For some time, self-service BI environments have been hailed by both marketing and technicians as a break-through that will enable business users to find data for themselves without the need of IT intervention.  Many of course are skeptical of these claims.  As a long serving BI professional, I too see the pitfalls of these self-service environments.  But done correctly, I do see potential for good things.

photo credit: ryk_neethling

I like to think of self-service BI environments as akin to common language bible translations.  Before bibles were available in the common languages of English, French or German, one had to seek the advice of a high priest for a religious interpretation of scripture.  Once the bible was available to any literate person, people could read the text themselves and make their own interpretations.  This of course led to an explosion of religious sects as many interpreted the scripture in different and sometimes contradictory ways.  Violent disagreement and often war would result from some of these differences.  The ensuing mayhem eventually led to freedom of religion being enshrined in most Western democracies.

In IT/BI, we also have a sort of priesthood with our own specialized languages (SQL, MDX) and we have historically held a position of power and trust on the subject of corporate data.  But when data does become available to the masses, we have to ensure that it is interpreted as uniformly as possible lest there be hundreds different interpretations of a corporation’s performance.

How can this be accomplished?

• Common metadata definitions – The data set metadata definitions must be communicated to the users and understand by all.  What is a customer?  What is a sale?  What is revenue?  Without controlling metadata definitions, the data is vulnerable to misinterpretation.
• Business user data literacy – Business users must have a good understanding of the underlying data.  This might involve an in-depth understanding of statistics, engineering or accounting depending on the data sets in question.  The better they themselves can understand it, the quicker they will see anomalies and resolve problems.  The ability of this group to understand data models is often instrumental as well.
• Use of Tools – Users must be well trained in their BI tools and understand when one tool is better to use than another.  Some tools work exceedingly well for some tasks but not for others.  Some effort must be made to control and monitor usage of tools – Report Studio should not be used to derive large scale data extracts for example; this is neither its strength nor its intended use.  Usage will tell you what is working well and what is not; where users are getting along well and where they are having problems.  Question the value of tools that are not being used, or ask users what they find lacking in that particular tool.
• Data Scope – Business users need to understand the scope of available data in the BI environment.  Any shortcomings need to be part of an ongoing BI plan, and users need to know when this data will become available.  Some industrious users may resort to extreme lengths to secure data sets when data is nonexistent, such as typing it into Excel spreadsheets themselves.  I have seen elaborate and labour-intensive Excel or Access solutions created by business users that needed to be reworked into the BI environment.

What this really comes down to is communication.  As Thomas Redman puts it, when there is data misinterpretation, you do not have a technological problem; you have a communications problem.  Like a good coach, BI must ensure that everyone understands the game plan and is on the same page.