HANA promises to cater for both, OLTP and OLAP, workloads. That allows to provide operational analytics within a S/4HANA system. The SAP-focused reader might wonder why, on earth, do you still want to have a BW/4HANA system in your landscape? This blog looks at 3 anonymised customer examples that reveal why having a data warehouse – such as BW/4HANA – is even more pressing in times of digitalisation than ever before. A data warehouse is thereby considered as the place that brings data and its underlying semantics from a variety of sources together in one place, either physically, virtually or mixed, either using an RDBMS, a big data environment or a combination thereof, either deployed on premise or in the cloud.
Example 1: Consumer Goods Customer
The first example comes from a leading consumer goods company. Figures 1a and 1b show details from 2 of their slides and list the sources of data that feed into their data warehouse. As expected, there is a bunch of traditional SAP systems, such as ERP (S/4), CRM and APO, but – as it has become common in days of digitalisation – also from sensors, logs, digitalised sales and marketing. Now, bringing that data semantically together – for example, to understand the financial impact of digital marketing on financial results – becomes mandatory. You need a system that is equipped with tooling and mechanisms (like modeling, security, transformation, connectivity, lifecycle, monitoring, governance in general) that allows that semantic consolidation. This is exactly what a data warehouse does. BW/4HANA provides this infrastructure while S/4HANA focuses on certain business processes.
Example 2: Fashion Customer
The second example is from a fashion customer who sells his products predominantly via on-premise stores but increasingly online. The latter triggers the need to look into more and more online behavioural data, such as clickstream or social media information, in order to answer questions about the products for which a customer has shown some interest or what the brand perception is etc. Fig. 2 lists data sources that this company is analysing. One aspiration is that demand can be better predicted by better understanding a customer’s interest indicators from clickstreams and social media. That in turn can impact demand, supply and other planning in – e.g. – a BW/4HANA system.
Example 3: Oil and Gas Customer
The third example is from an oil and gas customer. Fig. 3 shows the data sources that they connect to their data warehouse. There is obviously a mix of SAP and non-SAP sources. For instance, there is data on seismic measurements, oil rig sensor information, drill status (both for predictive maintenance), oil well status etc. Again, there is a number of scenarios or analytic questions that require to combine such data with data from an SAP system. To that end, a data warehouse approach is required. Simply copying such data into the HANA system underlying an S/4HANA instance would fall short in many ways: you would still end up creating a data warehouse on HANA that coincidentally sits on the same HANA as the S/4HANA instance.
These 3 real-world examples show that modern analytics requires data from an even larger variety of data sources than ever before. Big data, IoT, digitalisation etc. are trends that have added to that variety. Integrating data from those sources is more than just copying them together to or exposing them logically in one location. The need for a data warehouse remains as the place that brings the data together (physically or logically) and semantically integrates them through transformation, harmonisation, synchronisation etc. This is complemented by operational analytics inside a single operational system, such as S/4HANA, as it focuses and analyses data in there in an isolated way.
Hasso’s SAPPHIRE NOW 2017 Keynote Comments
Hasso commented in his SAPPHIRE NOW 2017 keynote (see here, at 0:37 to 0:39) that “he fought against data warehouses in the 1990s”. However, he also states that “there is still an application for data warehouses”. He then elaborates that not all analytics does have to sit in a data warehouse.
This is exactly the distinction and the point argued in this blog, namely: there is operational analytics (directly inside an operational system like S/4HANA and not necessarily in a data warehouse) and there is cross-system analytics (which needs something like a data warehouse). The latter is a problem that is not addressed by S/4HANA but that exists in the real world – see the customer examples above – and that is addressed by BW/4HANA.