Predictive Error Identification at the Machine and Operator Level | On Time Edge

This global life sciences manufacturer is a producer of high-stakes medical devices where precision, scalability, and global regulatory compliance frameworks are paramount to success. The organization focuses on high-throughput, precision assembly and has operations spanning multiple facilities around the world with regulatory oversight from agencies like the FDA and EMA. The company needed a smarter way to capture, contextualize, and act on production errors, particularly those arising from machine faults and operator variability.

Fragmented data prevents real-time, proactive error detection

Despite having deployed core systems including MES, SCADA, and QMS, the manufacturer struggled to proactively detect and prevent production errors. Key pain points included:

• Disconnected data streams: Machine level performance data, operator logs, and quality events were siloed across systems and file types, limiting traceability.
• Manual root cause analysis: Identifying the source of recurring production errors required cross-functional teams to manually search logs, sift through audit trails, and interview operators often days after the event.
• Lagging KPIs: Performance metrics were retrospective and often disconnected from real time operations, making predictive insight impossible.
• Scaling constraints: These issues created bottlenecks that threatened to scale along with the company’s growth trajectory.

Shattering data silos: Proving the effectiveness of unified data to drive predictive, scalable improvements

To address these challenges, the manufacturer partnered with On Time Edge and conducted a minimum viable product (MVP) initiative leveraging the ZONTAL data platform. The goal: ingest and contextualize machine, operator, and quality data in real time to reduce time-to-detection, accelerate root cause analysis, and build predictive models that scale. The effort incorporated four elements:

Data ingestion and contextualization — ZONTAL’s semantic data platform was deployed to unify structured (machine sensors, MES) and unstructured (operator notes, deviation reports, maintenance logs) data into a single, contextualized framework enabling near real-time access to a digital twin of the production process.

Edge-enabled integration — The On Time Edge team engineered secure, lightweight edge services to stream data from disparate sources including PLCs, operator terminals, and QMS databases into ZONTAL. Leveraging the ISA-95 model, On Time Edge is aligning data to equipment, operations, personnel, and quality events for high resolution traceability.

Root cause intelligence layer — The MVP tracks four critical data signals:

• Machine fault codes (structured SCADA/MES)
• Operator log entries (unstructured HMI/QMS)
• Resource availability and changeovers (MES)
• Non-conformance reports (unstructured QMS)

These are harmonized and visualized to detect patterns and causal relationships, enabling rapid identification of whether an issue stems from mechanical, human, or procedural root causes.

Predictive analytics pipeline — Initial AI models have been trained using historical and in-flight production data to flag early warning signs such as an operator repeatedly encountering faults during a specific shift, or a machine requiring maintenance outside of planned intervals.

Solid proof: Unified data model harnesses machine and operator data for real-time clarity to predict and prevent production errors

• Faster root cause identification: Errors that once took 48–72 hours to triage, flag and analyze reduced to hours.
• Improved operator coaching: Contextual dashboards help supervisors spot operator-level error patterns to proactively intervene with coaching and support.
• Reduction in repeat errors: Pilot lines are armed and focused on decreasing repeated error codes within 60 days of implementation.
• Scalable framework: The unified data model and edge-based ingestion pattern is being designed for rapid replication across additional lines and facilities.

Looking to the future, as the initiative progresses, the focus is on refining predictive models, integrating KPI-driven alerts for shop floor leaders, and embedding contextual intelligence into scheduling and resource planning systems. The long-term vision is a self-learning manufacturing environment that continuously improves performance, quality, and team productivity.

By combining On Time Edge’s deep integration expertise and capabilities with the real-time contextualization power of ZONTAL, this manufacturer is moving from fragmented, reactive operations to a predictive, scalable model of manufacturing intelligence. The MVP lays the groundwork for a factory of the future — where data doesn’t just report what has happened, but also shows you what’s about to happen.