Why the replacement of SAP BW on HANA is due now
The digitalization of business processes and the desire for data-driven corporate management make it clear that the time is ripe to replace SAP BW on HANA. Traditional enterprise data warehouses such as SAP BW are increasingly reaching their limits. While SAP is focusing on SAP Datasphere, companies are faced with a strategic decision: do we continue to rely on the SAP world – or do we choose future-proof alternatives such as Snowflake or Databricks?
For IT decision-makers, this results not only in a technological but also an economic consideration: How can the existing knowledge from SAP BW be transferred to modern cloud DWH platforms – and which architecture offers the greatest added value for business, analytics and governance?
Why SAP BW on HANA is reaching its limits
The decision to replace SAP BW on HANA often coincides with the desire to be able to flexibly integrate more data sources. For many years, SAP BW on HANA was the standard for structured reporting on SAP data. However, there are now clear limitations that make a migration necessary:
Rigid data model with limited enhancement options for non-SAP sources
High maintenance effort due to complex transformations and dependencies
High license and operating costs
Little flexibility in the integration of modern analytics and AI tools
Cloud maturity is limited – even the BW/4HANA cloud variants remain “SAP-centric”
Many companies are also facing another problem: BW experts are retiring, while the search for young BW talent is faltering.
Databricks and Snowflake: modern alternatives for a cloud data warehouse
Snowflake – more than just a DWH
Originally conceived as a cloud data warehouse, Snowflake has evolved in recent years into a state-of-the-art data platform that integrates key elements of a lakehouse architecture.
With components such as External Tables (access to external objects in the data lake), Snowpipe (streaming ingestion), Snowpark (data-related processing in Java, Scala or Python) and Cortex (integrated AI functions), Snowflake now offers much more than classic BI reporting.
The separation of memory and computing power, support for semi-structured data formats (e.g. JSON, Avro, Parquet) and a high-performance, scalable architecture with ACID transactions make Snowflake a lakehouse-capable platform. Especially in combination with Qlik Talend Cloud, dbt or Unified Star Schema modeling, Snowflake is suitable for modern, modular data architectures.
Even though Snowflake is structured more like a classic DWH in terms of its DNA, it now fulfills many requirements that were previously typically reserved for lakehouse platforms such as Databricks – with a strong SQL orientation and high user-friendliness.
Databricks – the lakehouse for analytics, data science and AI
In contrast, Databricks takes a different architectural approach. With the lakehouse approach, the platform combines the advantages of data lakes with those of classic data warehouses. This creates a flexible and powerful database – not only for reporting, but also for data science and artificial intelligence:
Delta Lake as the central storage format (ACID-compliant, high-performance, versionable)
Apache Spark-based – Ideal for big data and parallel processing
Native support for machine learning and AI
Python, SQL, R, Scala – all combinable in notebooks
Unity Catalog for Data Governance
Thanks to native support for Delta Lake, ML and streaming, Databricks is particularly well suited as a lakehouse platform. In combination with data vault modeling on Delta Tables, scalable, compliance-capable architectures are created – ideal for companies with an AI strategy.
Architectural considerations: Lakehouse, Data Mesh, DWH - what fits when?
Classic data warehouse (e.g. Snowflake “out of the box”)
Recommended if:
focus on clear structuring and governance
BI and controlling are the primary use cases
Structured data sources dominate
fast implementation with SQL-oriented teams is required
Lakehouse (e.g. Databricks or Snowflake with Qlik Talend Upsolver)
Recommended if:
structured and semi-/unstructured data sources are to be integrated
large amounts of data have to be processed with high performance
both classic reporting and explorative data science & AI are planned
a standardized storage format (e.g. Parquet) is desired
Snowflake – especially in combination with tools such as Qlik Talend Cloud (incl. Upsolver streaming) – can also be used as a Lakehouse-compatible platform platform. This enables a flexible architecture in which raw data is initially stored in a landing zone then transformed in Snowflake, modeled (e.g. via Data Vault) and then used for reporting, self-service and advanced analytics.
Hybrid architectures
A hybrid form makes sense for many companies:
structured SAP data flows into Snowflake via conventional pipelines
Event streams and logs are transferred to a semi-structured data design via streaming solutions such as Upsolver
The whole thing is consolidated in the Data Vault and exposed via a semantic layer (e.g. Power BI or Qlik)
Data mesh approaches
A data mesh can be particularly useful for distributed responsibilities in large organizations – and can be implemented with both Snowflake and Databricks.
💡
Excursus: MACH architecture as a strategic framework
Modern data warehouse and analytics platforms such as Snowflake or Databricks can be perfectly integrated into MACH architectures. MACH stands for microservices, API-first, cloud-native and headless – an architectural principle that aims to maximize flexibility, scalability and independence from monolithic systems.
Especially in the context of composable data stacks, MACH enables the combination of specialized tools along the entire data value chain – from ingestion and modeling (e.g. Data Vault) to reporting in Power BI, Tableau or Qlik.
