Manufacturing Intelligence: from isolated data points to clear decisions.

Manufacturing intelligence means consolidating and analysing production, quality and planning data into a single operational picture. Data from MES, ERP, metrology systems and spreadsheets is linked so that metrics such as OEE, scrap rate and process capability become continuously visible. The goal is shop floor transparency as the basis for data-driven manufacturing: faster decisions, fewer stoppages, more stable quality. metrologx connects existing data sources rather than replacing established systems.

An MES (manufacturing execution system) controls and monitors production at shop floor level: orders, machine states, quantities, times and quality data in real time. The difference between MES and ERP lies in time horizon and level of detail. ERP plans orders, material and capacity across days and weeks; MES executes those plans minute by minute and reports actual status back. When choosing MES software, look for open interfaces, modular design and SaaS operation.

OEE (overall equipment effectiveness) shows what share of planned production time actually results in good parts. The OEE formula is: OEE = availability x performance x quality. Availability captures downtime, performance captures speed losses, quality captures scrap and rework. An OEE of 85 percent is considered world class; many plants sit well below that. A reliable figure requires automated machine and production data capture rather than estimated times.

Improving OEE means addressing its three factors separately. Availability losses come from setup, breakdowns and missing material; performance losses from micro-stops and reduced cycle times; quality losses from scrap and rework. The first step is complete capture of downtime reasons, because without causes no measures can be derived. A manufacturing dashboard shows which machine, shift and fault causes the largest loss. The quickest gains usually sit in micro-stops and changeovers.

Production data capture (BDE) records order-related data such as quantities, times, personnel and fault reasons, usually entered at a shop floor terminal. Machine data capture (MDE) reads states, cycles and counters directly from the machine control, automatically. In practice the two interlock: machine data supplies the raw signal, production data the order context. Machine data capture software should support protocols such as OPC UA and connect older equipment via signal taps. Only both sources together produce a reliable OEE.

A manufacturing dashboard should show few metrics, but decision-relevant ones. The key quality metrics are scrap and rework rate, first pass yield, process capability indices (Cp, Cpk), complaint rate and open deviations. On the production side, OEE, downtime reasons, on-time delivery and lead time complete the picture. Digitalising shop floor management replaces the daily whiteboard round with current data. What matters is that every metric has a clear owner and a defined response.

Statistical process control (SPC) monitors manufacturing processes using samples and statistical measures, in order to detect deviations before defective parts are produced. Its core tool is the SPC control chart: measured values are plotted over time and compared against control limits derived from the process's natural variation. If a value breaches the limits or a trend emerges, the process is corrected. SPC software automates control charts, capability indices and alerts.

Process capability describes how reliably a process produces within specified tolerances. Cp relates the tolerance width to the process spread; Cpk additionally accounts for how the process mean is positioned. Cpk is calculated as the minimum of (USL minus mean) / (3 sigma) and (mean minus LSL) / (3 sigma). The automotive industry generally requires a Cpk of at least 1.33, and 1.67 for critical characteristics. Capability analysis is part of process approvals and audits.

The difference between Cpk and Ppk lies in how variation is calculated. Cpk uses short-term variation within samples and describes what the process can achieve under stable conditions. Ppk uses total variation across a longer period, including shifts between batches, shifts or tools. If both values are close, the process is stable. A markedly lower Ppk points to systematic influences such as tool wear or temperature drift.

A CAQ system (computer aided quality) bundles quality management tasks in one application: inspection planning, incoming and in-process inspection, SPC, complaint management, gauge management and documentation. Classic CAQ software is often designed for large enterprises and correspondingly heavy in rollout and licence cost. For mid-sized manufacturers it pays off mainly when customers demand audit-ready evidence or measurement data from several systems has to be consolidated. Cloud-based solutions can be extended step by step.

Measurement data management means holding all measurement data centrally, from coordinate measuring machines, inspection stations, hand gauges and supplier reports, in one data model linked to part, characteristic, batch and timestamp. Spreadsheets reach their limit quickly: consolidating measurement data from different systems is manual and error-prone, versions diverge, analyses are not reproducible, and audits require a complete history. Replacing Excel in quality management calls for software that imports data automatically and generates reports.

To evaluate coordinate measuring machine data automatically, measurement programmes export their results in a standardised format such as Q-DAS, CSV or XML. Measurement data management software imports these files, assigns them to parts and characteristics, and calculates statistics, trends and capability indices. Inspection reports can then be generated automatically, including PPAP documentation, first article inspection reports to VDA 2 and evidence for a VDA 6.3 process audit. Deviations surface immediately, not when the report is compiled.

Measurement system analysis (MSA) checks whether a gauge and its measurement process are accurate enough to assess the required tolerance reliably. Method 1 evaluates the gauge itself through Cg and Cgk; method 2 evaluates repeatability and reproducibility across several operators. The latter is the gage R&R study: it shows how much of the observed variation comes from the measurement system itself. Below 10 percent is considered capable, 10 to 30 percent conditionally capable.

A purchased part dimensional audit (Kaufteilmassaudit, KTMA) is the systematic, recurring check of dimensional conformity of bought-in parts, going beyond standard incoming inspection. Instead of measuring only at first article or complaint stage, purchased parts are checked continuously and results evaluated statistically per supplier and characteristic. This makes it possible to monitor purchased part quality before deviations appear in assembly. Monitoring supplier quality and digitalising incoming inspection requires centrally available measurement data, automated analysis and audit-ready reports.

Detecting process deviations early only works if measurement data is evaluated promptly and automatically, not at the end of a shift. The basis is automated SPC: control charts and capability indices update with every new measured value, and breaches of control limits trigger alerts. Anomaly detection on production data additionally identifies patterns that fixed limits cannot capture, such as gradual drift across several characteristics. This shortens the time between deviation and response, reducing scrap and customer complaints.