Brief Summary
This webinar introduces a useful but often overlooked feature in SAP Integrated Business Planning (IBP): machine learning for master data quality checks. The presenter explains how this feature helps identify and correct inconsistencies in master data, using a simple example of customer locations. The tool uses statistical pattern matching to detect anomalies, and users can define custom rules for data validation. While corrections can be made within IBP, the original source systems should be updated for data consistency.
- Machine learning in IBP helps identify inconsistencies in master data.
- Users can define custom rules for data validation.
- Corrections should be made in the source systems for data consistency.
Introduction
The presenter, Russ Saad, introduces a feature of SAP Integrated Business Planning (IBP) that is available on every module and is designed to assist with daily tasks. This feature, which utilizes machine learning, helps in identifying and correcting inconsistencies in master data. Despite being introduced almost two years ago, it has not received widespread attention.
Accessing and Navigating Master Data in IBP
The presenter logs into a demo system for IBP, which has a restricted view to focus on relevant features. He navigates to "Manage Master Data," where various master data types are stored and can be reviewed and corrected. This section is particularly relevant for those responsible for setting up customer, location, and product master data. The demo system contains over 100 planning areas, which is more than a typical system should have.
Customer Master Data Example
The presenter uses a customer master as an example, which includes customer ID, description, channel, business partners, customer group, country, region. The customer master contains 115 entries. The example shows how countries are assigned to regions (e.g., Austria to Dach, US to North America). The presenter then discusses how to identify and correct issues within this data using the "Check Data Quality" feature.
Using Machine Learning to Discover Patterns
The presenter explains how to use the machine learning capabilities within IBP to discover patterns in master data. He selects "Customer Country" as the condition and "Customer Region" as the consequent, aiming to verify the relationship between these attributes. The system analyzes the data to identify patterns, such as the US being in North America.
Analyzing and Correcting Data Quality Issues
The system detects patterns and identifies potential issues. For example, it finds that if a customer region is North America, the country should be the US. The machine learning is based on statistics and pattern matching, not reasoning. The system identifies an outlier where a customer in the US is incorrectly listed in South America. The presenter demonstrates how to accept the suggestion and correct the data directly within IBP.
Limitations and Considerations
The presenter discusses the limitations of correcting data within IBP. Corrections made in IBP do not automatically update the source systems (e.g., S4, ECC). IBP is designed for planning data and may include attributes not relevant to the ERP system. Therefore, corrections should be made in the source systems to maintain data consistency. A report can be generated from IBP and sent to the master data governance team for correction in the source system.
Further Analysis and Data Volume Requirements
The system only flagged the instance where a US customer was in South America but did not question whether France should be in the Mediterranean region. A certain volume of data is needed for the system to reliably recognize errors. A small number of incorrect entries may not be sufficient for the system to identify a pattern.
Preventing Data Errors with Master Data Checks
The presenter demonstrates how to prevent data entry errors using master data checks. When attempting to enter Austria in the Mediterranean region, the system prevents the change because it violates a predefined rule that Austria should always be in Dach. These checks are configured under "Master Data Checks," where users can define rules and messages for different languages.
Configuring Master Data Checks
The presenter shows how to create a master data check for the customer master data type. The check is set to show a warning and is currently active. Language-dependent messages can be created for different user groups. The check criteria involve creating a filter to detect specific combinations, such as ensuring Austria is always part of Dach. Wildcards and long lists of conditions can be used, and relationships between different master data types can be established.
AI Components: Machine Learning and Hardwired Logic
The presenter explains that AI in IBP combines machine learning and hardwired logic. Machine learning uses statistical analysis to identify patterns and group data. Hardwired logic involves predefined rules, such as those used in master data checks, to prevent incorrect data entry. The system is not intelligent but relies on pattern matching and predefined rules.
Numerical Checks and SAP Joule
The system can also perform statistical checks on numerical data to ensure values are within expected bounds. SAP Joule, a standard feature in SAP Cloud products, can provide assistance and explanations. Users can ask Joule how to perform tasks, such as running a master data health check or using machine learning to detect patterns. Joule provides relevant links to help documentation and explains the metrics used in the analysis.
Key Points and Limitations Recap
The presenter summarizes the key points: the feature is found in master data maintenance, requires at least 100 master data entries, needs a logical connection between attributes, and sufficient example data for correct pattern recognition. Limitations include that it does not change data in source systems and requires enough data to function effectively.
Q&A Session
The presenter answers questions from the audience. The relationships within the data are what the machine learning part is looking for and are not pre-coded. Master data checks need to be created by the user. The session recording will be available on the Westernacher YouTube channel. Joule is a standard feature in IBP, but advanced AI functions may require an additional license. Planning data maintained in SAP IBP can be fixed within IBP, but the source systems must be corrected separately.
Closing Remarks
The presenter encourages attendees to use the AI features in IBP to improve data quality and streamline planning processes. He emphasizes the importance of accurate master data and encourages users to utilize SAP Joule for assistance.
