A gear manufacturing plant in southern Germany checks every fiftieth shaft for dimensional accuracy. The sample has been within specifications for years. Nevertheless, a customer returned an entire batch because a press-fit process had been operating at the edge of the tolerance range for weeks. The error was visible in the process data. But no one had looked at it because quality control focused on the finished part rather than the process.
For years, the market has promised the opposite: camera systems, AI, and automation are supposed to eliminate defects entirely. The reality on the factory floor looks different. Quality control rarely fails due to a lack of testing technology. It fails because test results end up on paper, process data isn’t analyzed, and hours or days pass between when a defect occurs and when it’s detected.
Anyone who regularly visits production lines quickly recognizes the pattern: The companies with the lowest defect costs do not have the most inspection stations. They have the shortest response time between the occurrence of a deviation and corrective action. That is exactly what modern quality control is all about.
This article clearly defines the term, distinguishes quality control from quality assurance, compares common inspection methods, identifies the relevant standard requirements from ISO 9001:2015 and IATF 16949, and outlines a five-step process for manufacturing companies to transition from downstream inspection to process-integrated control.
THE MOST IMPORTANT POINTS AT A GLANCE
|
IN A NUTSHELL
|
What Is Quality Control? Definition and Scope
The terms “quality control,” “quality assurance,” and “quality management” are often used interchangeably in everyday language. In practice, this poses a problem because they represent three distinct tasks that require different tools.
Quality control is the act of verification: a product, a component, or a process result is checked against defined specifications, such as dimensions, torques, surface finishes, or functional values. The result is a decision: acceptable or unacceptable, approved or rejected. Quality control therefore always begins after a result has been produced.
Quality assurance, on the other hand, encompasses all preventive measures that ensure errors do not occur in the first place: capable processes, inspected tools, qualified employees, and robust work instructions. Finally, quality management is the overarching system that organizes both, with objectives, responsibilities, and continuous improvement in accordance with ISO 9001:2015.
| Criterion | Quality Control | Quality Assurance | Quality Management |
|---|---|---|---|
| Time | After a result is produced | Before and during the process | Continuously, system-wide |
| Question | Is the result acceptable? | Can the process produce errors at all? | Is the organization achieving its quality goals? |
| Typical tools | Measuring equipment, test stations, sampling, process data analysis | FMEA, process capability, operator guidance, calibration of test equipment | QM Manual, audits, CIP, management review |
| Result | Approval or rejection, test records | Capable, controlled processes | Certifiable system according to ISO 9001 / IATF 16949 |
| Perspective | Retrospective, descriptive | Forward-looking, preventive | Controlling, improving |
Important to note: Quality control is not the outdated concept it is sometimes portrayed as. It is the last line of defense before the customer and is simply mandatory in regulated industries such as automotive, medical technology, and aviation. Only one specific form is outdated: the purely downstream inspection of the finished product, whose results are documented on paper and never systematically evaluated.
In an engine plant with 40,000 bolted joints per day, the question is not whether inspections are performed, but where and how quickly: at the workstation in real time or days later in the inspection report.
Quality Control and Traditional Quality Assurance: Why This Distinction Is Crucial for Defect Costs
This distinction is not merely an academic exercise. It determines at which point in the value chain a defect is detected—and thus directly affects its cost.
The “rule of ten” for defect costs describes this relationship as a rule of thumb: With each stage of the value chain that a defect passes through undetected, the resulting costs increase tenfold. A component that’s screwed in incorrectly but never leaves the workstation costs a few seconds of rework. The same component at the end of the line requires disassembly, inspection, and reassembly. If discovered at the customer’s site, it results in a complaint, a sorting operation, and—in the worst case—a recall.
DEFECT COST STRUCTURE: WHERE A DEFECT IS DETECTED DETERMINES ITS COST
|
This leads to the core strategic message of this article: The goal is not more control, but earlier control. Every euro that moves defect detection closer to the point of defect occurrence saves many times that amount in consequential costs. Traditional quality assurance reduces the occurrence of defects, while quality control shortens the time to detection. Both levers act on the same cost curve.
A practical example from the automotive industry: For safety-critical features, suppliers work with target values in the single-digit ppm range—that is, fewer than ten defective parts per million. Such values cannot be mathematically verified using downstream sampling inspections. They require a control mechanism that evaluates every single process step as it occurs.
