Metalib and statistics


Metalib, one of the core components of Primo, isn’t working for NTNU Library; here’s the statistics:

NTNU-Metalib total sources searched:  1440

Consider that this is not 1440 “federated-search instances”, but 1440 individual sources searched using the federated-search interface. Metalib allows up to ten sources to be searched simultaneously; therefore, the worst scenario is that Metalib’s federated-search interface has been used 144 times in the last year.

And what were people searching?

20.6% Google Scholar
20.4% BIBSYS library catalogue
5.2% ISI web of science
4.1% SpringerLink (MetaPress)
4.1% Wiley Interscience Journals
3.9% JSTOR Complete
3.5% Oxford Journals
3% (Ingenta)
3% PubMed
2.6% ECO – Electronic Collections Online

OK, so Metalib isn’t just about federated search, it’s also a link server (or a collection of links to databases if you will). The statistics here are more encouraging, gathering around 20–25,000 clickthroughs to native interfaces per month.

A cost-effectiveness study would show that a simple link server isn’t worth paying so much money for — there are better ways of doing this kind of work.

Tags: , ,

3 Responses to “Metalib and statistics”

  1. Larry Donahue Says:

    That’s very interesting information.

    I work for a competitor of Metalib in the federated search space, Deep Web Technologies (, and would love to hear your thoughts on why you think the usage is so low.

    Is the federated search option easy to reach, visible, etc, for your users? Does the 1440 represent initiated searches or completed searches (as in are people not wanting to wait, and interrupt the search process)? Is it because 10 sources at a time aren’t enough? Is it because your users don’t like the experience, or otherwise feel the sources aren’t relevant enough to their research?

    With our customers, usage appears to be directly correlated to easy to reach the search input box. It’s been rare, but in those instances where usage has been low, we’ve attributed it to an incorrect mix of sources.

    Our technology is different from Metalib in many respects. While there are a number, three major differences that come to mind to help increase user satisfaction with the federated search are: No limit on simultaneous source searches, incremental results and outstanding ranking.

    We have some public implementations, if you’re interested in comparing technologies to assess whether that could be a factor:,, and

    Take care and again, thanks for sharing that information!


    • brinxmat Says:

      Hi, thanks for commenting!

      The usage is probably low due to a combination of the following factors:

      The federated search pane is fairly well hidden on our site
      The local setup of the federated search system in terms of sources is not particularly good
      The user interface and search options of this particular implementation of FS is not good
      User experience of FS is likely to be damaged by exposure to this product, and it will therefore not inspire users to return

      I think that you’re probably right in many aspects of what you say, though I’m unsure whether users are really that impatient. The mix of sources is important, but I also think that Norwegian educational institutions are a special case — there is a heavy dependency at lower levels (BA) on obligatory course readings (i.e. an exam question cannot typically require any knowledge beyond the scope of the course readings, and there is little incentive to find additional literature for term papers because there is a standard perception that obligatory readings are enough). BA students will typically be the users that benefit from broad, general searches of the kind facilitated by FS, whereas more advanced users will benefit from the specialized interfaces provided by a subject-specific database.

      Another national service provider, BIBSYS, provides a very competent FS-system (BIBSYS Mime) that searches several hundred sources simultaneously and very rapidly. This system is largely ignored by users also.

      I think that the connectors for the kind of databases we’re trying to search with FS require so much tweaking that it is rather futile for small universities to even attempt this kind of work. Ranking and cluster-faceting really help (actually, I rather prefer the latter, but then again I’m not a user :D), and Vivisimo really improves the usability of Metalib, however, I don’t think our users wait long enough to see the clusters and facets.

  2. Primo-BIBSYS redux « Infonatives Says:

    […] of its core components is Metalib, which is not a particularly good system (but then again, which federated search system for interdisciplinary research […]

Comments are closed.

%d bloggers like this: