Who is responsible for the accuracy of your data? In the data
warehouse, the data steward is the person.
A data steward should be from the user community and is responsible for
providing the correct algorithms to IT. IT is then responsible for
calculating that algorithm correctly. IT should not be responsible for
the algorithms.
For example, let's use a common metric in the hotel industry - RevPAR.
RevPAR is the revenue per available room. To IT, the formula might
logically be Total Revenue divided by number of rooms in the hotel.
However, that is incorrect. If there are rooms taken out of service
(i.e. upgrades, repairs, etc.), then they do not count toward RevPAR.
When I worked with a hotel chain, I asked someone in upper management
how they defined RevPAR. The response was "That is a fairly known
metric in the industry." I replied that while I understood that, I
wanted her definition. The reason was that IT had been getting two
different definitions of RevPAR from the sales and marketing
organization. I never received an answer.
This is where the data steward becomes important. The data steward is
the user community's liaison between the business and IT. The data
steward defines what the algorithms are and then it is IT's
responsibility to "code" it correctly. If there are any problems with
the data, the data steward is the final authority of what should go into
that column. How many data stewards should there be? One per column.
One data steward can do more than one column, but you want only one per
column.
The role of the data steward needs to include, but is not limited to:
1) The business rules (i.e. Can a sales person work in more than one
store?)
2) Quality expectations (i.e. Do our addresses have to be perfect for
the Postal System, or what is sufficient to be successful?)
3) Security (i.e. Who will have access to this information?)
4) Where does the data come from? (i.e. If there are multiple source
systems that have the same data element, which one is the source for the
data warehouse - commonly called the "system of record").
Some organizations create a committee to be the data stewards. While I
prefer a single person with a backup, some organizations like groups.
We must work within the organizational structure that fits the company's
culture. If not, we will risk having a less than successful business
intelligence environment.
