Manufacturing Data & AI Glossary

AI that doesn't just analyze or recommend — it takes action toward a goal, querying systems, making decisions, and triggering steps with limited human input. On the floor, an agentic system might detect a quality drift, diagnose the likely cause, and open a work order or adjust a setpoint — not just chart the problem. It sits at the top of the Data Maturity Model (Stage 5, Autonomous) and only works on a foundation that's already connected, visible, and predictive.

A single, governed layer that links every plant-floor and back-office system — PLCs, SCADA, MES, ERP, QMS, and IoT — into one clean, trusted source of truth, structured so BI and AI run on it reliably. It's the difference between collecting data and being able to act on it, and it's Stage 2 of the maturity model — the part most "dashboard" vendors skip. See the full guide: What is a connected data foundation?

The rules, roles, and controls that decide who can access what data, how each metric is defined, and how quality is maintained. For manufacturers it means one agreed definition of a number, access limited to the right people, and an audit trail — which becomes non-negotiable in regulated sectors like pharma, food, and aerospace. Without it, a data foundation slowly degrades into unusable "dark data."

A modern architecture that combines the cheap, flexible storage of a data lake with the structure and query speed of a data warehouse. For manufacturers, it means raw sensor and machine data and clean, modeled business data can live in one place — ready for both dashboards and AI.

The automated route that moves data from a source system into your foundation — extracting, transforming, and loading it on a schedule or in real time. It replaces the weekly manual spreadsheet export. Reliable pipelines from PLCs, SCADA, MES, ERP, and IoT are what keep a foundation live instead of stale.

A central repository where cleaned, structured data from many systems is stored and modeled for analysis. It's one component of a connected data foundation — the place trustworthy numbers live so dashboards and models can query them. Often built on Azure or Snowflake.

A live virtual model of a physical asset, line, or process, fed by real-time data from the real thing. Manufacturers use twins to simulate changes and test "what-if" scenarios without touching production. A twin is only as good as the data feeding it — which is why it depends on a connected foundation underneath.

The system that runs the business side: orders, inventory, procurement, scheduling, and finance. It holds what was planned and what it cost — but not what's happening on the machine right now. Reconciling ERP with MES and the shop floor is one of the core jobs of a data foundation.

Extract, Transform, Load — or Extract, Load, Transform: the two common patterns for moving data through a pipeline. ETL cleans data before storing it; ELT loads it raw and transforms it inside the warehouse. Either way, it's how messy source data from your floor becomes clean, query-ready data.

The percentage of units that get through a process correctly the first time, with no scrap or rework. It's one of the clearest signals of process quality — and a number you can only trust if your quality data (QMS) is connected and clean.

Networked sensors and devices on equipment that stream data like temperature, vibration, pressure, and energy use. Manufacturers generate enormous IoT data — and most of it goes unused (commonly cited at around 90% in manufacturing). Capturing and connecting it is the raw material for predictive maintenance and real-time visibility.

An AI model trained on vast amounts of text that can understand and generate language. For manufacturers, LLMs power tools that turn documents, SOPs, and tribal knowledge into answers a team can query in plain English — especially when paired with RAG.

The system that tracks production as it happens — what's being made, in what order, against which work order, and at what rate. It's the floor's source of truth, but it often doesn't reconcile with ERP or quality data. Connecting the two is foundational.

The practices and tooling that keep AI models reliable in production: deploying, monitoring, and retraining them as conditions change. Without MLOps, a model that worked at launch quietly drifts and decays. It's what turns a one-off pilot into a system that compounds — Stages 4–5 of the maturity model.

The gradual loss of a model's accuracy as the real world moves away from the data it was trained on. New products, changed lines, or shifting demand all cause drift. Left unmonitored, a predictive model slowly becomes wrong — and trust erodes with it. Catching drift is a core job of MLOps.

The average time it takes to fix a failure and get equipment running again. It's a key reliability metric and a direct lever on downtime cost. Lowering MTTR depends on learning about failures fast — which is where real-time visibility and predictive maintenance pay off.

The standard measure of how well equipment is used, combining Availability × Performance × Quality into one percentage. It's the headline metric of manufacturing productivity — but only meaningful if the underlying availability, speed, and quality data are connected and trustworthy.

The percentage of orders delivered both on schedule and complete. It's the metric customers actually feel, tying production performance to delivery reliability. Hitting OTIF consistently takes a clear, connected view across the floor, the ERP, and logistics.

The rugged industrial computer that controls a machine or process, executing the logic that runs equipment. PLCs generate rich runtime, cycle, and fault data — but it's often trapped on the machine. Getting it off the PLC and into your foundation is step one for real visibility.

Using data and AI to forecast equipment failure before it happens, so maintenance is done just in time — not too early (wasteful) or too late (downtime). It's one of the highest-ROI uses of manufacturing AI, but it depends entirely on clean, connected sensor and machine data. Stage 4 of the maturity model.

Analytics that don't just predict what will happen but recommend what to do about it. Where predictive analytics says "this press will likely fail in 200 hours," prescriptive says "move the die change to Thursday." It's the bridge from forecasting to action.

The system that records inspections, non-conformances, scrap, rework, and corrective actions. It holds the truth about quality — but when it's a stack of disconnected spreadsheets, defects get spotted too late. Connecting QMS data is essential for FPY and for traceability, especially in regulated sectors.

A technique that lets an AI model answer using your own documents and data, not just what it was trained on. For manufacturers, RAG turns scattered manuals, SOPs, and a retiring veteran's notes into a system your team can ask questions of — with answers grounded in your real material.

The system that monitors and controls industrial processes in real time, sitting above the PLCs and aggregating their signals. It's a key real-time data source — and a key thing to connect into your foundation, so floor conditions show up in your analytics, not just on a control-room screen.

The inventory that's started but not yet finished — sitting between steps on the floor. Too much WIP ties up cash and hides bottlenecks; too little starves the line. Tracking it accurately needs connected MES and ERP data.