Transitioning from Reactive to Proactive Management: Capitalizing on Data Utilization for Factory Floor Optimization

Global manufacturing relies on factory floor data to assess the health of production lines, monitor Key Performance Indicators (KPIs), and streamline operations where applicable. Producing and capturing data has never been an issue, as manufacturers produced over 2 Zettabytes (ZB) of data in 2025, with data produced expected to exceed 4.4 ZB by 2030 (see ABI Research’s Industrial and Manufacturing Survey 2H 2024/1Q 2025: Data Utilization in Manufacturing (PT-3722)). The vast majority of these data are produced by sensors, edge devices, industrial robots, Radio Frequency Identification (RFID) tags, and Programmable Logic Controllers (PLCs), indicating that manufacturers have sufficient smart infrastructure to provide foresight into production operations.

While manufacturers produce copious amounts of data, a clear issue arises about how data are being used. Only 5% of factory floor data are processed for insight generation and value-added use cases. Meanwhile, 95% sits in cold data storage, never to be used. With high data production, but low usage rates, discrete and process manufacturers are missing out on operational gains that translate into more efficient production.

Data utilization is more important now than ever before. With the rising prominence of Generative Artificial Intelligence (Gen AI) and Agentic AI initiatives such as Industrial-Grade AI from Siemens and Nexus Black from IFS, proper data architecture that includes collection, standardization, and contextualization will become mandatory.

In 1Q 2025, ABI Research conducted a multi-client survey that assessed the technological development, capabilities, sentiment, and challenges of both discrete and process manufacturers. Based on the survey, 65% of discrete manufacturers and 68% of process manufacturers have the ability to run real-time analytics. Real-time analytics are imperative to managing the current status of production lines, but fall into the bucket of descriptive analytics. Descriptive analytics is decades old and is used as a reactive method to fix production issues as they arise. With modern data analytics, manufacturers have the ability to take the step from descriptive analytics to advanced use cases in the form of predictive and prescriptive analytics. 

The worrying statistic is that 31% of discrete manufacturers and 29% of process manufacturers have the capabilities to run prescriptive analytics. In the context of use rates for descriptive analytics, what this shows is that most manufacturers run into  impediments when stepping up the value chain from descriptive through prescriptive analytics. This impediment is fueled by a lack of understanding in the offerings from data analytics vendors, and what the prerequisite requirements are to prep data for advanced use cases.

Manufacturers are evidencing difficulties moving up the value chain from descriptive to predictive and prescriptive analytics, implying there may be further concerns about future AI adoption rates. Vendors with AI-based solutions that train models with customer data should pause to assess whether a manufacturer truly has a stable data foundation to support the accurate training of Agentic AI. If due diligence is not undertaken by both the manufacturer and vendor, the impact of Agentic AI will not just be imprecise responses that can be ignored, but rather costly errors that significantly affect the bottom line of the end user.  

Manufacturers have done the hard work of installing the relevant hardware such as sensors, PLCs, and other Internet of Things (IoT) devices to capture factory floor data. What has not happened is a transition from significant data capture to significant data usage. This has led to manufacturers missing out on core value drivers such as expected downtime management, root-cause analysis, and quality control. Not only are manufacturers missing out on the opportunity cost of not deploying predictive and prescriptive analytics, but they also foot the bill for unused data storage, cutting into profit margins.

Discrete and process manufacturers need to have a clear understanding of how they can profit from their own data. First and foremost, manufacturers must look for solutions that can standardize and contextualize data so that predictive and prescriptive analytics can be run. DataOps providers such as Crosser, HighByte, and Litmus Automation enable manufacturers to streamline and automate data standardization and contextualization by building out pipelines that connect factory floor machines and relevant software such as Manufacturing Execution Systems (MES) and Quality Management Systems (QMS). By connecting machines and software, Crosser, HighByte, and Litmus Automation can provide manufacturers with a centralized platform to manage, observe, and interrogate the data being produced on the factory floor.

After data are prepped in a centralized DataOps platform, manufacturers should look to data analytics providers such as Alteryx, Augury, Ekhosoft, and MachineMetrics. These vendors can utilize standardized and contextualized data from individual machines, full production lines, and multiple factories to derive predictive and prescriptive insights. Manufacturers that have successfully deployed solutions from DataOps and data analytics providers see significant reductions in machine downtime, increased batch quality, and productivity gains through efficient worker action.

Once a coherent strategy consisting of sufficient data collection, a DataOps platform, and a further analytics solution is built out, manufacturers will be ready to take the next step toward deploying Gen AI and Agentic AI. One of the largest impediments facing Gen AI and Agentic AI adoption by manufacturers currently comes from a lack of trust that answers will be reliable. With proper data governance, organizations in both discrete and process industries will be able to run predictive and prescriptive analytics, along with deploying Agentic AI models that operate within an acceptable degree of accuracy.