The Bottleneck of Post-Process Inspection: A Comparison of Methods and Standard Requirements
There is no single “right” method of quality control. There are characteristics with varying levels of risk, and for each characteristic, there is an economically sound inspection strategy. An overview of the four basic methods:
| Method | How It Works | Suitable for | Limits |
|---|---|---|---|
| Sampling inspection (e.g., according to AQL) | Inspection of a statistically defined subset, drawing conclusions about the entire batch | Non-critical characteristics, incoming goods, large lot sizes | Residual risk always remains; unsuitable for ppm targets and safety-critical characteristics |
| 100% inspection (inspection of all parts) | Every part is inspected, either manually or automatically | Safety-critical characteristics, customer requirements, sorting operations | High inspection costs; manual visual inspection typically achieves only an 80 to 90% detection rate |
| Statistical Process Control (SPC) | Monitoring of process parameters (cp, cpk) using control charts; intervention when trends are detected | Mass production processes with measurable continuous characteristics | Requires stable processes and consistent data collection; responds to trends, not to individual defects |
| Process-integrated inspection (inline inspection) | Process data such as torque curves or press-fit values are evaluated against tolerances in real time—for every part, at every step | Joining processes (screwing, press-fitting, welding, bonding), comprehensive documentation requirements | Requires networked equipment and a system that collects and evaluates the data |
In practice, these methods are combined. A typical scenario on the assembly line: process-integrated inspection for all joining processes subject to documentation requirements, SPC for dimensionally accurate machining processes, spot checks for packaging and labeling, and 100% inspection only where required by the customer or a standard.
Purely downstream inspection becomes a bottleneck for two reasons: statistically, because ppm targets cannot be verified using sampling; and regulatory, because standards require significantly more than just a filed inspection report:
| Standard / Regulatory Framework | Requirement | Implications for Quality Control |
|---|---|---|
| ISO 9001:2015 Section 9.1 | Monitoring, measurement, analysis, and evaluation; decisions based on data | Test results must be systematically evaluated, not merely filed |
| ISO 9001:2015, Section 6.1 | Risk-Based Thinking | Determine the scope and method of testing for each characteristic based on risk, e.g., using FMEA |
| IATF 16949, Section 8.6 / 8.6.2 | Product approval, layout review, and functional testing | Every approval decision requires verifiable, traceable test records |
| IATF 16949 Section 8.5.2 | Identification and Traceability | Test results, parts, lots, tools, and dates must be linkable |
| IATF 16949 Section 8.5.1 | Controlled Production Conditions | Monitor process parameters; do not merely inspect final results |
| EU Product Liability Directive 2024 | Extended liability, reduced burden of proof for injured parties, expanded definition of “manufacturer” | Anyone who cannot provide complete testing and process documentation in the event of damage bears the liability risk |
| VDA Volume 5 | Suitability of testing processes (measurement uncertainty) | The test equipment and test processes themselves must also be demonstrably capable |
Two issues regularly become problematic during audits.
-
First, retention: Test records for characteristics subject to documentation requirements must remain available for 15 years or longer, depending on customer specifications and the product, even if the manufacturing system has long since been shut down.
-
Second, traceability: A test report without a clear link to the part, batch, and process parameters is virtually worthless in the event of a complaint because the source of the defect cannot be pinpointed.
How process-integrated quality control works with process data analysis
In most plants, the maturity level of quality control can be gauged by a single question: How long does it take for a deviation in the process to trigger a response? In practice, the answer ranges from seconds to weeks.
Level one is paper-based control: workers’ self-inspections using clipboards, inspection records in binders, and evaluations conducted at most once a month. Level two is digitally recorded but downstream control: inspection values are entered into Excel or standalone systems, and evaluations are performed manually. Stage three is process-integrated monitoring: equipment and testing instruments automatically send their data to a central system, which immediately evaluates and documents every value and triggers an alert in case of a deviation.
Process-integrated inspection has a structural advantage: It does not generate additional inspection costs per part because it uses data that the equipment produces anyway. A screwdriving system measures torque and angle for every screw connection. The only question is whether these values are evaluated, documented, and immediately reported in the event of an error—or whether they remain unused and are lost in the control system’s memory.
The difference is measurable. In pilot projects using process-integrated test data acquisition, the rework rate typically drops by 20 to 40% because production defects are stopped after the first part rather than after the hundredth. At the same time, the manual documentation effort—which can easily take 2 to 5 minutes per inspection in paper-based systems—is eliminated.
