Feature Article
Company News
Customer News
Product News
Professional Services News
Feature Highlight: Understanding InfoFlex
Anyone who has created serious budgeting and forecasting models in P4SS knows that InfoFlex is an invaluable process for transforming data to suit a variety of business requirements. Some examples of where InfoFlex may be useful include:
- Performing single, multi-tiered, and top-down allocations
- Seeding baseline data to future years in long term forecasts
- Incorporating growth factors to trend existing data
- Performing mid-year reforecasts
- Moving data between cost centres, or different time periods (intra-period forecasting)
- Copying data between versions
Enhanced Spread Methods
Beginning with P4SS 4.2, the spreading has been enhanced to include the ability to spread on any dimension member combination. In addition, it is now possible to spread based on the current target member combinations. In prior releases, spreading was locked to a single time-account-version selection irrespective of the member currently being processed.
The interface has also changed slightly with all dimension spread selections now moved to a separate tab in InfoFlex. By default, if Spread based on data is chosen and nothing is selected, the Current target member will be used for spreading during each iteration of the Common Dimension members. In the example below, data will be allocated and spread based on Gross Sales for the currently iterated member combinations of Versions, Time, and Currency. Hence, if the current iteration is Budget-July 2007-Native, data will be spread to the target members based on Gross Sales-Budget-July 2007-Native.
Performance Considerations
There are several factors that may affect the performance of InfoFlex. In addition to the actual server hardware in use, original model design decisions such as the number of dimensions, choice of dimension (parent-child vs. regular), data density, and presence of interacting account calculations in the cube, all contribute to overall efficiency.
Ultimately, the setup of the InfoFlex parameters has the greatest effect on actual performance. As InfoFlex can affect a large number of cells in the cube, it is best to be as precise as possible when creating the definitions. Always attempt to reduce the number of possible member combination to be updated. This information is presented in the Common Dimension tab. Avoid updating a large number of combinations (over 100,000) at one time. If you need to update lots of combinations, it is better to divide the task into multiple smaller InfoFlex processes instead of creating one with an enormous scope.
In P4SS 4.2, a new SQL processing option is introduced to improve performance of basic tasks such as copying data between versions or members. If InfoFlex detects these types of activities, it will automatically select SQL processing. The SQL option forces the calculation engine to operate directly on data stored in the fact tables. If the InfoFlex parameters are more complex, and SQL processing cannot be used, MDX processing will be used instead. In general, manipulating data on the fact tables with SQL will be much faster than its MDX equivalent. However, SQL processing is not possible if the InfoFlex definition contains any of the following:
- Source Data Aggregation option is set to Average including empty values
- The Contra Account option is enabled
- Time or dimension spread based on anything other than Evenly or Store
- The same spread methodology is not used for both time and dimension when spreading is required
- The Common/Target account has a time conversion other than Sum if Spread Evenly is selected
- A calculated version is selected as the Source / Common
- A calculated measure (i.e. YTD) is selected as the Source
In these instances, SQL is unavailable because the actual data is not easily derived from the fact table and must be resolved using an MDX query. Although logic is incorporated into InfoFlex to verify the validity of selected parameters, it is still up to the user to understand where data is stored in P4SS models because not all potential cases can be anticipated.
^ Top | Copyright ©2008 PROPHIX Software