“The customer need is more important than the technology”

Thanks to the Internet of Things, processes can be digitised and thus often made more efficient thanks to data exchange via sensors and applications. Dominik Schatzmann is a professor of Industrial IoT at BFH. In an interview with the magazine Technik und Wissen, he explains what to look out for when evaluating sensor and actuator data in the cloud.

How does the data get from the field level to the cloud?

There are different ways, but you often see the following elements. First, there is the sensor that generates the data. From there, it goes via a gateway to the cloud, where it is stored in a data lake. An analytics layer then generates usable information from the stored data. These insights are then integrated into a business process with the help of a process engine, from which results can be derived.

What criteria must be fulfilled for the digitisation of a process to make any sense at all?

Dominik Schatzmann is a lecturer at BFH specialising in Communication Systems, Security, Data & Analytics.

The connection of the insights gained through the collection and aggregation of data must bring tangible added value for customers or the company, be it a reduction in costs, an improvement in quality or the creation of a new type of customer experience, for example.

Digitalisation pursues different goals depending on the industry. Is there nevertheless a common denominator?

The common denominator in digitalisation is the customers or the users. For example, if a process is to be simplified, the hope is that savings can be made that can be passed on to the customers. Of course, this is my personal opinion and others may see it differently, but for me, the customer is at the centre.

To what extent must a process to be digitalised be seen from the customer’s point of view?

This customer perspective is extremely important, because otherwise there is a danger that you will be guided in your approach to a solution by the technical possibilities and not by the problem that needs to be solved.

Can you illustrate this with an example?

Let’s take a machine component that is monitored via IoT. In order to plan its replacement and procure the spare part, a certain lead time is needed. In other words, the problem cannot be solved, or not optimally, with the current data evaluation. Therefore, an attempt must be made to predict what will happen in the current week with the data from a week or a fortnight ago. A real-time streaming layer is not necessary in this application because the event that indicates the failure has already been recorded days or even weeks before.

What are the typical challenges in digitising an end-to-end process?

What I just mentioned. The problem has to be analysed from the customer’s side to identify the true needs and their elements. If you are guided by the existing technology, you may develop a system that does not solve the problem at all.

What else should be considered when defining a process?

The simplicity of the solution is crucial. Ideally, it should be based on the skills of the employees and the IT department. Let’s stay with the example of the discussed failure of a component. If you have thought about it properly, you will realise that you do not always need complex real-time analytics to improve the process. This task can often be solved with a simpler batch job that provides the recorded information once a day, for example.

Modern sensors generate a lot of data. How do you know which data to use in the first place?

This is a big challenge of analytics. I have seen many projects where data was searched for days without a clear result because there was simply too much. My personal advice is to work backwards with domain experts from the business problem and start with a small set of data. If we stay with the example of component failure, their developers can best explain which parameters are indicators of impending failure. This speeds up the process, as the data scientists do not have to research and find out these correlations themselves.

What needs to be considered with regard to data protection?

Right from the start, attention should be paid to privacy by design and security, i.e. that personal data is protected from attacks or intrusions. As a rule, competent contact persons can be found within a company or organisation who can support the design and protection of a process.

Assuming that a company does not find the necessary support within its organisation, what should it pay attention to when dealing with personal data?

Ideally, the data should be anonymised before it is used. Another aspect is the reliability of the data processing systems. This also includes that they are maintained and receive regular updates.

To what extent must a system be future-proof so that it can be easily adapted to new requirements?

First of all, it should be clarified which features of a system can be transferred to another one and reused there without much effort. Once this has been decided, the system must be built in such a way that the units or components can be easily separated. Crucial to this are the interfaces, which must be designed accordingly.

In your view, what else should definitely be said about this topic?

For me personally, the interaction between the clients, the technology and the employees of the service company is important. In order to achieve a good result, these three elements must be brought into harmony.

About the person

Dominik Schatzmann studied electrical engineering and information technology and holds a doctorate in communication systems from ETH Zurich. He has held various positions within Swisscom, most recently as Principal Product Manager in the area of Data, Analytics and AI, responsible for the channel analytics portfolio of the Swisscom mass market. He is a lecturer at the BFH with a focus on Communication Systems, Security, Data & Analytics and Linux.

Learn more

This article was published in Technik und Wissen and the author is Markus Back.

Creative Commons LicenceCreate PDF

Related Posts

0 replies

Leave a Reply

Want to join the discussion?
Feel free to contribute!

Leave a Reply

Your email address will not be published. Required fields are marked *