Short answer: S/4 Hana does need a data warehouse for Business Intelligence (BI), but one with different qualities. Operational reporting moves from BW back to S/4 Hana – where it belongs.
Definition of Business Intelligence
Business Intelligence means to analyze all the corporate data to see how the company is doing and making fact-based decisions whenever possible. Data warehouses, data lakes, ETL, and semantic layers are just technological concepts to support this goal.
Focusing on the first requirement, analyzing all data, requires all data to be available in the system, either physical or virtual, and companies must be able to query it in sub-seconds. The S/4 Hana system contains a lot of corporate data, but not all. Firstly, it is lacking historical ERP data in many areas. It contains the current data, for example a sales order document in its final version, but it does not contain the information that this order had more line items previously and some got deleted because the delivery date would have been delayed otherwise. Master data is also stored in its current state only, for example when a customer has moved from one region to another. The consequence is that yesterday’s report opened today shows different results than before. That’s not good. Making decisions based on wrong data is worse than making gut-based decisions. Something about “Going into the wrong direction with confidence…”
Hana does have the option to retain the full history, it is just very expensive. Furthermore, S/4 Hana reporting is not prepared to take advantage of it. Adding non-SAP data into S/4 Hana is very uncommon, too, again mostly due to cost.
Granted, the Hana database has all the methods to also query non-SAP data, either by loading it into additional tables (to benefit from the database’s speed) or by keeping it in the original system and accessing it via data federation (SAP Smart Data Access). Option 1 requires bigger hardware, option 2 delegates the queries to the source system and hence returns the data in source-system-speed only.
At the end of the day, it will usually be cheaper to copy the ERP data into a secondary database, one that retains the data in historical context and is optimized to incorporate non-SAP data. Such a database is called the Enterprise Data Warehouse, and BW is a prime example.
Unfortunately, the above assumption that S/4 Hana can cope with all Business Intelligence requirements caused a sequence of suboptimal decisions that SAP customers must cope with somehow. For example, if no data has to be extracted, no BW Extractors have to be built for areas where the data model changed from ECC 6.0 to S/4 Hana. A full copy of the data is the only option for many of those BW Extractors, even today.
When SAP found out that many BI and operational reporting queries have a higher complexity and even Hana cannot beat the laws of physics, the obvious solution was offloading those queries into a second Hana database. This is the Hana System Replication feature. It creates a 1:1 copy of the ERP data, hence has the same issues discussed above. What would have been needed is a real-time system replication into SAP BW and its data structures.
This would require the transformation of the data. SAP Data Hub, now called Data Intelligence, was developed for exactly that – not because it is the best tool for the job but rather in the hopes of getting some sales revenue. All other SAP products that were better suited for the task got no further funding. Hana Spark Connector, Hana Smart Data Integration, and SAP Data Services all exist, but the release notes show little progress over the last years. With no appealing data integration story remaining, the only option was to rebrand the old Sybase database as a data lake in the hopes customers would put all their data into that (see also Hana Cloud Data Lake).
The final blow to SAP’s data warehouse strategy is the requirement to make business decisions in real time: How do customers benefit from being told today what decisions they should have made yesterday?
For customers, this is nothing new – I get a lot of requests to move SAP data into Snowflake for Business Intelligence these days. This is not optimal, either, but at least it’s cheaper and more future proof. SAP tools are usually only used for operational reporting with S/4 Hana data.