Enormous amounts of data are generated in the Industrial Internet of Things from sensors on machines and systems or smart products. It is not always entirely clear what should happen to the digital capital. Five dimensions can provide clarity here.
Although the first IoT scenarios have been successfully implemented, this can only be the beginning, given the available amount and variety of data. Many manufacturing companies still lack a clear idea of consistently using data stocks and developing into data-driven companies. This is confirmed by the figures from the Bitkom study “Industry 4.0 – Germany’s factories are so digital” from 2020: 88 per cent of the companies surveyed stated that they were developing products and services for platform-based business models as part of Industry 4.0 initiatives. In contrast, only 18 per cent developed products and services for data-based business models.
In this respect, companies should get clarity regarding their data handling to set the right framework. This applies to the organisation, the processes and the technology – especially the analytics architecture. It is worth looking at dimensions: interest in knowledge, data, users, costs and system landscape.
Dimension 1: Interest In Knowledge
The cognitive interest can be understood as the foundation for the other four dimensions. It is also crucial for the motivation to deal with the handling of data. First of all, it is instructive to look at the degree of maturity.
So far, data analysis has been limited primarily to condensing data from the past into key figures or key performance indicators ( KPI for short ) and compiling them in reports. For example, you can see how high the overall equipment effectiveness was – in a specific period, at a specific location, for a specific product. This descriptive analytics provides answers to the question of what happened. Evaluations can now also be used to answer why something happened (diagnostic analytics), what will happen (predictive analytics) and how companies can make something happen ( prescriptive analytics). With this increasing degree of maturity, the degree of implementation of a data-driven culture also increases, which leads to an increase in the added value of the entire company.
Each company must decide for itself which level of maturity makes sense. This also applies to the field of action in which comprehensive knowledge is worthwhile. A few general assumptions about what is particularly relevant for manufacturing companies can still be made.
As a supplement to the Key Performance Indicator Overall Equipment Effectiveness, it is possible to diagnose how the result came about and the reasons for particularly poor or particularly good performance. This knowledge can be used to define suitable measures. Analysis methods also make it possible to predict or simulate how these measures will probably work.
Production Planning And Control
Today, production planning and control are based on data that is as comprehensive as possible and on tools that bring resources and orders together. In some cases, feedback from the production area is taken into account. With new analysis methods, significantly more data can be processed and related to each other across all phases of planning and control. This makes it possible further to optimise details – such as the tape speed. People in an artificial intelligence delegate can also make decisions. All in all, this increases the utilisation of available resources, shortens throughput times and helps to meet commercial requirements. The vision of batch size one production at competitive prices can thus become a reality.
Product Development Product
Data used by customers can be condensed into valuable insights into quality and usage. The findings can help product development to optimise products continuously.
Dimension 2: Data
On the one hand, the interest in knowledge largely defines which data is required for the analysis. On the other hand, an overview of the available data provides the impetus for identifying useful use cases and limits it to what is feasible. To gain clarity, it is advisable to use Gartner’s 3V model as a guide:
- Volume: What data volume should be processed?
- Variety: Which data sources (e.g. SAP systems, non-SAP systems, IoT devices) and which data types (structured, unstructured) should be processed?
- Velocity: What speed should be achieved when processing the data?
Dimension 3: Users
Analyses are only useful if the users work with them. In many companies, this was not always the case. With a lot of effort, umpteen Excel reports were created, gathering dust on the hard disks unread. Under which conditions the users derive the greatest benefit from the analyses should be clarified. This increases acceptance and frequency of use. Above all, requirements can be derived from what certain user groups want to use the data.
Dimension 4: Cost
Costs also play a role when deciding on architecture. Identifying all costs and assessing them is not at all. On the one hand, several factors must be taken into account. On the other hand, the costs must always be set concerning the benefit that is realised and expressed in monetary terms:
Capex Or Opex
A few years ago, it was common to buy an IT solution and pay its licence costs. So there were CAPEX costs. There are now several subscription models in which the software is not purchased, but the use is paid for – and thus, Opex costs arise.
In addition to the costs for the IT solution itself, there are costs for operation. How high these depend largely on who takes care of the software and how intensively.
With the help of the analytics architecture, different use cases are implemented to bring added value. For example, costs can be reduced, sales increase, new products and services are created, innovative business models are established, and improved user experience. How high the added value will be can rarely be said exactly in advance. Nevertheless, an approximation is indispensable for a well-founded decision.
Dimension 5: System landscape
Finally, requirements can arise from the existing or desired system landscape. On the one hand, this applies to deployment: on-premises, cloud or hybrid? On the other hand, it is about the architecture’s scalability – in terms of functions and performance.
Embedded Analytics Or Enterprise Analytics?
Dealing with the five dimensions contributes significantly to developing a data-driven culture. In addition, aspects that are elementary for the selection of an analytics architecture are clarified. There are two architectural approaches, each of which has its specific strengths and can also be combined: embedded analytics and enterprise analytics. With embedded analytics, data and functions are located in the transactional system – usually in the ERP system – in the case of SAP, i.e. in SAP ERP or SAP S/4HANA. With enterprise analytics, data and functions are stored in dedicated components outside the transactional system. Usually, a data warehouse is a backend and specific modelling, analysis, and visualisation solutions as a frontend. The two approaches can also be defined concerning the different levels of an analytics architecture: With the embedded analytics approach, a system comprises the levels of collection, storage and usage, while with the enterprise analytics approach, there are special systems f
or all levels that support a single point of truth.
SAP implemented the embedded analytics concept for the first time with SAP S/4HANA. Analytics functions are embedded here that access the data stored in the HANA database. At the same time, SAP continues to offer dedicated components for an enterprise analytics approach – including SAP Analytics Cloud (SAC), SAP BW /4HANA and SAP Data Warehouse Cloud (SAP DWC). Both approaches have some specific characteristics.
The task now for companies is to reconcile the characteristics of the approaches with their intended handling of the data. For example, if data from production is to be used to improve production planning and control further, the embedded analytics approach is most suitable. To analyse the enormous amount of data from products in use, only the enterprise analytics approach can be considered simply because of the connection of sources.
What is important here is that companies do not create any irreversible facts by deciding on an architecture. Because the individual components build on each other, if you conclude today that the embedded analytics options of SAP S/4HANA are sufficient, you can introduce the SAP Analytics Cloud as a supplement tomorrow. And for those who operate an elaborate enterprise analytics architecture with SAC, SAP BW/4HANA and an additional data warehouse in the cloud, embedded analytics still makes sense in certain use cases.