Self-service BI/Analytics is the new black, it seems. And why not? Why should decision makers have to ask IT people for a new report they want, or to load a new dataset into the EDW so they can mash it with some existing stuff. The quicker they can get their answers, the quicker they can make decisions that will help the business, right? Well…
The trouble with mashups is that unless you understand, preferably in some detail, all of the datasets involved, it is very easy to produce results that are, well, wrong. Take this story published in a UK newspaper last year – Just 100 cod left in North Sea. The BBC has an explanation for how this article came about – North Sea cod: Is it true there are only 100 left?. In summary, they were talking about adult fish and they based the definition of “adult” on the typical lifetime of a cod, but they took that typical lifetime data from a different part of the world where cod typically live twice as long! This mistake, along with other misunderstandings, meant their final number was a factor of 1,000,000 out.
Admittedly this is an extreme example, but it shows that it is easy to make mistakes through misunderstanding the data you are manipulating. An example closer to home – I did some analysis of Service Management data in our ERP system by feeding it into Lumira Desktop. One of the things I did was produce a profile of spend throughout the year, based on the costs on the service order and the date on the notification. The resultant chart can be seen in this blog – First experience with SAP Lumira. We spend more in July than any other month, by quite a lot. Except we don’t. The dates came from the notifications, which are generated in advance, in July, but the work is spread over the whole summer. There is a three month peak, not a one month peak. Fortunately I was just experimenting with Lumira rather than trying to predict our cash flow… 🙂
Enabling decision makers to make decisions more quickly is a good thing. But only if they are good decisions, based on good data, surely? Are we requiring that our decision makers are all data analysts? If not, how are these misunderstandings avoided?