The method came first.

The professional starting point

I am Zoltán Dózsa, a mechanical engineer. I began my career as a lead design engineer in the aluminium industry and, after a few turns along the way, found myself in the automotive industry — a high-pressure, demanding and fascinating world. I became deeply familiar with the world of systems: serious and powerful, but sometimes also overly bureaucratic and more focused on compliance than on real value.

I spent the final four years before retirement at Nokia’s manufacturing site in Komárom, Hungary, working as a process improvement expert. Nokia provided an exceptional human environment, with outstanding colleagues who were genuinely open to innovation.

At the time, the factory was facing major financial losses caused by phones being reworked because of visual defects, together with a large amount of capital tied up in reworked inventory. My colleague Javier — in many ways the conscience of the factory — pointed out that visual inspections were highly unreliable, tended to overreact to defects and were generating substantial losses. He asked me to assess the capability of the inspectors.

Coming from the automotive industry, I confidently said yes. After all, I knew the AIAG MSA (Measurement System Analysis) methodology. What I had not yet considered was that the world appears simple in only two cases: when we are geniuses, or when we know almost nothing about the problem. I am not a genius, so at that point I clearly belonged to the second group.

I intended to start with Attribute Gage R&R, but I ran into fundamental questions almost immediately: How large should the assessment sample be? What should its composition be? How easy or difficult should the assessment be? How should reliable reference decisions be established?

Neither the American AIAG nor the German VDA methodology provided sufficiently practical answers to these questions. Then came the evaluation itself. Statistical software produced plenty of results, but there was no single qualification value that could say, in an understandable way, that an inspector’s capability is X% — and whether that level is good enough or not.

The established approaches left too many important questions unanswered and were not practical enough for the problem we were trying to solve. That is how the development of the VQI methodology began. It evolved considerably through experience gained in real applications and eventually became a coherent system that can be integrated into an organisation’s quality-management framework.

The core idea

Inspection capability is typically assessed by asking inspectors to classify every item in a multi-piece sample several times. Statistical methods can then be used to measure consistency (repeatability), the correctness of decisions, and the frequency of the two fundamental types of decision error: rejecting an item that is actually good, or accepting an item that is actually defective.

However, the result of such an assessment can be strongly influenced by a number of factors and conditions. These factors must be controlled and, as far as possible, kept constant from one assessment to another if results are to remain comparable. I gradually realised that this was precisely where the established methods offered too little practical guidance.

The main factors that can significantly influence the result include:

• a clear definition of what constitutes a capable or “good” inspector;
• the quality of the specification, including defect descriptions, boundary samples, clarity and completeness;
• the stability, size, composition and difficulty of the assessment sample;
• the number of repetitions;
• the reliability of the reference decisions; and
• the method used to evaluate the assessment.

There should be one — and only one — principal metric for inspector capability, because that is what makes the result understandable to everyone involved. In VQI, this is Dozsa’s WSR% (Weighted Summary Result), a weighted combination of two relevant factors: repeatability and decision correctness. The result is then assigned to a qualification level, allowing the organisation to determine whether a person qualifies as an inspector and, at the highest level, as a VQI Master.

Professional development of the method

An especially important issue is the reliability of reference decisions. Within VQI, these are established through the unanimous decision of at least three highly capable inspectors — VQI Masters — ideally with the involvement of a customer or OEM representative where appropriate.

Assessments using the VQI method may only be conducted after appropriate training. Training is delivered by VQI Masters, who in turn participate annually in a VQI Workshop. The workshop serves both as continuing professional development and as a forum for jointly examining difficult cases, defects and lessons learned from practical application.

The statistical evaluation of assessment results is another essential part of the methodology. A key distinction is whether sample-based or decision-based statistics are used. The VQI method can also account for asymmetry in the assessment sample, which may otherwise significantly distort results. It can even incorporate specific management priorities by giving additional emphasis to particular types of inspector error in the final evaluation.

The methodology is not limited to human visual inspection. It can be applied more broadly to classification-based or qualifying inspections, including both binary classifications and ordinal, ranked classifications. Visual inspection remains its most common field of application, but the same decision-based framework can also be used to evaluate camera-based inspection systems and compare their capability with human inspection.

Selected publications: Magyar Minőség, October 2021, Vol. XXX, No. 10; and Minőség és Megbízhatóság (EOQ-MNB), 2025/4.