IR costs

When it comes to IRs, the software profile of institutions that are members of the Association of Research Libraries (ARL) seems to differ from the profile of IRs worldwide. The evidence for this appears in the latest in ARL's SPEC Kit series (no. 292), INSTITUTIONAL REPOSITORIES. This is a subscription series, although copies of this edition can be bought individually.

The implications of the principal findings from the report, which can be found in the freely available executive summary (pdf 22pp), have been debated by Charles Bailey, one of the report's co-authors, Stevan Harnad and others (see Charles Bailey's original notice posted to the American-Scientist-Open-Access-Forum and follow the thread.) This blog entry tries to highlight key elements in the report as identified in the debate. First a few words of caution and qualification:

CB: "Although the Executive Summary is much longer than the typical one (over 5,100 words vs. about 1,500 words), it should not be mistaken for a highly analytic research article. Its goal was to try to describe the survey's main findings, which was quite challenging given the amount of survey data available.

"While the survey attempted to be as comprehensive as possible, its length had to be restricted. As it was, it was certainly one of the longest (if not the longest) survey of this type that ARL has ever done."

On to the findings, first the profile of IR software usage.

Software profile

ARL: "By a large majority, the most frequently used local IR software was DSpace, with DigitalCommons (or the bepress software it is based on) being the system of choice for vendor-hosted systems."

SH: The ROAR figures for total US OAI archives are:

DSpace: 55
EPrints: 52
Bepress: 44

The corresponding figures worldwide are:

EPrints: 210
DSpace: 167
Bepress: 53

or more specifically for IRs rather than OAI archives generally

dspace/bepress/eprints

ARL IRs: 23d/7b/0e
US total IRs: 36d/40b/33e
World IRs: 111d/47b/123e

Costs

It's not clear to what extent this software profile of ARL-based IRs affects costs, although David Goodman usefully dissected the costs quoted in the report: "planners of IRs should consider the summary data with care."

ARL: "For start-up, 67% of budgets fall below $75,000, 14% are $75,000 to $125,000, and 19% are $150,000 or greater. The maximum start-up budget ($1,800,000) is far greater than the next highest ($400,000) and is from an institution that included extensive software development and testing costs in its start-up budget. For ongoing budgets, there is a similar concentration at the ends of the ranges: 50% are below $50,000 and 50% are $100,000 or greater. The maximum ongoing budget ($500,000) is also much greater than the next highest ($300,000) and is reported by an institution that has a major role in a state-wide IR initiative."

DG: "They also give the median, which for those have implemented the repository is $45,000 start up (with a minimum value of $8,000) and $42,000 operating costs (with a minimum value of $8,600).

"If i were giving a quick summary, I would emphasize the low end, especially for the start up costs: It can be done for less than $10,000, even at a large research university."

CB: "Regarding the idea that IR start-up costs are in the $10,000 range,
4 out of 15 respondents with operational IRs reported costs below $25,000;
5 reported costs between $25,000 and $49,999;
2 reported costs between $50,000 and $74,999;
1 reported costs between $75,000 and $99,999;
and 3 reported costs greater than or equal to $150,000."

There was some concern over IR staffing costs, and whether these reflected actual costs or the way they were calculated

CB: "Once you get past the first unit, a wide range of library units become involved in IR support, not just core technical units. This may explain why respondents with operational IRs identify staffing and benefits as the top start-up and operational costs.

"One section of question 9 asked: "Please estimate the percentage of the budget allocated to each of the following categories."

"For respondents that had an operational IR, the mean results for start-up costs were:

Staffing and benefits: 63.3%
Hardware acquisition: 25.6%
Software acquisition: 23.0%
Hardware maintenance: 9.2%
Software maintenance: 6.0%
Vendor fees (if IR is hosted
by an external vendor): 70.2

"For respondents that had an operational IR, the mean results for ongoing operation costs were:

Staffing and benefits: 68.3%
Hardware acquisition: 23.3%
Software acquisition: 14.5%
Hardware maintenance: 10.3%
Software maintenance: 11.5%
Vendor fees (if IR is hosted
by an external vendor): 73.8%

"The staffing requirements issue is one that certainly deserves further study."

Purpose

Perhaps the profile can be explained by the purpose of the ARL IRs:

CB: "As to whether the main purpose of most of these projects was OA, question three asked "What motivated your institution to establish an IR? Check all that apply."

