thorough grounding in the realities of cost. There are certain assumptions which precede this state of affairs, among them the notion that digital libraries somehow will be cheaper than print libraries, perhaps even free. One suspects that this arises from the misplaced hope that digital libraries will liberate us from the difficult cost dynamics of print libraries. There is also a presumption that electronic access will mean added value to library patrons, but it begs the question if the access is at a cost patrons are unwilling to pay.”

TULIP proved to its participants beyond a doubt that building digital libraries will be a costly and lengthy process. The future is highly uncertain: is it a remote access, single computer serving the world or the continuation of the building of local libraries, with local patrons, spending local budgets? The easy and undoubtedly correct answer is to say that it is both. But for both libraries and publishers, this may be the most difficult to deal with, as we have feet in both worlds and cannot easily balance the economics of parallel systems.

“The current publishing environment and the cost analysis of the work at Carnegie Mellon does not support the notion that digital libraries are about to happen. The world of publishing is anything but monolithic....... In addition, the necessity to repeat capital expenditures on servers and storage will not encourage the development of digital libraries, even though hardware is becoming more of a commodity and storage costs may be trivial in the not-too-distant future.”

V 2.2. Economic models examined

The TULIP project did not provide definitive answers to the questions about economic issues surrounding the transition from a paper based situation to the electronic dissemination of (scientific) information. Testing of economic models has been limited, as described in some detail below.

V 2.2.1 Internal charging at universities

Some universities had plans to test some economic models on their campuses by charging for article printing. This proved to be impractical or unfeasible within the context of this project. For instance at the University of Michigan:

“From the project’s inception, the team anticipated gaining a base of knowledge about system building for electronic resources, user behavior with digital journals, and economic models applicable to the evolving hybrid print and digital environment. They were successful in meeting the first two objectives but were unable to test economic models. The rapid evolution of the World Wide Web as an information environment presented the developers with a better approach to TULIP than possible during its initial implementation. However, the Web does not yet support the kind of data collection that is necessary for testing economic models nor the level of security for financial transactions that we felt was necessary”.

V 2.2.2 Consortium model

Discussions had started between Elsevier Science and the University of Washington to refine a consortium model by working with NorthWestNet, but unfortunately that was not possible because of problems which Washington experienced with their implementation.

V 2.2.3 Subscriptions to electronic-only material

Elsevier Science wanted to explore a few possibilities to test specific economic models based on the subscription model, as both sides know that model and are comfortable with it. As the electronic versions of the subscribed-to TULIP journals were free, Elsevier Science offered the participating universities some additional options:

• Elsevier Science offered electronic copies of titles to which the participants did not have a paper subscription at a reduced rate from the beginning. There were few takers of this offer, partly because participants did not want to negate/counter collection development decisions, partly due to budget constraints, not only concerning the journal subscriptions, but also (and possibly even more so), because of additional storage/infrastructure costs;

• Elsevier Science also offered lower prices for a second group of titles for which again response was very limited, the major reason being budget constraints (cost of content as well as of storage). Other reasons included, that some universities perceived their systems to be incomplete (for instance the user interface), and did not want to invest in more content at that point. Also, as some universities were planning to discontinue providing access to the TULIP files after the experimental period, they did not want to extend the journal list.

In conclusion we can say that making additional funds available for electronic content will not be a trivial issue for these universities. As Carnegie Mellon University described it:

“In the first place, it seems clear that libraries will not have large amounts of new funding with which to purchase electronic materials, although it is not a zero-sum game. It follows that publishers may not expect to have large sources of new profits from the sales of electronic products that represent the scholarly information published today in books and journals”.

V 2.2.4 Pay per use - article delivery

Elsevier Science offered two article delivery related options:

• All participants were offered the opportunity from the beginning to use the files for article delivery to other non-participants if a royalty was collected; none of the participants actually implemented this. The reasons for this option not being taken up are somewhat unclear, however it seems that one important reason is that most universities’ document delivery demand stems mostly from ILL requests, which they are bound to fulfill at little or no cost to the receiving university on the basis of reciprocity. Although the costs of fulfilling these requests by hand are much higher than the royalties to be paid in the TULIP option, these royalties are “out of pocket” as opposed to the personnel cost involved with copying the original articles. Elsevier Science does not permit the electronic files to be used for ILL;

• Similarly, Elsevier Science encouraged universities to consider providing article delivery to their own constituents for titles not subscribed to (but included in the bibliographic information) by using Ei, in which case Elsevier would subsidize Ei’s handling fee and only royalties would be collected; again, none of the participants did. It seems that the reasons here are the technical difficulty to implement this, or in some cases rather the lack of priority to implement it, as well as the inability to charge through to end users in combination with budget constraints.

In conclusion: it seems to be quite difficult to move away from the status quo as far as article delivery is concerned, either because of “political differences” between publisher and universities, or because of the inadequate infrastructure provided by current technical and economic solutions.

V 2.2.5 Conclusions

There are no blueprints available of economic models in an electronic environment. In the TULIP project we took the first steps, addressing many of the issues - technical, organizational, user needs and preferences, and to some extent economic. Some of the practical problems with establishing and testing new economic models are mentioned above. However, towards the end of the project it also became clear that the expectations of (a number of) the universities and Elsevier Science on what kind of economic models would be tested were somewhat different. Discussions continue between individual universities and Elsevier Science, in the context of the Elsevier Electronic subscription (EES) program. A final comment from the University of Michigan report (appendix VI):

“A review of faculty and student comments and patterns of use suggests how important a better understanding of economic models is if the TULIP concept is to be expanded. Clearly, we continue in a transitional phase with some users reporting their inability to sustain additional subscription costs for personal copies of favored journals and with the library, too, forced to make choices because of budget constraints. However, the scientists and students report a desire to retain a paper version as well as the electronic files as long as possible. Certainly, we need to know what are feasible pricing options that will enable us to continue to support our researchers’ and students’ information needs”.

V 3. Conclusions and recommendations

The TULIP partners - universities and Elsevier Science - have learned many valuable lessons in the course of the project that will enable them to better face the long transition phase towards digital libraries. These lessons are shared in this report and the university reports with anyone facing similar challenges now or in the future. The TULIP project has been rewarding for all partners, even though not all objectives may have been met completely:

“Sustained communication among libraries and Elsevier Science as we probe electronic frontiers together has surely resulted in higher levels of understanding about the differing concerns that we bring to the arena of electronic journals. TULIP libraries have gained first-hand experience with a publisher’s perspective on the costs of changing from a traditionally accepted format to an experimental, rapidly changing electronic medium. Perhaps Elsevier Science has benefitted from working closely with libraries in the developmental process to discover values and constraints that inform library acquisitions decision-making.”

“Elsevier Science sponsored TULIP at a timely juncture for the future of scholarly research and publishing. TULIP represented a unique collaboration among publishers, librarians, academic faculty and researchers for exploring delivery of full text to the desktop....... At the time TULIP started, librarians and their clientele had had little exposure to scholarly electronic journals, and Gopher was an innovative development. With the advent of Web capabilities, we discovered a much more viable medium for image delivery. The fact that Elsevier Science journal tables of contents are now available via the Web suggests that this direction holds promise for the future”.

--------------------------------------------------------------------------------