The following guidelines are useful to those who are new to BI or are embarking upon implementing a BI solution for the first time. They are based upon many years of BI solution delivery experience.
| 1 |
A BI project should be a business led initiative supported by IT, not an IT led project.
|
| 2 |
The dependency upon operational systems to support BI users should be removed. Resist the temptation to integrate BI reporting tools with operational databases for anything other than prototyping, you will not get the information consolidation that you require and you will compromise your operational capabilities too.
|
| 3 |
The key performance measure definitions should be defined early, centrally and shared by all business departments in order to propagate a single version of the truth across the business.
|
| 4 |
Do not overestimate the quality of your operational data. All organisations think that their data integrity is excellent, in reality it seldom is, especially if information is held within legacy systems. Data integrity should be addressed as part of a BI project. A BI implementation should close the loop with operational systems, highlighting any anomalies in operational data and drive the resolution of data quality issues. Often 60 - 80% of the data consolidation effort can be directed towards data cleansing issues.
|
| 5 |
A pragmatic mixture of dimensional and relational data modeling often facilitates the best reporting and analysis solution. Don't assume that a single approach will always work. BI tools nearly all work best with a dimensional model. For future proofing though, you may also want to hold your key information in a relational model too.
|
| 6 |
Do not try to answer all questions from the start. Concentrate on your key reporting measures at the beginning of a BI project. A lot of value is gained from delivering key information in a consolidated, accurate and timely fashion. Once users get to grips with key performance information, they will ask much more poignant business questions than they will before they have access to this information.
|
| 7 |
Allocate appropriate time and resource to quality assurance testing of your data. Remember, It takes only one inaccurate report to discredit your BI implementation. Do not assume that because the BI database is available and your data loads are error free that the information is accurate. Test it thoroughly, especially if you intend to publish KPI's outside of your organisation. Getting it wrong may have a dramatic effect upon the share price!
|
| 8 |
Publicise your BI system internally. You’ve invested a lot of time and money in implementing a solution, get people to buy into its usage. You want it to be the single point of access for business information in order to discourage the departmental cottage industry approach to BI.
|
| 9 |
Choose solution components and a technology that are right for your organisation. Do not go with niche technology if you have no in-house experience in that area. It will be difficult to maintain the long term commitment required if you cannot find the skilled resource needed and niche skills are often expensive to hire. This advice becomes even more critical when considering tip 10.
|
| 10 |
Understand and accept that there is no end to a BI project. BI systems are subject to Pandora’s Box syndrome. Access to quality information leads to more searching questions which in turn feeds the requirement to consolidate more information.
|
