I was reading a post by BI Monkey and found myself in agreement with what he says. It also got me thinking about the wider implications of the problem of not “helping you make better decisions”.
I’d add to BI Monkey’s question with the specific question I always ask business analysts as they begin to list out the attributes they want added to a dimension in a requirements meeting. “What question does this answer?”. It’s a BI specialist’s responsibility not just to resolve the technical aspects of the requirements, but also to help steer the requirements so that something useful is delivered at the end.
I think this is one of the reasons why support from the business can fade. And without champion users in the wider business, it’s very likely that a BI implementation will be paralysed. People don’t understand the information provided and so don’t use it. Why keep funding the project if nobody uses it? Or the last (over-scoped) project took so long that the team isn’t trusted to take on new projects. Here are some scenarios…
The BI implementation provides lots of data, but not much information
As stated in BI Monkeys post, what is the point of all of these attributes and transaction records if your users can’t interpret them into useful answers. I have found that users have a very low change threshold. If trying to fathom out a new report takes too long, they will more than likely go back to their preferred version of the truth, the good old Excel spreadsheet.
Useless questions which add to the development, but provide no business value
Why does a campaign datamart need to contain sales order data? What benefit is there in a student attendance datamart of being able to drill down into a student address? If compelling arguments can be made then by all means include it. But the compelling argument shouldn’t be “because it’s there in the source system.”
The scope for the project is too wide
You’ve pitched your ideas to get buy-in from the executive. You’ve presented a utopian BI scenario where every department in the corporation will be reporting from your beautifully conformed dimensions. “There will only be one version of the truth from now on, and look at the possibilities with real-time data, superfast cubes that can provide you with dashboard reports in seconds instead of an hour” you say. Maybe you throw in data mining as a possibility just to seal the deal… They love it and set your team to work to deliver this vision. This happens all to often and no BI team can deliver a project like this unless a mature data warehouse already exists. Manage expectations carefully or your sponsors will be left disappointed.
Datamart is so big that performance is slow
Perhaps this sounds like a technical issue, but it may be a requirements problem too. An example I saw of this was a requirement for all of a website’s click data to be loaded on to the datawarehouse. The volume of data was so great that the load time deteriorated affecting the entire organisation and all associated queries took so long that nobody bothered to ask the question they originally wanted to answer any more. More disappointed customers. Given this data was already contained in summarised form elsewhere in the data warehouse, the BI specialist may have been wise to steer the requester away from this line of inquiry.
The success of a BI team is totally dependant on how their work is perceived by the clients they serve. A fantastic high speed ETL loading 10 source system’s data in real-time might be technically impressive, but if the marketing / finance / HR / executive team can’t get the answers they really need, then future work may be harder to find. So it’s in everyone’s interest to make sure that BI requirements are achievable and provide the answers to the questions, not questions about the answers.
What examples do you have of disappointed BI customers and what could have been done to avoid it? I’d be interested to hear your views.