PRACTICAL TIP: IPM—PROCESS DATA AS INSPECTION RESULTSIPM, the process data management system within Manufacturing OS, captures all data from the production process: torques, press-fit values, welding and bonding parameters, and inspection results. Each value is evaluated in real time against the stored tolerances. In the event of deviations, the system alerts operators immediately, allowing corrective action to be taken before additional defective parts are produced. All results are recorded in a product history file for each product, including automatic archiving. This transforms quality control into audit-proof quality documentation. Learn more at csp-ipm.com. |
The human factor is also part of digital quality control. A significant portion of the errors detected later during inspection result from assembly errors: the wrong part, the wrong sequence, or a missed step. The PG worker guidance system intervenes before the inspection, guiding employees step by step through the assembly process and confirming each step in the system. And one often-overlooked component is the inspection of the equipment itself: Tool inspection with QST ensures that testing and assembly tools are verifiably capable, as required by VDA Volume 5 and DIN EN ISO 6789 for torque tools.
5 Steps to Process-Integrated Quality Control
The transition from downstream to process-integrated quality control is not a software project, but rather a change in methodology. The following approach has proven effective; depending on the scope of the production line, it typically takes 3 to 9 months to reach the first operational area.
Step 1: Identify Critical Characteristics and Risks
The starting point is the FMEA or the characteristic classification: Which characteristics are safety-related or subject to documentation requirements, which are functionally critical, and which are non-critical? Only with this classification can the testing effort be allocated on a risk-based basis, as required by ISO 9001:2015, Section 6.1. A rule of thumb from practical experience: 10 to 20% of the characteristics account for over 80% of the risk of defect costs.
Step 2: Define a testing strategy for each characteristic
The method is specified for each characteristic: process-integrated inspection for joining processes subject to documentation requirements, SPC for continuous machining characteristics, and sampling for non-critical characteristics. The result is an inspection plan that specifies the method, inspection equipment, tolerances, and action plan for each characteristic.
Step 3: Automatically collect process data
The equipment provides the data, and a central system collects it: screwdriving systems, presses, welding systems, and dosing systems are integrated so that every process value is recorded automatically, completely, and tamper-proof. Data discontinuities, such as manually typing in display values, are eliminated. This is the point that determines whether subsequent inspections are based on reliable data.
Step 4: Set up real-time evaluation and response logic
Every recorded value is immediately evaluated against tolerances. An escalation process is defined for deviations: Who is notified and how quickly? When does the line stop? When is rework performed? When is production halted? The goal is a response time measured in minutes rather than days. Important: The decision to approve or block remains a human decision; the system provides the basis for it.
Step 5: Archive and systematically evaluate records
All inspection and process data is fed into a product history file for each product and archived in an audit-proof manner for the entire required retention period. Within Manufacturing OS, CHRONOS handles this long-term archiving, regardless of the lifecycle of the source systems. On this basis, trends, clusters, and process capabilities are regularly analyzed so that monitoring leads to improvement.
WHEN PROCESS-INTEGRATED QUALITY CONTROL WORKS
|
Case Study: Quality Control Using Process Data at Mercedes-Benz
The Mercedes-Benz plant in Hamburg demonstrates what process-integrated quality control looks like in series production. There, IPM provides process monitoring with automated notifications when thresholds are exceeded and daily updated web reports. The software is strategically integrated into the IT strategy and forms part of the internal product lifecycle record.
“IPM enables preventive quality assurance. The tool provides all the necessary basic data from the process for quality planning. The software has therefore been strategically integrated into Mercedes-Benz’s IT strategy.”
— Stephan Ivanauskas, Quality Systems Administration, Mercedes-Benz Hamburg Plant
The key point of this example: Quality control here not only provides approval for each part but also the basic data for quality planning. The same data collection enables real-time monitoring, documentation, and continuous improvement. Customers such as BMW and Knorr-Bremse also use a combination of worker guidance and process data monitoring to safeguard manual assembly processes in line with their zero-defect strategy.
→ Mercedes-Benz Success Story: Read in detail how the Hamburg plant set up process monitoring with IPM, from connecting the equipment to generating daily reports. Request your free copy now
AI in Quality Control: Possibilities and Regulatory Limits
AI-supported methods measurably enhance quality control. Anomaly detection in process curves—for example, using Curve Anomaly AI—identifies irregularities in screw-tightening or press-fit curves that are technically within tolerance but would be missed by rule-based inspections. This makes gradual process changes visible before they lead to scrap.
