Alexander Thamm, Founder and CEO of the Data Science and AI consultancy Alexander Thamm GmbH and Dr. Alexander Borek, Global Head of Data & Analytics for Volkswagen Financial Services AG, introduced the Data Festival 2018 by providing 10 Key Principles of how to become a data-driven leader. These two experts discussed the common pitfalls of digital transformation for (traditional) companies.
Rule 1: Accept that easy digital transformation is a myth.
While many (digital) start-ups already have Big Data and Analytics in their DNA, many more traditional, pre-2000 companies struggle with digital transformation. This development mainly started in Germany about 2-3 years ago, but, many CDOs are now confronted with a major dilemma.
In traditional companies, the DNA of the firm has to change, which means that “what you do and also how you do it” becomes a fundamental question and major changes turn out to be necessary. However, it is not easy to change things about the company that people are familiar with. CDOs and other leaders of change within the company have to deal with one major dilemma: On one hand, they have to persuade employees that change is necessary and offers great potential and, on the other hand, they have to be honest and accept that this is not going to be an easy step.
One way to do this is by presenting scenarios that demonstrate the potential of data across the company. However, all these (data-driven) business models have to be implemented into the core of the company once they have been developed in a lab. This step requires extensive change management. It is also important to distinguish between potential and the actual realization of that potential. In the same context, it is necessary to be brutally honest about the fact that digital transformation is going to be tough.
Rule 2: Demonstrate how basic beliefs in your industry are being turned upside down due to digital disruption.
There are some basic beliefs in companies that need to be rethought in the context of digital transformation/disruption. One of these beliefs or paradigms is the lack of investment infrastructure in Germany and in the EU. This is particularly true for traditional, non-start-up companies. So, what does this mean?
In Germany, many companies are focused on their return on investment (ROI). While a focus on a quick return on investment might have worked well in the past, many digitally connected companies such as Zalando or Amazon show that high initial investments are necessary. These companies invested large amounts of money over a long period of time before achieving an adequate ROI. However, this investment strategy turned out to be a crucial step on their way to success.
In order to achieve digital transformation into a similarly digital company, traditional enterprises need to push their comfort zone a little. So, what does a little mean, and how much is too much? For example, it does not go down very well to tell your employees that their work will soon be automated and executed by a robot. A more successful approach would be to illustrate that time-consuming use of excel sheets can be replaced by an algorithm, which allows employees to leave earlier and play with their kids.
Rule 3: Share the simple equation: Digital = Data + X.
This “Digital is Data + X” equation sounds like a dummy equation. However, it can be interpreted in two different ways, both of which are of crucial importance for successful Data Science projects.
First of all, the equation expresses that if you want to pursue digital transformation, you need both data and analytics. This might seem like common sense, but is not clear to everyone. Many people are thrilled by the idea of implementing all the “cool data stuff”, such as recommendation systems, data lakes etc., but not everyone is aware of the data and analytics that are actually behind all this, and huge investments, e.g. data lakes, are necessary to implement the cool stuff.
The other interpretation of this equation means that you have to control the X. If you do not do that, you can develop as many algorithms as you want, yet you might never be able to implement them. According to the experts, many people, including graduates in IT programs, imagine that Data Science is basically Machine Learning. However, this is not very realistic. There are many important things that also need to be taken into account.
Dr. Alexander Borek gives an example of a great recommendation algorithm that was developed by several experts over 6 months. However, this great algorithm was never able to be implemented into the website as it did not fit the data pipeline. Therefore, the development of the algorithm was basically useless.
For illustration purposes, Alexander Thamm compares Machine Learning to cinnamon: It is a great spice that can significantly improve some dishes. However, if you eat pure cinnamon, it is not exactly enjoyable – it requires other ingredients to become delicious.
So, how should these two issues be handled? Basically, it is helpful to educate people / employees and demonstrate the value of data, but at the same time, it should be emphasized that data is not the only part of a data product.
Rule 4: Train your existing workforce in Data Analytics – everyone can learn it.
On their way to becoming a digital company, many companies today are facing a lack of Data Scientists, and a much greater lack of Data Engineers. How can companies deal with their need for Data experts? Instead of hiring new people from the market, a more practical approach might be to train your own workforce, as they already know the internal processes and specific culture within the company.
