Blog
Creating impact through data science is hard. Doing that for an industrial company is even harder. Adopting data science practice to drive outcomes for GE has been a challenging, yet an extremely rewarding experience at the same time. It is well known that domain knowledge is important in applying data science to solve hard problems. But, for an industrial or an enterprise company it is not just important but extremely critical to understand specific domain and the related outcome. This allows generating insights with correct context for the targeted users (or decision makers) of those insights.
It is evident that data science as a practice and a field has surged in the past five years. This happened not just because of advancement in tools and technologies, but enterprises are realizing the complexity of dealing with large amounts of disjointed data systems in their organizations and challenge to figure out how to generate more value from the data they have or continue to capture.
There are many reasons why organizations employ data science practices. At GE, we see three primary scenarios where data science allows us to make an impact for our own goals and for those of our customers.
First, a customer has a large amount of data but doesn’t know how to extract valuable information. They have collected this data over many years, through various processes and systems, but don’t know how to use it in order to drive business value or additional productivity.
Second, and one of the most common scenarios, is where an organization has a specific problem or an outcome in mind and wants a data-driven solution to the problem. For an industrial company with critical assets to manage, this often happens when there is an engineering problem, but a purely engineering solution is not feasible through analytical modeling or traditional engineering techniques. In these cases, a data-driven approach may sometimes provide a more flexible and cost-effective solution. Another situation could be where the industrial assets generate non-sensor data such as maintenance records, field problems, work orders, and operator usage. In this case, an automated analytics solution is well suited to generate insights from wide variety of data.
Lastly, data science plays a critical role for customers intending to start a digital industrial transformation journey. This involves multiple outcomes and use cases centered around data-driven decision making, increased and perhaps real-time visibility of key performance indicators, improved productivity, and uncovering new areas of product or business opportunities. Data science activities can often become the first step by bringing hidden insights from a multitude of data to the forefront for guiding the right mix of solutions as digital industrial transformation occur in an asset-centric organization. For example, data scientists can help identify key variables from all data streams essential to develop any key business metrics often in real-time. Further, more insights and knowledge of data can be used to define a roadmap for future analytics.
Whatever your organization’s motivation or need to use data science is, it is clear that data science, if done right and with correct context, has the power to deliver impact regardless of the industry. But, the story does not end here. There is lot more to be thought about as the industry matures. If you are a business leader of an industrial company who regularly relies upon driving outcomes and decision making through data science, you may be thinking, “How do I go beyond analysis and insights? How do I get continuous value?” Even if you are new to adopting data science in your enterprise, you should consider these aspects.
But, you can’t rely on data science alone to get continuous value, optimize processes, improve productivity, generate new revenue streams, and spur growth. You have to go from doing data science to building data products. And if you thought that building data products is hard in the consumer world, it is even harder for industrial companies. Even though the end result is a software product, there is a distinction between building a software product and a data product, which is delivered as software. Misunderstanding of this critical distinction leads to failed products before even they are developed.
There is a growing trend in democratizing data science, adopting machine-learning techniques to drive data products, and cultural change to find new innovative ways to develop products in a cross-disciplinary team.
In the next part of our data science blog series, I will describe these trends, define data products, identify unique challenges of developing data products for industrial companies, and share lessons that can be applied to develop industrial data products.