Modeling approaches in comparison: Data Vault vs. unified star schema
Modeling approaches: From structured storage to self-service reporting
Modern data architectures benefit from strategically combining different modeling layers. Two established approaches – Data Vault and the Unified Star Schema (USS) – fulfill different roles in the overall picture.
🧱 Data Vault (with Databricks or Snowflake)
Data Vault is a methodical approach to flexible, historicized and scalable data storage. Structured hubs, links and satellites create a clearly comprehensible architecture – ideal for governance-driven organizations.
- Historization and auditability through structured hubs, links and satellites
- ELT processing with modern engines (e.g. Delta Lake for Databricks or Snowflake Streams/Tasks)
- Scalability through separation of structure and content
- High automation capability with dbt, Qlik Talend and Co.
Data Vault is particularly suitable for hybrid organizations with strict governance requirements and complex data landscapes.
🧠 Analogy: Data Vault is like a well-thought-out closet – no matter how many new things you add, it stays tidy and easy to find.
📊 Unified Star Scheme (Snowflake)
The Unified Star Schema (USS), developed by Bill Inmon and Francesco Puppini, is an agile, standardized reporting model for self-service BI. It abstracts complex data landscapes into a user-friendly structure.
- Agile & resilient: adaptable to new data sources and business requirements
- Holistic data view: avoids fan and chasm traps
- Self-service-friendly: Optimized for Power BI, Tableau or Qlik
Application scenario: Companies aiming for a group-wide BI model with self-service access.
🧠 Analogy: USS is like a voice-controlled interface for your closet – “Show me all blue shirts with long sleeves” – and immediately delivers the right data.
Migration strategy: From SAP BW on HANA to Snowflake or Databricks
1. analysis of the existing BW landscape
Which objects (InfoCubes, queries, transformations) are active?
Which data sources are connected?
Which user groups and reports exist?
2. mapping to target architecture
BW Queries ➝ SQL Views / Materialized Views in Snowflake
DSO / InfoCubes ➝ Delta Tables (Databricks) or optimized table structures (Snowflake)
Process chains ➝ orchestrated pipelines (e.g. with Qlik Talend, dbt, Azure Data Factory, Apache Airflow)
3. selection of the appropriate migration approach
Lift & Shift (short-term, low transformation)
Greenfield / re-design (medium-term, new architecture with modern model)
Hybrid migration (combination, e.g. for quick wins and long-term optimization)
4. development of a data platform
Central landing zone for raw data
Automated loading processes (ELT)
Modeling (USS / Data Vault)
Semantic layer for self-service
Role and rights management
Monitoring & Cost Control
Decision matrix: Snowflake or Databricks?
| Criterion | Snowflake | Databricks |
|---|---|---|
| Target group | BI, controlling, reporting | Data science, AI, engineering |
| Data formats | Structured + semi-structured | Structured + semi-/unstructured |
| User interface | SQL, classic BI | Notebooks, code-first |
| Modeling | Data Vault, Unified Star Schema (USS) | Data Vault, USS possible (Lakehouse) |
| Platform architecture | Cloud DWH with Lakehouse functions (e.g. Snowpipe, Snowpark, Cortex) | Lakehouse (Delta Lake, Unity Catalog, ML native) |
| Governance & Sharing | Stark (Data Sharing, Masking, RBAC) | Strong (Unity Catalog, RBAC, Lineage) |
| Cost model | Usage-controlled (storage & compute separate) | Compute-intensive, granularly scalable |
| Skillset in the company | SQL / BI | Python / Notebooks / Engineering |
Conclusion: The right way out of the SAP BW world
The replacement of SAP BW on HANA is more than just a technical project – it is a strategic step towards modern, flexible data architectures.
Snowflake is ideal when the focus is on clear structures, SQL-based workflows and BI reporting. However, with External Tables, Snowpipe, Snowpark and Cortex, Snowflake can also be used as a modern lakehouse platform – especially in combination with Qlik Talend, dbt and semantic modeling.
Databricks scores particularly well in exploratory data analysis, machine learning and the processing of large, heterogeneous data volumes. The native Lakehouse architecture with Delta Lake and Unity Catalog offers maximum flexibility for engineering and AI teams.
Both Snowflake and Databricks can be operated in a lakehouse-like manner – depending on the tool stack, architecture strategy and data engineering expertise in the company. While Databricks maps this architecture natively, Snowflake can also be expanded into a high-performance lakehouse solution in combination with Qlik Talend Cloud (Upsolver). Further information on the Lakehouse architecture can also be found in our blog post: ” A differentiated view of the architectural differences ”
Both platforms can be operated in a modular, high-performance and future-proof manner. The most important decision is: What goals are you pursuing with your data strategy – and which architecture suits your organization?
🚀 Act now: Start assessment or pilot
Do you want to know what a concrete replacement of your BW system could look like? We offer:
- Quick analysis of existing BW architecture
- Workshops for target architecture Snowflake or Databricks
- Pilot setups incl. Data Vault or Unified Star Schema
- Technology-independent advice from experienced architects