An analysis of the use of digital collections in a scientific research library network : Part two of a case study from CSIC , Madrid , Spain

s 26,038 45,163 2002 Journals % out of total No. of % use of Range of articles per range of journals articles ScienceDirect More than 3,000 articles 3 0.21 11,294 4.31 Between 3,000 and 2,000 6 0.43 14,370 5.48 Between 1,999 and 1,500 18 1.28 30,346 11.58 Between 1,499 and 1,000 35 2.48 42,272 16.13 Between 999 and 500 75 5.32 54,613 20.84 Between 499 and 250 148 10.50 51,594 19.69 Between 249 and 50 379 26.87 46,094 17.59 Between 49 and 10 400 28.37 9,987 3.81 Fewer than10 346 24.54 1,429 0.57 Total 1,410 100.0 261,999 100.0 Figure 6. Downloaded articles 2002 An analysis of the use of digital collections at CSIC Agnès Ponsati and Mercedes Baquero Serials – 18(1), March 2005 56 Range of articles % out of total of journals 2002 % use of ScienceDirect 2002 More than 3,000 articles 0.21 4.31 Between 3,000 and 2,000 0.43 5.48 Between 1,999 and 1,500 1.28 11.58 Between 1,499 and 1,000 2.48 16.13 Between 999 and 500 5.32 20.84 Between 499 and 250 10.50 19.69 Between 249 and 50 26.87 17.59 Between 49 and 10 28.37 3.81 9 or fewer 24.54 0.57 Total 100.00 100.00 Figure 7. Percentage of journals and articles out of total downloaded in 2002 2003 % out of total No. of articles % use of Range of articles Journals of journals per range ScienceDirect More than 5,000 articles 5 0.3 32,992 7.1 Between 4,999 and 3,000 13 0.8 45,329 9.7 Between 2,999 and 2,000 32 1.9 74,832 16.0 Between 1,999 and 1,000 71 4.1 100,599 21.5 Between 999 and 500 131 7.6 93,886 20.1 Between 499 and 250 167 9.7 60,129 12.9 Between 249 and 50 403 23.3 48,779 10.4 Between 49 and 10 377 21.8 9,387 2.0 9 or fewer 529 30.6 1,072 0.2 Total 1,728 100.0 467,005 100.0 Figure 8. Downloaded articles 2003 If we look at a graph with only these two parameters (percentage of journals and percentage of articles out of the total of those downloaded), we can see how the columns in the two tints rise and fall in opposite directions:


Background and history of compiling usage statistics at CSIC
IDEAL from Academic Press was the first electronic journal platform which was available to the CSIC community in a trial.This took place from October to November 2000.The web page, which provided access to the trial from the Library Network Server, included a brief questionnaire.This first survey showed that the users were familiar with electronic information resources and that they took a favourable view of electronic versions gradually replacing the print collection in their centres.
It also showed attitudes and 'desiderata' which have remained prevalent over the last three years: ■ What is on offer is fine but new titles and publishers were always suggested.
■ Librarians are always more conservative with print than the users and want to guarantee physical copies for the future.
■ Users identified the changes brought by these electronic journals both in their work and in the library environment: it would be necessary to invest in computer equipment; it would decrease the demand for interlibrary lending; small research centres would benefit the most from the centralized acquisition of packages.Furthermore, four years ago they demanded an integrated, transparent access point to the digital collection.
In 2002, ScienceDirect (Elsevier) and SpringerLink were contracted as well as IDEAL.There were free trials from Kluwer Online and Wiley InterScience for two to three months to allow users to experience and evaluate them.The results of these surveys confirmed the interest in these products.Electronic journals have advantages over printed ones when it comes to finding out the real use that is made of each title.This has enabled CSIC librarians to be able to justify the investment they demanded in order to make the CSIC Network more effective by purchasing digital information resources.With print, it is not possible to measure use both in terms of quantity and also what type of use a user has made.This obstacle has been overcome with electronic journals and we have accurate and contrastable information about which journals are used and which are not, what the core of the collection is and what can be dispensed with.

An analysis of the use of digital collections in a scientific research library network: Part two of a case study from CSIC, Madrid, Spain
The introduction of electronic journals into Consejo Superior de Investigaciones Científicas' (CSIC's) collections has been gradual and constant.As the collections have been stable since 2002, the data which has been obtained from analysis of their use in the period 2002-2003 can be looked upon as giving a meaningful picture of developments during these years.The process for establishing the digital collection has been set out in our earlier paper. 1 Even though the objectives of this process were clear, it was also evident that it was necessary to count on the backing of the researchers as well as the financial and political authorities of the CSIC.As with any new initiative, it was crucial to undertake trials through which levels of usage could be evaluated and investment justified.

