Flexibility Is KeyBy Michael Linhares, Dr. | Posted 2009-01-26 Email Print
Virtual data marts add flexibility, security and scalability to Pfizer’s rigorous drug development process.
Flexibility Is Key
We have introduced maximum flexibility into a necessarily rigid, standards-based framework, with the understanding that changes can be made as needed without crippling the process. For example, let’s say that a business request is not specific enough, or is left up to the deployment team to interpret. After going with the initial interpretation and putting the information in the hands of the business team, we may need to start over if the interpretation is incorrect.
Because we haven’t made huge investments or spent weeks developing an analysis that is ultimately rejected, our teams can start over quickly and easily. This ability engenders a different engagement and thought process throughout, emphasizing continuous improvement.
One of our biggest continuous-improvement efforts has been providing a single trusted source of information for the teams and eliminating the individually managed spreadsheets and workbooks. Further, we have introduced data transparency to the teams and departments, so everyone has access to the same information at the same time. This has improved decision making significantly, while also building trust within the teams.
Now, when a team member has a new idea, it can be incorporated or tested by using the cache capabilities in Composite Studio to store, in a single location, a defined set of data on a routine basis: several times a day, nightly or monthly. We call this approach “light ETL” (extraction, transformation and loading) because we avoid the development cycles required for the strategic ETL tools that we use for other cases.
When changes are requested to a data set, we can update the view quickly without complex processes. We are able to reuse common dimensions across multiple subject marts, ensuring that the same vocabulary is used across different data presentations (for example, consistent project naming, team members’ formatting and project status indicators). This eliminates confusion for all users and allows more widespread integration.
To date, we have achieved the following measurable results from the project:
• 90 percent reduction in the time that’s needed to create a new report using the data abstraction layer;
• reduction by a factor of three the time-to-value (business benefits) for new views and applications;
• 100 percent increase in key business analyst productivity by reducing the wasted effort of creating permanent data mart solutions for one-time, nonrecurring data integration projects;
• 5 percent improvement in R&D project delivery through faster deployment of strategic planning information to key line-of-business senior executives and managers;
• elimination of the spread marts.
Extending the Success
Our virtual data mart program has been a success, judging by the increased delivery of value and the growing number of information requests. This has led us to consider new areas for additional business value. For example, we maintain a candidate project database for managing project portfolios.
Today, this database is hand-coded and stitched together in heavily customized ways. The data quality is questionable, and some necessary data—schedules, inventory, safety, etc.—isn’t included. Plus, few people understand the database. To correct this situation, we’re planning to virtualize and abstract this information for easier access by those most likely to benefit from access to it.
By deploying virtual data integration that adapts to our changing business requirements, we have found a way to strike a balance between adhering to complex standards, while remaining flexible enough to change course when new information points to promising breakthroughs.