Metadata modeling workflow

Import and verify metadata

Once your requirements are gathered and the supporting data sources are available, in the metadata modeling tool, you can import the required metadata to support the user requirements. You might be tempted to import everything, and then decide what to use. From a model maintenance, readability, and usability standpoint this is not recommended. Import only what you need, and then add more later as the requirements change.

For more information, see Import and verify metadata.

Remove ambiguity

Ambiguity in model design refers to potential misinterpretations of relationships and their cardinality by the Cognos® Analytics query service. You can remove ambiguity in the model by ensuring that the correct join paths are used in your queries, and that the intended fact tables and dimension tables are always treated as such. This design practice produces the expected aggregation for your measures.

For more information and background on resolving ambiguity, see Remove ambiguity.

Consider model design

In this phase, you need to consider how to present objects to the users in a clear, logical, and concise manner. You can consolidate multiple tables into one view, consolidate logically grouped tables into one area of the model, add filters and calculations as required, and so on.

For more information, see Model design and presentation.

Identify and configure for multi-fact, multi-grain queries

There might be scenarios where facts from different fact tables are stored at different levels of granularity, which refers to the scale or level of detail that is present in the set of data. For example, the Inventory Fact table stores values at the month level while the Sales Fact table stores values at the day level. Different levels of granularity might introduce scenarios where one fact is inadvertently double-counted (aggregated more times than it should based on the nature of the data). As a metadata modeler, you should identify these potential scenarios and configure the model accordingly to prevent double-counting.

For more information, see Multi-fact, multi-grain queries.

Enhance the model with additional features

Based on the needs and requirements of users, metadata modelers might use various techniques and features to enhance the model. For example, some users might want to do relative date comparisons of the data. Each metadata modeling tool has a way to accomplish this type of request.

Framework Manager uses dimensionally modeled relational (DMR) models for this purpose. A modeler creates dimensional objects that allow users to drill up or down through the data based on defined hierarchies. Dimensional functions can also be used to extract and compare data from different time periods or segments of the business. In data modules, modelers can implement navigation paths and relative date calendars to accomplish similar user requirements.

Other features to enhance the model or its performance include, but are not limited to, ensuring minimized SQL and controlling the way data is aggregated.

For more information, see Enhance the model with additional features.

Consider performance

As you develop your model, you need to constantly test for performance. Performance starts with the data source that you report on. However, there are also some key optimizations that can be accomplished within Cognos Analytics, such as leveraging data caches, data sets, and join optimizations across heterogeneous data sources.

For more information, see Optimizing query performance.

Iterate

Again, as with any project, the metadata modeler develops the model based on requirements, tests often, and then iteratively changes the model until the desired results are achieved.