MERCEDES BAQUERO CSIC Libraries Co-ordination Unit Madrid, Spain
Nowadays we know more about the use of journals and the habits of their readers than ever before.The statistics on electronic journal use available from the publishers' own systems include numbers of accesses, the number of articles downloaded from each title, whether indexes or summaries are consulted, whether the texts are displayed in HTML or in PDF, etc.This knowledge has limited use in one respect -even though CSIC may know the titles which are not used we are not going to stop subscribing to them as the final cost does not change.However, these statistics have confirmed that what we have been buying in print is also what our researchers consult most often in the electronic version, which is to say, we have a very relevant collection.This is something which, up to now, we had expected but which we could not prove.Moreover, we can see that there is a large number of titles never subscribed to on paper which have been highly successful with researchers and this justifies the initial decision not to acquire only online versions of subscriptions which already exist.
At the end of 2002, we drew up statistics on use of the platforms, which had by then been available for the 12 months of the year (SpringerLink, ScienceDirect and Academic Press, the latter having been taken over by Elsevier half-way through the year).In addition, statistics from other products, which were used over shorter periods (WileyInterscience and Kluwer Online, or PCI Full Text), were included.We did the same at the close of 2003 so that it was possible to observe the developments in use of these platforms over two years.(Other new products -AIP, APS, Blackwell Synergy, MUSE, Nature, IOP -were contracted for 2004.A first assessment of the statistics, made in April 2004, shows that the levels of use of those services acquired in the previous years is being maintained and that Blackwell Synergy is receiving the most use.)

Characteristics of the statistics on use of the electronic journals
In 2002, it was extremely difficult to produce comparable statistics as there was no uniformity in the terminology or usage measurements publishers had adopted.The advent of the COUNTER project, and the adoption of its recommendations on which measurements of use publishers should make available, have altered this situation significantly.
The only constant element was the 'articles downloaded as complete texts'.This data was common to all the publishers and was the one taken as the key measurement in this study.When the editors provided more information (relation of session articles downloaded, localization by browse or by direct search, number of personal profiles registered, etc.) this was also processed but it has to be accepted that few publishers provided this type of detail in 2002.
For the CSIC -due to its organizational structure and the nature of our electronic journal contracts -there are some shortcomings for us in the statistical reports provided by the publishers, which are perhaps not as important for other institutions.Moreover, those of us working with this statistical information are not in contact with the end users and it is impossible for us to have firsthand knowledge about the more than 4,000 electronic journals that we have made available to our users.In our opinion the statistical reports provided by publishers have the following deficiencies for us: ■ It is usual for publishers' reports to provide data on global use rather than by individual CSIC centre.Each CSIC centre requires information specific to themselves, which we cannot give to them.When the publishers do in fact give this data by centre, it is not usually very definite as they mix our data with that of the universities and interpret accesses coming from the authentication systems erroneously.
■ Reports are made on lists of journals which are not always those that CSIC subscribes to.On occasions the publishers offer their complete list to report the number of downloads.But if a title had '0' downloads it may be that it is not a CSIC subscription and is not accessible for our users -it does not mean that they are not interested in it.Also, on certain occasions, the lists which are sent monthly or covering threemonthly periods vary in the number of journals they are giving information on, which makes it very difficult to match up the data title to title for the whole year.
■ Little detail is given on rejected accesses.We would like to know which journals that are not subscribed to are required by the users or which years (retrospective) are requested but we do not have.When a complete package has not been contracted this data is very useful and we miss having it.
■ It would also be appreciated if these lists of downloaded titles/articles included the data on coverage of the journal in such a way that it would be easy to know if it was a little-used title because it closed some time ago and so its interest for the reader has diminished or if it has split or continued through others which do in fact register use.
■ Additionally, it would also be very useful for publishers to provide reports that give information on the years for which searches are made for each title, to indicate whether the title is currently available or ceased publication and on the demand on the retrospective archive of each title.

Searching habits
We consider that the module of use reports from Elsevier provides a greater quantity of information for the client.For this reason and because this platform is most used in the CSIC, we have drawn up a profile of the habits of CSIC researchers from the reports obtained from this module of ScienceDirect: (a)  It is very interesting that some of the journals which are in the leading positions by number of complete articles downloaded hardly had any abstracts consulted.Conversely, there are titles which are in important positions in the consultation of summaries or abstracts but do not have the same position when it comes to ordering by articles downloaded.