However, the boundaries are clearly defined, and honesty is essential here. AI is a decision-support tool, not a substitute for human responsibility. The EU AI Act requires transparency and human oversight for such systems, and fully autonomous approval decisions made by AI are not permitted by regulation in safety-critical industries such as the automotive and medical device sectors. IATF 16949 and the EU Product Liability Directive 2024 also require traceable decisions for which a human is accountable.
And there is another limitation that applies regardless of regulatory requirements: Without structured process data, it is impossible to implement AI-supported quality control. The data foundation always comes first; AI comes second. The article on AI in manufacturing quality assurance discusses in detail how to successfully get started with AI-supported quality assurance.
Frequently Asked Questions About Quality Control
What is quality control?
Quality control refers to the inspection of products, components, or process results against defined specifications such as dimensions, torques, or functional values. The result is a decision to approve or reject the product, accompanied by documented test records. Quality control always begins after a result has been produced and is therefore the verifying—not the preventive—component of quality management. In regulated industries such as the automotive and medical technology sectors, it is mandatory for critical characteristics.
What is the difference between quality control and quality assurance?
Quality control checks whether a result that has already been produced is acceptable. Quality assurance encompasses all preventive measures to ensure that defects do not occur in the first place, such as capable processes, inspected tools, and worker guidance. Put simply: control finds defects; assurance prevents them. Economically successful manufacturing combines both and shifts the focus toward prevention, because a defect that is prevented is always less costly than one that is found.
What quality control methods are used in manufacturing?
The four basic methods are sampling inspection (inspection of a statistical subset, such as according to AQL), 100% inspection of all parts, statistical process control (SPC) using control charts and process parameters, and process-integrated inspection, in which process data such as torque or press-fit curves are evaluated against tolerances in real time. In practice, these methods are combined depending on the characteristic: process-integrated inspection for joining processes requiring documentation, SPC for continuous characteristics, and sampling for non-critical characteristics.
When is a 100% inspection appropriate, and when is a sample inspection sufficient?
A full inspection is advisable for safety-critical or documentation-required characteristics, for customer requirements, and for sorting operations following a defect. Sampling is sufficient for non-critical characteristics with stable processes and large lot sizes. It is important to recognize the limitations of statistics: defect rates in the single-digit ppm range, as required in the automotive industry, cannot be verified using sampling. This requires an inspection that evaluates every single process step, typically integrated into the process via equipment data.
What is the cost of missing or delayed quality control?
As a rule of thumb, the “rule of ten” for defect costs applies: With each stage of the value chain that a defect passes through undetected, the resulting costs increase tenfold. A defect caught on the production line costs seconds of rework; the same defect at the customer’s site requires sorting operations and special shipments; in the field, it risks a recall and liability. Quality-related costs in manufacturing typically amount to 5 to 15% of revenue, with the majority attributable to the costs resulting from defects rather than the inspection itself.
Which standards govern quality control?
ISO 9001:2015 requires, in Section 9.1, systematic monitoring, measurement, and evaluation, as well as data-driven decisions, and in Section 6.1, a risk-based approach when determining the scope of inspection. IATF 16949 specifies requirements for the automotive industry: Sections 8.6 and 8.6.2 require verifiable release decisions; Section 8.5.2 requires complete traceability; and Section 8.5.1 requires controlled production conditions. VDA Volume 5 is the authoritative standard for the suitability of the testing processes themselves. The EU Product Liability Directive 2024 further increases the pressure to be able to provide testing and process documentation over the long term.
How does digital quality control work?
Digital quality control replaces paper logs and manual data entry with a closed data chain: equipment and testing instruments automatically transmit their readings to a central system, which evaluates each value against tolerances in real time, triggers immediate alerts in case of deviations, and archives all results in an audit-proof history file for each product. In pilot projects, this typically reduces the rework rate by 20 to 40% because production defects are stopped after the first part rather than after the hundredth. Systems such as CSP’s IPM implement precisely this principle.
Can AI take over quality control?
No, not entirely. AI-supported methods, such as anomaly detection in process curves, enhance quality control by identifying anomalies that are technically within tolerance limits. However, fully autonomous approval or rejection decisions made by AI are not permitted by regulations in safety-critical industries: The EU AI Act requires transparency and human oversight, and IATF 16949 as well as the EU Product Liability Directive 2024 require traceable decisions for which a human is accountable. AI is therefore a decision-making aid, not a substitute for human responsibility.
