Gaining business intelligence from user data
Posted by benwynne2 on 29 July, 2010
I attended a JISC workshop on use of user activity data on 14th July. The purpose of the workshop was to present some recent JISC funded and other work in this area and to consult on what additional work JISC might usefully fund in the future.
The presentations and discussion largely focussed on user activity data gathered by library management systems such as search histories and circulation data. Other types of activity data would include usage statistics for electronic journals at both title and article level and activity data generated within VLEs.
So, what might you want to use such activity data for?
Well, as the title of the workshop suggests, one use is to gain ‘business intelligence’ of different kinds. So, in a library context, to see what resources are being used and by whom. This is, of course, very topical in a challenging economic environment where difficult decisions need to be made about what to spend money on and what not.
Another possible use, is to use activity data to enhance the ‘user experience’ in different ways. The best known example of this is probably Amazon. So, ‘customers who bought this item, also bought these items’. In a library context, this could become, ‘people who borrowed this item, also borrowed these items’.
The best known (only?) university library in the UK which has been using activity data in this sort of way is that at Huddersfield – and the workshop included a presentation from Huddersfield’s Library Systems Manager, Dave Pattern. This illustrated that borrowing of unique titles has increased at Huddersfield since the ‘recommender’ features were introduced – suggesting that students are benefitting from a new means of locating related resources.
The JISC funded MOSAIC project has been exploring how user activity data can be extracted from library management systems and combined with data from student record systems to provide recommendations along the lines of ‘Economics students who borrowed this item, also borrowed these items’.
And Ex Libris now have a recommender system called Bx which uses activity data taken from open URL resolver log files.
Issues which arose during the discussion included:
Why haven’t more libraries done what Huddersfield has done? Possible reasons touched upon included lack of data (some systems do not log the necessary activity data), limited access to technical skills, competing priorities, concerns about data protection and other legal issues.
Why would a senior manager commit time and resources to exploiting user activity data? Business drivers might include usage analysis to demonstrate/assess value for money; improving students’ ability to find relevant resources, therefore enhancing the student experience and improving performance and retention rates.
Who ‘owns’ user activity data and who should manage it? Issues here include user consent, trust and the purposes to which data is put.
One thing which the workshop underlined for me is that there are many kinds of user activity data which can be used for many different purposes. This diversity of data and potential purposes can complicate the discussion at times.
Discussion suggested that what may be needed now is some practical use cases, addressing practical needs which provide more evidence of practical benefits in an HE context. And that’s on the recommender system side of things.
On the business intelligence side of things, the case for committing staff time and resources to analysis of user activity data seems easier to me to make at the institutional level in current circumstances.
Last but not least, JISC is currently funding development of a Usage Statistics Portal – which will provide libraries with a mean of accessing e-journal usage data in one place. That certainly is addressing the ‘business intelligence’ aspect of user data.