Analysis of use of the collection
The ScienceDirect usage statistics module (now in compliance with COUNTER) offers an alphabetical list of titles available with the number of articles which have been downloaded throughout the year.We have organized this list by quantity of articles downloaded, so obtaining a series of data, which is shown in the following tables and graphs where we compare the developments in use in the years 2002 and 2003.If we look at a graph with only these two parameters (percentage of journals and percentage of articles out of the total of those downloaded), we can see how the columns in the two tints rise and fall in opposite directions: The first figures in these tables show that in both 2002 and 2003 they complied with a traditional prediction in the evaluation of collections, which had been drawn up for journals on paper, but which is also repeated for the electronic collections: 80% of use is concentrated on 20% of journals.
In 2002, the first six ranges of journals (up to the parameter 499-250 articles downloaded) represent 78.03% of use out of 20.22% of the journals available.In 2003 these same 6 ranges were 87.40% of use out of 24.4% of the total of journals.
There is clearly a core of about 50 journals in this ScienceDirect service which are essential for CSIC researchers -titles which repeat their position in the internal journal ranking which we have drawn up over the two years studied.
We also see that during these two years there are still 50% of the titles (those from which less than 49 articles have been downloaded) which we could say have little or no interest for CSIC researchers.This is a very high percentage, which shows that it is necessary to reconsider the titles which are being acquired.However, so as not to be so negative, let us remember that abstracts or summaries have been consulted from all but 13 of the journals, although they have not been interesting enough to download a complete article.We have also realized that the titles which show '0' use (291 in 2003)

Comparison with subscriptions on paper
The CSIC subscribes -in a centralized way through the procedure of public tenders -to 415 printed journal titles from Elsevier.(The number of subscriptions is higher due to duplicate subscriptions.)The comparison between the journals which are bought on paper and this internal ranking of the most frequently used titles illustrates the relevance (or not) of the titles which the CSIC has been subscribing to for years in the printed version.
In line with the figures shown above -that 80% of use is from 20% of the collection -we see that in 2003, 87.40% of use (24.4% of the collection) is borne by 419 journals.Of these, CSIC subscribes to 252 on paper and does not subscribe to 167.This means that the CSIC has subscribed, on paper, to 60.14% of the journals that interest it most.And also, thanks to the acquisition of the full ScienceDirect service, researchers consult nearly 40% of the journals they did not have before.If we limit ourselves to the first ranges the percentage relevance of the collection on paper is much higher; there are 121 journals from which more than 1,000 articles have been downloaded (54.40% of the total use of ScienceDirect).Of these the CSIC subscribes to 115 on paper, that is to say, more than 95% of the titles which hold interest.
At the end of this list there are 529 journals (30.6% of the total of ScienceDirect) with insignificant use: fewer than nine articles have been downloaded.They represent 0.2% of the demand on Elsevier.Out of these 346 journals, 20 are subscribed to on paper.
The drop in interlibrary loan in the CSIC Library Network has to be looked upon in relation to the existence of all these available journals (not only in Elsevier; the percentages noted are similar in the other publishers) in electronic version which have not been in our libraries before.
The balance of the relevance of the collection on paper, in the light of this data, is quite positive.It could be concluded that for decades the CSIC has been subscribing to the titles from Elsevier which are most requested by its researchers.The next step should be to co-ordinate, in the most rational and economical way possible, the coexistence of this print collection with the digital collection.
But it must be taken into account that two years is still a very short time to establish trends in use.However, there is one thing that is clearly shown and that is that electronic journals are perfectly implanted in the community of CSIC users.

Range of articles
Figure 3.

Figure 9 .
Figure 9. Percentage of journals and articles out of total downloaded in 2003

data on use: articles downloaded in 2002/2003 in the CSIC by publishers: PCI-FT 2002 2003
Serials-18(1), March 2005 Agnès Ponsati and Mercedes Baquero An analysis of the use of digital collections at CSIC 53 Quantitative

Serials -18(1), March 2005 56 Range of articles % out of total of journals 2002 % use of ScienceDirect 2002
Figure 6.Downloaded articles 2002An analysis of the use of digital collections at CSICAgnès Ponsati and Mercedes Baquero Figure 7. Percentage of journals and articles out of total downloaded in 2002

2003 % out of total No. of articles % use of Range of articles Journals of journals per range ScienceDirect
Figure 8. Downloaded articles 2003 are subject to various criteria: