Many a times, a lot of time is lost in getting the timely decisions from the stakeholder for BI. Unlike a transaction system, where most of the decisions and their impacts are visible in terms of functionality, impacts are more difficult to visualize in case of BI. This is more pertinent, when we are modeling and designing the core components. For example, if there is a decision to not to include certain dimensions or measures in the dimensional model, the overall impact for the business stakeholders will be the analysis they will not be able to do or the reports they will not be able to see. Please refer Analyze Well, but be decisive.

One may like to refer unique challenges of Data Warehouse initiatives and assessing organization readiness for Data Warehouse, to appreciate on how BI projects are different animals and require different techniques to tame them.

Here are some areas of decisioning and tricks to better articulate for faster decisioning:

Information Needs:

Information needs by a stakeholder can have an impact on timelines as well as cost. The articulation would include:

• Which information need is leading to how much impact on timelines and cost?
• Are there any middle-points where partial information needs can be met by minimal impact on timelines and cost?
• Are there any manual workarounds or temporary methods to satisfy the information need in short-term?
• What can be the options in the level of detail a user need to go into? It might be possible that enabling higher level analysis could be much easier compared to the detailed data availability.

Information Quality Needs:

Business typically looks for 100% data quality. The best way to drive decisioning about the level of optimal quality is:

• The effort business has to put in to improve the data quality. Business is the provider of data which gets into the source systems that feed data into BI. The demand for high data quality places a direct demand on business to work harder. This statement should not be taken as BI vs Business. It is important for the business customers to know the cost-effort estimate on getting improved data quality.
• The BI cost impact on data quality: Sometimes the effort on transformation to ensure improved data quality can be huge. Please refer transformation, data correction and enrichment for more details on this subject. The expected cost of some time-guzzling data quality issues should be highlighted to business.
• The workaround and tolerance to data quality issues: Some range of data quality issues might be acceptable to business, given the usage objective. For example most of the data used for analytics, may take some shortfalls- Like pin-code based sales analysis. If 2% of pin-codes are faulty, the analysis may still work.

User-Base Needs:

The number of users needed for designing and viewing the outputs of BI tools. The number of users should ideally not have significant impact too soon, if the sources they are accessing are well distributed. For example, if all the users are accessing the DW and OLAP layer directly, it can have scalability issues. Alternatively, if the users are accessing most of their information from the reports based out of the reports repository in the enterprise reporting tool, the impact on DW will be lesser.

Frequency of Refresh Needs:

There is big hype now-a-days on 'real-time BI', which claims to provide the data on real-time basis from the source system. We have reservations against this concept of real-time BI to be used as an enterprise solution. Even if we are not pursuing real-time BI, there are demands for multiple refreshes in a day. One needs to share the impact of multi-refresh on the availability of the data during the refresh. One also needs to look at the reasons for multiple refresh. If one questions hard one may realize that the multiple refreshes in a day may be driven by select information needs. One may work-out methods by which:

• Multi-day refresh to be done not for all but for few data-sets.
• The interim information for the select information groups can be taken directly from the production systems, instead of from BI platform. The chances are that this kind of 'high-frequency' information need will be purely related to operational domain.