Worst Practices in Business Intelligence

I came across this white paper from Information Builders on worst practices in BI. Aside from having to register to download the white paper, Kevin Quinn does a good job highlighting the 4 key areas of failure:

Worst Practice #1: Assuming the Average Business User Has the Know-How or Time to Use BI Tools

This issue is very accurate. Even with the advances in user interfaces and hiding of the complexity to aggregate and join information, the tools are still overwhelming. What is interesting is what people are usually migrating from is desktop productivity tools like MS Access or Excel (see worst practice #2).

Worst Practice #2: Allowing Excel to Become the Default BI Platform

Excel becomes the default BI platform inherently because of worst practice #1, meaning the tools are too complicated. The simplicity, flexibility and ease of use along with the overall ubiquitous familiarity of Excel makes it the default tool of choice. Typical organizations at some maturity point will move from something like Excel to a standard tool (see worst practice #4). Nothing can be more frustrating then purchasing a BI platform and having to fall back to the legacy solution, which can be error prone and unmanageable.

Worst Practice #3: Assuming a Data Warehouse Will Solve All Information Access and Delivery

Data Warehouses are meant to store data, in whatever preferred design approach that makes sense (hub and spoke, data marts, enterprise warehouse, federated, etc). The BI application implementation should leverage this single view of the organization's data. The point that is made in the white paper is that the information needs typically go beyond what is housed inside the warehouse. The BI platform should have the flexibility to consume these non-warehouse sources. As much as I agree with this point, it can be inherently dangerous to use sources of "unknown nature". That's not to say that all data outside the warehouse is not trusted, but consistency in the usage of this information across the enterprise needs to be assessed.

Worst Practice #4: Selecting a BI Tool Without a Specific Business Need

Absolutely (see worst case #1 and #2). From my experience, the most successful selection and implementation processes have happened when both the business users and the IT group run the selection process together. I have also seen where one group runs the process independently from the other, the "opposing" group rejects the selection. So it is not just the business users that need to make the selection.

What additional worst practices have you encountered? The list is endless, but it would be great to start sharing other experiences individuals have.

In Search of BI Mashups

An area that has had a tremendous impact on the consumer aspect of the web is the concept of a “mashup”. The history of the term goes back to DJs and mixers combining different songs together to create new music. The term has evolved to more generically represent an application that is built by combining two or more data sources (if this isn’t the definition of a business intelligence application, I am not sure what is). The Senior Director of Engineering at Adobe put it best when he said

“a lot of talk about Web 2.0, web mashups, Ajax etc., which in my mind are all facets of the same phenomenon: that information and presentation are being separated in ways that allow for novel forms of reuse.” - Sho Kuwamoto

The same statement can be applied to enterprise data…separate the organization’s data from the different ways it can be presented. Where mashups come into play is when enterprises start presenting this data beyond grids and charts. In addition, as I have discussed on this blog, enterprises can combine traditional and non-traditional data sources to provide further context. Thus, the case for BI mashups.

If you perform a quick search for examples of BI mashups, you primarily find sample applications from different BI platform vendors. The first example I came across was from open source BI vendor Pentaho, which combines sales data with Google Maps to plot customer performance. Additional examples from Information Builders and Oracle offer similar examples. The trend with the majority of these examples is that they plot spatial data into geographic maps to show enhanced visualization. Not quite what Tufte would recommend, but certainly an enhancement over traditional BI. Adding non-structured data into the mix such as blogs and customer surveys through RSS feeds would enhance the experience even more.

For the technical audience, here is a very in-depth article by Larry Clarkin and Josh Holmes on mashups including examples, architectural components and key considerations when developing your first enterprise mashup. There is a wealth of information within the article, but one of the key elements applicable to a BI mashup is providing “rich visualization of data” for users that they won’t get from a typical chart or grid of data. If you are considering your first enterprise mashup, I would get familiar with this article as a first step.


There are some great resources available if you are looking for more examples of mashups. The most well known site is Programmable Web, which tracks interesting mashups, Web 2.0 applications and new web platforms. And if you have a short attention span and would prefer to see a video, check out this YouTube video.