Explaining modeling options

Dimensions represent categories that provide perspective on your numeric data; for example, product category, date, region, cost center, etc. Dimensions can contain properties that further describe a dimension. For example, you may have a dimension for customer which has properties such as phone number and address to further describe the customer dimension.

Dimensions can also be rolled up into a hierarchical view; for example, time (year, quarter, month), geography (country, region, location), employee structure (executive, manager, employee), etc.

Measures represent the numeric values that you are analyzing; for example, sales revenue, salary, number of employees, quantity sold, etc. Sometimes these quantities are contained in a single dimension referred to as an Account type dimension (and probably with the name Account, or something similar). In this situation, the numeric values represent the line items on a corporate balance sheet, income statement, profit/loss statement, etc. But you can also present the numeric values as individual elements called Measures.

Together, dimensions and measures are the framework for viewing data, whether it be a trend line of revenue over time or a tabular comparison of gross margin across different regions.

Models are comprised of dimensions and measures and represent a specific subset of data; for example, sales, production, financial, shipping, etc. Models are the primary data sources for SAP Analytics Cloud stories and analytic applications.

In SAP Analytics Cloud there are two styles of models:

• Analytic model: read-only
• Planning model: read/write

We will look at each style in more detail in the next concepts.

Watch this video to learn about the data source options for creating a simple SAP Analytics Cloud model.

An analytic model is used strictly for read-only data reporting and analysis. A date dimension is available but is not required, and you can remove it from the model during the design stage. Why is a date dimension optional? One scenario is that the model represents only current data. If users know the data is always "current," there is no need for a date dimension.

Planning models are pre-configured with required dimensions for time and version. These dimensions are required because planning activities are dictated by time frames, and the planning numbers are intended for different purposes – budget, forecast, and planning. They offer support for multi-currency and security features at both the model level and dimension level. When working with a planning model in a story, users with planning permissions can create their own versions of model data. These users can also write data to the model by typing new values, copying and pasting data, and using data actions.

A dataset is a simple collection of data usually presented in a tabular format. You can use a dataset as the basis for a story.

SAP Analytics Cloud has two types of datasets:

• Embedded - Embedded datasets are embedded into a story and are unique to that story. They cannot be shared outside the story or refreshed.
• Public - Public datasets are standalone datasets and can be shared among different stories.

Both types of datasets can be enhanced with basic data transformation/wrangling functionality. Neither dataset can be scheduled for a refresh; you must manually re-import the updated data. SAP Analytics Cloud automatically matches the columns of the newly acquired data to the columns of the existing data, but any prior data transformations will be lost. If you imported data from a flat file, you can only re-import a compatible file: a file that has the same number of columns as the original file, and with the same column names and data types as in the original file.

Both datasets can be secured to allow users access to the dataset or not. Specific column-based or property security, however, is not supported for any datasets.

You can convert an embedded dataset to a public dataset. However, a limitation to a public dataset is that you cannot change its datasource. For example, if your public dataset was originally created from a flat file but you now want to use a BW query, you have no option to make that change. Embedded datasets, on the other hand, do allow you to change the data source via the "Add New Data" option.

You can also convert an embedded dataset into a model, but any transformations you made to the dataset are lost and must be recreated in the model.

Note

A public dataset cannot be converted to a model.

Overall, datasets and models complement each other. Datasets are perfect for ad hoc, ungoverned use cases based on acquired data. Models are used when the use case requires more governed data analysis and planning scenarios.