Smart Discovery

Let Smart Discovery analyze your data by running a machine learning algorithm to help you explore your data in a specific context, to uncover new, or unknown relationships between columns within a dataset.

In SAP Analytics Cloud, you can use Smart Discovery to clearly define the context of the business question you want to ask your data. To form the context, or topic of this business question, you must choose values for some settings in the Smart Discovery side panel.

Smart Discovery Settings

Some of the settings options that you see in the Smart Discovery side panel are determined by the type of underlying model you're using - a classic account model or a new model type. The main difference between the model types is how they handle measures, which are the structures in your model that hold numeric values. To learn more about the different model types, you can refer to the chapter called Get Started with the New Model Type.

Classic Account Model

In a classic account model, model values are stored in a single default measure. To define the context of your business question when using a classic account model, you must select a value for these settings:

Target: The Target is the measure or dimension you’d like to know more about.
Entity: The Entity defines the dimension or dimensions you’d like to explore in relation to the target, and the level to which Smart Discovery must aggregate your data to analyze it for you. Within the dimension or dimensions you choose to define your entity, you can include all, or exclude some dimension members. There's no need for any data preparation. Simply select the Entity value from the Smart Discovery side panel.

Example

If you want to explore revenue by customer, you'd select 'revenue' as the Target, and 'customer' as the Entity.

You can also refine the scope of your topic by adding Filters, but the filter selection is optional.

New Model Type

The new model type exposes measures as single entities and lets you add and configure multiple measures with aggregation and units to fit your data. To define the context of your business question when using a new model type, you must select a value for these settings:

Target: The Target is the account or dimension you’d like to know more about.
Entity: As above
Measure: The Measure you select determines the values that are assigned to the account dimension in the model, to be analyzed by Smart Discovery, ignoring all other measures in the model.

Example

Select 'revenue' as a Target, 'customer' as an Entity, and 'amount in euros' as a Measure to get Smart Discovery results only for revenue by customer, where the accounts are consuming the values for the measure 'amount in euro'. So, your Smart Discovery results show only those accounts in the model that have values for 'amount in euros'.

Note

If your new model type contains an account dimension, all accounts that appear in your Smart Discovery results are based on the Measure you select in your settings.

Note

You can run Smart Discovery for measures associated with the "average", "minimum", and "maximum" exception aggregation types. These measures are included in the Smart Discovery analysis and can be selected as Key Influencers.

You can also refine the scope of your topic by adding Filters, but the filter selection is optional.

Restriction

Account dimensions that have multiple hierarchies are not supported.

Refining Your Smart Discovery Exploration

You can always ask more than one business question. Explore your data from different angles by asking Smart Discovery to analyze the same target in relation to different entities, and it produces different results.

You can use custom names for the target, target group, and the entity. The custom names you choose are then used instead of the default name, wherever they appear in Smart Discovery.

To narrow down the scope of the topic, or to optimize running Smart Discovery on large datasets, you can create filters to exclude specific records from dimensions in your model.

Smart Discoveries are driven by users working on stories. You can exclude particular dimensions or measures from the analysis, focus on particular dimension members, and display relationships from a list of influencers.

Smart Discovery Results

Smart Discovery always gives an overview of your data by automatically building charts to begin discovering more about your data. The generated content shows the target for the selected entity in relation to interesting or relevant dimensions and measures. Depending on the data you select to analyze, Smart Discovery can also show the influencers on the topic of the discovery, how they relate to one another, and key members or value ranges. Smart Discovery also alerts you if no significant or insightful relationships are present in your data.

After running Smart Discovery, you can simulate numeric targets, using the results of the key influencer analysis. You can also view data records that are highlighted by the predictive model as being unexpected.

The Smart Discovery results can be saved as story pages that you can share within your organization.

Note

Currently, you can't run Smart Discovery on more than one million cells. By filtering the data first, you can reduce the number of cells included in your dataset. The number of cells is calculated by multiplying the number of measures by the number of records.

Note

The Export/Import functionality via the Content Network does not work for Smart Discovery.