"The top four reasons for respondents that had operational IRs were:

Increase global visibility of institution’s scholarship: 97%
Preserve institution's scholarship: 95%
Provide free access to institution’s scholarship: 89%
Collect and organize institution’s scholarship in a single system: 89%

"The top five types of digital objects that respondents with operational IRs identified as being in their IRs were:

Electronic theses and dissertations: 67%
Articles, preprints: 61%
Articles, postprints (author modifies preprint to match
published work: 61%
Conference presentations: 50%
Technical reports: 50%

DG: "The main purpose of most of these projects was not OA. It was ETD (Electronic Theses and Dissertations.) (p.18) For three-quarters of the libraries this was the major type of deposit."

Impact of the report

Assessing the impact of the SPEC Kit on libraries without IRs, Heather Morrison supports the effect of the report: "The news from the ARL Spec Kit is, in my view, overwhelmingly positive for the open access archiving movement. Close to 80% of ARL libraries either already have, or are in the process of planning, institutional repositories. Reporting this trend via the SPEC Kits, highly regarded by academic library managers, is likely to increase this tendency."

Queensland University of Technology in Australia was among the first to adopt an institutional policy mandating deposit of papers in its IR. Paula Callan, the project officer for the repostitory, provided an update on the success of its strategies and an outline of costs in this list posting:

"QUT's repository was launched in January 2003 and we used a range of strategies to promote the service and to populate the repository. The most important strategy was the endorsement, at the highest level, of an Eprint Repository Policy

"However, getting a policy endorsed by the University Academic Board did not immediately translate into having the whole community aware of the policy and actually acting on it. Hence we also used various other strategies ...

Costs:
"The QUT eprint project got off the ground with a University "Initiative grant" that funded a full-time project officer position for 6 months and about A$3,000 for promotional materials and events.

"For 2 years after that, the project officer position was continued half-time; funded jointly by the Library and the Division of Information & Learning Support. It would have been impossible to fund the full costs of this project from the Library's normal operating budget.

"Now that our eprint repository has reached a critical mass of participating authors, it seems to have developed a momentum of its own. The others will hear about it from their colleagues and will not want to feel they are missing out on something good. It will still need ongoing promotion and support but our faculty librarians will now keep the ball rolling. Two library assistants have will now undertake most of the metadata reviewing/editing. This will incorporated with their other duties."

There appear to be some misunderstandings about institutional repositories and the need for more functionality. The phrase "run before you can walk" comes to mind in some cases, but this desire is perhaps predictable and not unreasonable.

The misunderstanding is to believe this functionality comes for free. Where the software is open source it remains free, of course, whatever extended functionality is supported, but to implement additional functionality comes at a price. Managing an IR costs money; to support more complex functions costs more money. This was illustrated quite graphically by a spreadsheet on the costs of setting up and maintaining Open Source repository software produced by Rebecca Kemp of Yale University, and shared on the liblicense list. Quite startlingly, as Rebecca commented: "There was a range, from about $6,886.62 for a set-up cost, all the way to over $1 million."

The EPrints examples tended to come at the lower end of the cost range, prompting us to produce this story, EPrints: "World's Best Practice for an institution commencing an institutional repository". The platform for our story was a brilliant article by Arthur Sale, whose phrase we used in the title. Arthur's emphasis for repository managers is the easiest and fastest route to set up an IR at lowest cost. On all three counts, according to Arthur, EPrints scores best.

Why would you choose anything else? Well, we're back to the quest for more functionality. EPrints has always had a focussed agenda, the support of open access. By emphasising ease and speed of set up and low cost as well, we seem to be suggesting limited software. It's not the case. Since it was the first and original IR software, EPrints has been in development the longest of all IR softwares. It has a hidden treasure chest of features to make the most of an IR. The skill of the developer has been to provide these without distracting from the core requirement of a low-cost IR for all.

Not all IR softwares are the same, but don't just look at the publicity materials to find the differences. Look at the origins of each one, study the development path and the motivations that have shaped it. Remember, IRs are not intended to be whole digital libraries and manage the business information of an institution all in one place, but to be components of a range of digital institutional services. In fact, the 'IR' may become a number of interacting repositories. If you are looking for a single package as the complete solution you are likely to be disappointed; or at least looking at the higher end and above of Rebecca Kemp's cost examples.

Instead, if you have more realistic and practical objectives, the key features to look for are fitness-for-purpose now and interoperability for the future. EPrints can give you what you need now, with expandability in terms of the the hidden features you can discover when you need them, and the ability to interface via Web services to other repository software that can form part of an emerging institutional information architecture.