Skip to main content

Posts

Showing posts from August, 2016

Costs of doing BI the hard way

I am preparing a business case to justify the building of a data warehouse for the University.  This has some challenges.  While everyone acknowledges that our current BI reports need improved, it is far from obvious how to measure the benefits of improving our BI.  Suppose we our student satisfaction score increases in two year's time: how much of that would be due to which specific initiatives, and which of those would result from decisions made with better BI data?  It's a tenuous thread of causality. Nonetheless, if we believe that the decisions made by staff have some impact on outcomes, and that by having better information available to them they will make better decisions, it follows that a successful data warehouse project will have a positive effect.  Even if the impact on University income is one tenth of one percent across the board, that would quickly repay the cost of developing and running the service. Another approach is to look at the costs of not having an

BI advice from Southampton Solent University

We were fortunate to have a visit from Neil Randall of Southampton Solent University last month.  Neil and his colleague Paul Colbran gave an excellent presentation at this year’s UCISA conference about their experience of setting up an effective BI service.  I invited Neil to visit Edinburgh to explain their approach to our team and to review our proposed BI architecture. We began the day with Neil reprising his part of the UCISA presentation and discussing several points arising.  We presented our draft architecture, and then we discussed topics including how to structure and manage a BI service, which ETL tools to use, how best to model data, and how to integrate a data warehouse with relationship management (CRM) software.  We had a very information conversation about “Extract, Transform, Load” (ETL) tools, which load data from source systems into a data warehouse.  Neil recommended we look at file-based tools rather than database-oriented tools.  Without this advice, we probably