For every SAP data-centric data platform, data engineering teams must decide whether to extract data from Core Data Services (CDS) views or directly from SAP tables. While CDS views offer significant advantages in data modeling—ensuring that models are fit for purpose—they are not always the best option. Many organisations may lack the right technology, skills, or processes to fully leverage CDS views for data extraction. Having gone through many implementation projects myself, and prompted by a post on Linkedin from my friend and colleague Ronald Konijnenburg, I thought I’d share my views on the topic here.
What Are SAP CDS Views?
SAP CDS views are a powerful abstraction layer that allows for defining data models within SAP applications. They simplify data access and help organise SAP data for analytics consumption. For an introduction to CDS views, see for example this article on the SAP community site. However, despite their benefits, CDS views are not always the best choice for data extraction. Below, we explore the key technological, skill-based, and process-related challenges associated with CDS view-based extraction compared to table-based extraction.
1. Technology
A common approach for extracting data from source system is the use of a Change Data Capture mechanism. This mechanism is widely available for databases, using either trigger- or log-based replication. Both SAP and third parties provide out-of-the-box solutions for a wide range of databases and applications, including SAP ERP. These methods enable efficient, incremental data extraction.
CDS views, however, are SAP proprietary, and CDC mechanisms are not readily available in third-party tools. To the best of my knowledge, only SAP DataSphere and SNP Glue offer an efficient, CDC based extraction for CDS views.
OData extraction is available for CDS views, but it does not scale well for enterprise-level workloads. Large-scale data extractions using OData often run into performance limitations, making direct table extraction a more viable approach for enterprises handling high data volumes.
2. People (Skills)
Maintenance on CDS views require SAP-specific skills, which are not always available in data & analytics teams. This skill gap can create a bottleneck when changes or troubleshooting are required. Many CDS views are not delta-enabled, meaning they cannot support incremental data extraction. This makes them unsuitable for large-scale data pipelines that depend on efficient updates. Not all SAP tables are included in CDS views. If a required table is missing from existing CDS views, the engineering team must either create a new CDS view or revert to direct table extraction.
3. Processes
Changes to CDS views must follow the SAP S/4HANA change process, which is significantly more rigorous than data platform governance. This level of control is justifiable, given that SAP is the core system of record in most enterprises, but it means even small adjustments can take a long time to implement. In contrast, adding a table to an extraction process can be as simple as selecting a checkbox, making it a far quicker and more flexible option.
Conclusion
If you are on a technology stack which supports CDC for CDS views and you have a team comfortable with the maintenance of CDS views and the governance processes around it, then using CDS views for data extraction can give you better data products more quickly.
If you wish your data & analytics team to be independent from the SAP governance processes, you don’t have an ETL tool supporting CDC for CDS views and you don’t have easy access to SAP skills, then you are probably better of with direct table extraction. It’s often said that creating models on SAP tables is very difficult as SAP has several hundreds of thousands of tables. You only need a fraction of those though, and it really is not that complicated to create a data model suitable for analytics. The initial hundred or two hundred tables will easily support the first use cases and even for larger implementations I often see fewer than a thousand tables being used.
With the considerations outlined in this article, organisations can prioritise and score the different approaches and choose the data extraction strategy that works best for them.
