Organizing and establishing data science and machine learning within the company: evolution instead of revolution? Dr. Lars Schwabe, head of the Data Insight Lab, provides his answers to this question in order to explain how companies can transform into data-driven organizations.
Becoming a smart enterprise with big data
Lufthansa Industry Solutions was recognized as a relevant player in the digital factory in the three use case clusters “Predictive Analytics & Maintenance”, “Traceability” and “Asset & Plant Performance Monitoring” by teknowlogy. According to teknowlogy, the leading independent European market analysis and consulting company for the IT industry, LHIND has proven to be able to address all related use cases in these categories.
Big data technology has arrived in companies, but its installation is not the end of the road to becoming a data-driven, digitalized company. What makes a company a data-driven company? How can data scientists and engineers take their achievements “to the streets”? Being successful as a data-driven company usually requires a change in corporate culture that combines empirical and entrepreneurial thinking. This can be achieved when data scientists at the company are able to disseminate their approaches toward thinking and solving problems, both in the departments and in IT. To do so, they have to create the proper organizational framework.
There is a broad range of ways to evolve from a company into a smart enterprise. Whereas some firms have only just begun their journey to becoming data-driven companies, others have had a strong emphasis on data for some time now and are already experimenting with new business use cases to complement their existing business and to dominate in the long term.
Using data science to develop new business use cases
Sometimes, companies already have the technologies required in-house, often as a result of initiatives from an IT department that has taken a proactive approach to its new management role. But technology alone does not make a company a smart enterprise. In some cases there is talk of companies investing in big data technologies and only then asking the question: “And what do we do with it now?” It is here that data-savvy business analysts come into play – people with a good sense of what can be achieved with the help of data science and machine learning methods. This gives rise to new business use cases that can be tried out, for example, in proof-of-concept implementations or in laboratory-like environments.
One conceivable scenario typical of project work is converting successfully tested business use cases into IT back-end systems and implementing them in operations. A data-driven company then views and continuously analyzes its own data. It monitors business use cases and develops new ones. And they have to be tested in the real world: Only cost-effective business use cases that function both technically and methodically will survive. But this is nothing particularly new. The fundamentally new aspect in this regard is that, today, this can all take place with very good data available for the quantitative assessment of the situation – and this at very high speeds.
Data science departments as company staff units
Successful researchers are very familiar with the following operation: formulating a hypothesis, collecting data, analyzing the data, coming to and then utilizing analytical conclusions. Data scientists and machine learning experts are the researchers of the company. But what is the best way for them to establish their experience within the company? In this regard the following thoughts: The establishment of data science and machine learning within the organization must be thought of as a series of establishments. Those who have a long-term objective in mind – for example company-wide, data-driven thinking – should choose the steps they want to take to achieve it carefully. For example, where could a “data science” department be set up? As a staff unit? In the departments? In IT? Why not begin with a staff unit that is well networked within the company and then integrate the data scientists into IT or the departments after a few years? LinkedIn embarked upon a similar journey. There, one of the first “data science” departments was ultimately dissolved and its knowledge has migrated into the departments along with its staff.
Data science is a transient topic. If data scientists and software manufacturers who create the appropriate tools are successful, “data science” departments become superfluous. Those who still have their own “data science” departments in five years’ time have failed to digitalize. The empirical thinking that data scientists exemplify should be imitated and internalized.
Machine learning is essential for data-driven companies
Machine learning is a task for IT departments. Whereas analysts are chiseling out and driving predictive analytics and next-generation business intelligence (BI), the productive utilization of machine learning is dependent on developers who primarily see themselves as such, but also have an eye for the essentials in the domain of machine learning. Methodically speaking, there is a smooth transition between business intelligence, and data science and machine learning. However, machine learning is certainly more IT and informatics based in at least one aspect, which is its strict emphasis on mechanization and automation. We are only now finding out what the tasks will be for applied machine learning. However, we are already seeing that the combined expertise of developers and DevOps enthusiasts with an eye for the essentials of machine learning will be essential for data-driven companies.
This article by Dr. Lars Schwabe was recently published on euroforum.de within the scope of the CeBIT Innovation Day “Becoming a smart enterprise with big data” in 2016.