Only DataOps Can Drive True Industrial Transformation
Industrial DataOps may be the key to securing your ultimate industrial transformation success.
- By Geir Engdahl
- July 25, 2022
In May, Germany earmarked $220 billion to fund industrial transformation. It’s a clear sign of intent for this industrial-leading nation, marking a move away from legacy and towards greater efficiency, sustainability, and safety among the heavy asset players. Although Germany’s industrial progress will naturally be stimulated by this ample funding, time has taught us that transformation only truly sticks when the workforce goes along with it.
The biggest mistake German industrial operations can make is to pour their funding into one-off projects or buy a batch of robots for a single use case. The industrial companies that have succeeded with their own reinvention are the ones who have seen the bigger picture, who have looked at the totality of their operations and recognized the opportunity to scale solutions. They understood the power of using their existing data to drive their own transformation, extracting it and operationalizing it to reap value.
This idea of a data-driven reinvention of industry is what we at Cognite call industrial data operations (industrial DataOps). It’s a discipline that gives the workforce the tools, processes, and organizational structures to transform the company into a more data-focused one. It requires collaboration, the collapse of silos, and the widespread availability of shared and contextualized data -- data that is accessible and understandable for a wide range of workers.
DataOps Can Eliminate the (Data) Time Drain
According to a 2019 IDC survey, only one in four organizations is analyzing and extracting significant value from data. This shortfall is caused by data that is dispersed and remains trapped in silos when the company lacks the tools and processes to bring the data together and make it meaningful. This is a time drain, as workers waste around 90 percent (according to IDC) of their time searching, preparing, and governing the data rather than using it for transformative purposes.
Industrial DataOps serves as a much-needed methodology to industrialize data management and the data analytics value chain. It applies automation and agile methods to the data life cycle, which in turn boosts time to value as well as the overall quality, predictability, and scale of data analytics. I see three key steps to adopting DataOps in an industrial setting.
Step 1. Make industrial data available
The industrial reality today is made up of many systems, so finding data can be a tedious and often dead-end process. Bringing it together is the first step towards industrial DataOps. This requires the integration of IT, OT, and engineering data, as well as other data sources that may exist, and unifying the data into a single platform. Quality is critical and companies require assurance that the data is correct for it to be usable, hence my next point.
Step 2. Make the data valuable
Once the data is integrated in one place, contextualized, and made securely available, explorable, and actionable for all types of users, only then can it be called useful. Contextualizing the data is about connecting it and establishing meaningful relationships between the different sources. This capability should be at the core of a DataOps platform, otherwise it will be impossible to get value from the data.
Maximizing data value requires using advanced models to generate insights for decision-making. These models are a combination of data science and physics, complemented with machine learning and deep learning. Even greater value emerges when users can develop low-code or no-code applications using the platform, lowering the threshold for moving from insight to action for users of all skill types.
Step 3. Give data to everyone
DataOps, when deployed across the operation, can drive democratization of the actual data, making it useful for both domain and data workers and enabling them to develop and innovate with it. It’s no longer only in the wheelhouse of the data scientists to do so because industrial DataOps enables even those without a data science degree to develop applications and drive the digitalization and transformation of their industries.
Think Big and Broadly
The key to industrial DataOps success is to not think small, but rather to think of everyone. Consider how to scale the digitalization use case before you even begin. Evaluate how other areas of the operation can benefit. Make sure that almost anyone can relate to the data, even if a worker isn’t part of the data team. Simultaneously, continue to strive for greater efficiency, safety, sustainability, and ultimately a more autonomous industrial operation.
Germany is off to a good start in terms of its billions in funding, but industrial DataOps will be key to securing their ultimate industrial transformation success.
Geir Engdahl is the co-founder and CTO of Cognite, where he runs the R&D department, leading the development of Cognite Data Fusion. In 2011, Engdahl founded Snapsale, a machine learning classifieds startup. He served as CEO and CTO until Schibsted acquired the company in 2017. Before that, he spent three years working as a senior software engineer for Google in Canada. Engdahl has a master’s degree in computational science from the University of Oslo.