Big Data needs architecture

SD models offer a rigorous enterprise architecture for firms’ data. Pharmaceuticals firms are  blessed with huge amounts of data on just about everything. But a competitive war-gaming project just completed for one of the big players was surprising. Sure, we could get data on things needed to model the firm and its rivals, but this seemed to come from lots of different sources, in many different forms – mostly big spreadsheets and slide-decks.

The data was inconsistent in units and time-scales – some was weekly going back >5 years, some monthly for the last 2, 3 or 4 years and others [especially the financials] quarterly or annual [last year: this year: %+/- ]. The data also lacked any sense of “what causes what” – What has the varying number of sales reps calling on doctors over the last 4 years done to the number of doctors starting to, or ceasing to, prescribe our product? … This raises 2 questions:

  • Just how much value are firms going to get from mining their Big Data if it has these problems?
  • How much better could they do if their data-collection and analysis was informed by the rigorous structure of a properly-built system dynamics model?

We might have allies in the IT world where “enterprise architectures” are seen as an essential foundation for planning an organisation’s information systems. But EAs seem to focus on the processes that make the enterprise function – how a new customer is taken on, how service calls are delivered, how the hiring process works. An SD architecture in contrast focuses on the “stuff” itself – how many customers, staff, service calls etc – and the causal interdependencies between them.  I am not for a moment suggesting this as a replacement for a process-oriented EA, but perhaps a complement?

Leave a Reply

Your email address will not be published. Required fields are marked *