For example, almost any employee, especially those with a quantitative focus in their job, can be trained in Machine Learning and Data Visualization. Even if they might not be the best data experts, they can leverage the value of data significantly.
In addition to this, it is important to train every employee to work with data products from a business perspective. Besides the training of employees, it is also crucial to engage them. This is especially important to avoid making people feel threatened or cut out by new technologies. This is part of the soft side of change management in the digital transformation – take people on board and have them play with new technologies.
Rule 5: Ask the board to delegate decision-making power to cross-functional data analytics roles and bodies.
In contrast to the soft part of change management, this rule draws on the hard part of change management. Most data topics cross functions within a company. For example, sales require data from quality management. If you do not have a cross-functional committee that steers changes and decisions, you have to consult the board with small and/or technical decisions on a regular basis.
In order to avoid this obstacle, it is highly recommended to establish a cross-functional committee that steers changes, decisions and functions as a pre-layer to the board in order to shorten decision-making processes.
Rule 6: Free your data.
Freeing data might be the biggest task of the mandate from rule 5 –someone is needed to free the data and make it available to everyone in the company who needs to work with it, while aligning with the allowances of the General Data Protection Regulation (GDPR). According to the two experts, GDPR can be a great opportunity for companies to “check and tidy the basement”. What data are we storing and what are we using? This question is important if you want to free your data.
Many projects fail because people only focus on collecting (or even hoarding) data. Then, issues can occur when someone tries to use this data, because knowledge about this data has neither been documented nor shared. In this example, the data was not collected correctly. This leads directly to rule 07.
Rule 7: Share knowledge on data.
GDPR is a great opportunity to start collecting data knowledge. In the context of GDPR, many companies found that they do not actually know much about their data. For some data, companies do not even gain access for its usage.
A common problem is that data is local, which means that local processes are involved and those local processes need to be understood in order to make sense of the data. To gain that understanding, a lot of people have to be consulted, which takes up lot of time. So, once you have made sense of the data, it is crucial to document it. Otherwise, everyone has to spend a lot of time on extensive research about the data. In summary, building data products requires an understanding of what that data means.
Rule 8: Automate Data Preparations.
Data Scientist = Data Waiter? That can happen, if you don’t automate your data preparations. You might think that this process is only necessary once – but it might have to be iterated every time new data comes in or if a mistake is identified in the existing data. What does this mean? Alexander Thamm recommends using Python rather than R. Why is that?
Once data use cases come into production, it is crucial to build a proof-of-concept, so that you have a theoretical foundation that can be used on the final product. Consider, for example, a client who requests reports all the time. 80-90 % of those requests could be automated and that would be the main use case. Sometimes, leveraging data does not necessarily mean developing a fancy face recognition algorithm or recommendation engine – it can also mean saving costs by designing IT processes more efficiently, e.g. by automating the data pipeline and as a consequence, drastically reducing licensing or labor costs.
So, what is the core message of this rule? Imagine a bottle that you fill with liquids and that can be used by everyone. This “bottle” may be filled with a customer retention score or another product that everyone can use, share and think of as a component. Something might only be a by-product for you, but it can be crucial for another department and be helpful for the digital transformation of your enterprise.
Rule 9: Embrace an AI and cloud first strategy, embrace open source.
Cloud first – whenever you start a new product, try to use a cloud. A cloud offers a safe environment for data. If you use a European cloud, it aligns with important regulations such as GDPR or banking regulations. It is even recommended to put in-house data in the cloud, because IT systems cannot keep up with Big Data Analytics and its innovations.
AI first – Whenever you think about developing apps, think of making it operable by natural voice and language instead of manual touch.
Open source – considering the high rate of innovation, innovative products often contain certain open source components. It is not necessary to go for free versions, though. However, most open source components are both innovative and strategic.
These three recommendations can enhance companies’ digital transformation.
Rule 10: Balance Data Analytics, innovation and transformation.
The process of improving in data usage is not a linear path – that goes hand in hand with the first rule. Improvement can be achieved by an iterative process of pushing on different sides. These could include a strategic side, a lab side, the technological side, the PLC side, the prototype side, and so on. This is not an easy task, but it is a crucial step in the digital transformation of a company.
You want to learn more? Check out the full-length keynote here..