What makes it so difficult to manage operations data and achieve AI success in batch process industries? There are many valid answers, and it’s crucial to address them all. In this white paper, we explore special issues around mindset, complexity, multiple disciplines, and a variety of dispersed data. We also address how these batch process issues intersect with the five main facets of data management.
Click here for the full report (no registration required), courtesy of our sponsor, Quartic.ai.
Table of Contents
- Executive Overview
- Taking Action to Get Value from a Flood of Data
- Building a Data Foundation
- Facet 1: Data Collection and Extraction
- Facet 2: Data Harmonization and Storage
- Facet 3: Contextualization
- Facet 4: Analytics
- Facet 5: Applications
- Consider a Complete Platform
- Recommendations
- References
- About the Author
Executive Overview
Urgency has never been greater for makers of biopharmaceuticals and consumer products to be faster, more efficient, and more confident in serving customers. As experts retire and regulators increase pressure, this need grows. Companies are look to Industry 4.0 technologies for digitalization and ideally predictive approaches to improve performance. At the heart of whether they succeed is their ability to manage operations data effectively.
Many companies are experimenting with new technologies, particularly advanced analytics such as machine learning (ML), but most cannot build scalable and sustainable solutions. This raises a host of questions:
- Why are so many projects failing to meet expectations?
- How can we get past proofs of concept (PoCs) and move on to enterprise-wide digital transformation?
- Beyond failing fast, can we finally move on from those experiments and succeed?
Many projects fail because there are needs that are not obvious. Creating the industrial data foundation for artificial intelligence (AI), ML, or other new technologies is possible but has several stringent requirements. Batch process manufacturers lag some others for various sound reasons; these companies:
- Have a mix of digital and non-digital assets
- Operations data vary in provenance and usage
- Have to satisfy a wide variety of stakeholders by discipline and level in the organization
- Have many types, formats, cadences, and volumes of data to bring together: from operational technology (OT), IT, and business partners such as suppliers
- Need to contextualize a set of varied data streams differently for various stakeholders, such as asset, product, process, or time
- There are thousands of use cases and applications to satisfy, and even in a moderate-sized organization, many of these situations arise unexpectedly
Data management has five main facets: collection and extraction, harmonization and storage; contextualization; analytics; and applications. Because of the complexities, nearly every batch manufacturing company is missing some elements of some of these facets today. They cannot succeed without a reasonably complete set of them.
Fortunately, there are solution providers with a broad footprint today. Many use an array of modern technologies that can interoperate with most IT and OT environments. A few data management platforms have grown up in the batch process industries and focus on enabling all five facets for their customers. Finally, we can begin to tackle these issues with some commercial support
Recommendations
Based on industry experience and research for this report, Tech-Clarity offers the following recommendations:
- Focus on Value: Set your sights on how to stay focused on optimizing our business, allowing solution providers to deliver the industrial data management platform.
- Map It: Operations data management is a journey each piece building on the next. It pays to map it out in advance and get prerequisites in place at each step.
- Set Scope: Set small goals and craft projects that deliver business value and some piece of the foundation simultaneously. More than a one-off PoC, but not ”boiling the ocean,” and always with a long-term view.
- Commit Big: Don’t expect a silver bullet – this is hard work, just as ISA88 and 95 were; each facet is likely to be that much work
- Communicate: Educate everyone that getting to actionable insights requires all five facets of the data foundation to be robust and tightly integrated.
- Multi-level: Realize that this data foundation layer must include edge, fog, site, enterprise, and ecosystem layers.
- Ask Hard Questions: Don’t assume anything internally or with solution providers. Many projects make significant gains just in mapping the current reality and spotting issues to address immediately, even without new technology.
- Involve All: Form a team for the initiative with all stakeholders – IT and OT, production, quality, maintenance, site-level, enterprise-level.
- Partner: Don’t plan to do all the work yourself. You probably can’t afford the time, money, and problems of DIY experimentation.
- Test the Partner: If an industrial data management platform provider seems like a good fit, craft a proof of concept (PoC) in the context of a larger long-term strategy.
- Gain Benefits: Rather than fail fast, seek to succeed relatively quickly in getting to the next step with industrial data management
Click here for the full report (no registration required), courtesy of our sponsor, Quartic.ai. If you have difficulty obtaining a copy of the research, please contact us.