How is Predicitive Analytics changing existing processes and in which direction is the digital environment moving?
Arvato-expert Michael Freundlieb provides answers to the most important questions regarding predictive analytics and gives an insight into current processes.
1. For how long has Arvato Supply Chain Solutions been focused on the topic of business intelligence/predictive analytics? Why is the topic so important?
Business intelligence is important for us as an outsourcing service provider in many ways:
1. For one thing, the operational processes of our customers run largely on our live systems. This means that our customers risk losing transparency, and we are successfully averting that through EDI interfaces between our live systems and those of our customers, as well as through comprehensive business intelligence solutions.
So, we use business intelligence to restore the transparency of outsourced business processes. The key questions there are often: which services are we providing for our customers, of what quality, in what time frame and at what cost? In this way, we are also restoring our customers' ability to provide information to their end customers, for example with regard to the current status of an order.
2. As a service provider to many customers with similar businesses, we are able to pass on best practices and recommendations on business intelligence solutions to our customers. Because of the large number of customers who are served on a common business intelligence platform, we can also use state-of-the-art technology, such as our in-memory database, that would otherwise be unaffordable for our customers.
3. In addition to the added value for our customers, we also use our business intelligence solution internally for data analysis and to improve our processes, for example in the areas of logistics, customer service, financial services, account management and our own company management.
Since the start of 2017 we have built up a data science team within the Business Intelligence unit which is dedicated to predictive analytics, machine learning and artificial intelligence. Unlike the more “traditional” business intelligence analyses that tend to look at the past, the focus of the data science team is on making statements about the future through mathematical/statistical methods or simulations. Currently, we are using these procedures primarily to improve the planning of human resource allocations internally in logistics and in customer service by forecasting order, return and contact volumes. We also make our forecasts partially available to our customers as an additional service.
Further use cases in the area of fraud and churn prevention are currently being implemented and will be offered to our customers as an additional element of our service portfolio in the future.
Overall, we see huge potential in the field of data science for further internal optimization, as well as for additional data-based services for our customers.
2. Is Arvato a pioneer and an “opinion leader” when it comes to this subject, or is it just one of many? How does Arvato’s approach differ from that of its competitors?
To reiterate, Arvato has the unique advantage, as a service provider, of not only looking at one customer transaction but also bringing together extensive experience from a large number of similar customer transactions and combining it all into a best-practice approach. Therefore, in addition to providing the technical implementation of business intelligence/predictive analytics solutions, we can also give our customers advice: How do other customers do it? What are the key figures? How is the data prepared and made available to the various user groups in the best possible way? Which methods and technologies are suitable for making reliable forecasts in the area of predictive analytics?
Moreover, one of Arvato's key USPs is that much of the required data is already within the Arvato system landscape and can be connected using existing interfaces. In contrast to, for example, cloud-based, third-party business intelligence solutions, our customers no longer have to put their own effort into operating interfaces. You get the ‘all-inclusive package' from one place.
We run our business intelligence and predictive analytics solution for a three-digit number of customer businesses and have a large team working to implement and expand these solutions. This means that we are always up-to-date on new technologies and concepts and can implement solutions that would not be feasible for many of our customers by themselves.
3. Where do the particular challenges lie?
We have a number of exciting challenges ahead: For one thing, of course, there is the balancing act involved in offering low-cost, easily scalable standard solutions that also have to be adaptable and expandable according to the often very individual requirements of our customers.
Moreover, business intelligence is subject to constant technological change: fifteen years ago, it was still state of the art to create a set of standard reports in several hours each month and to make them available to users in a fixed and static form.
Today, data is updated at least daily, or is even real-time. Our customers and internal users want to work flexibly with the data themselves, not wait for it to be implemented by an IT department. Additionally, looking at the past is no longer enough. Instead, business intelligence solutions are increasingly being used to develop forecasts or even specific recommendations for action. And this is all being done, mind you, with volumes of data that are practically exploding.
For us, there is a constant challenge to keep up with the increasing demands, but also the ever-improving technological capabilities and concepts. But that's why business intelligence appeals to me: new challenges are constantly arising, but so are new ways of responding to them.
A permanent challenge that is often underestimated lies in data quality. User-friendly analysis tools, snazzy visual display options and sophisticated forecasting methods are pointless if the underlying data is technically flawed or misinterpreted.