Lean Blog Series - Part 3: Lean in Quality Assurance

No-one knows the cost of defective product. Don’t tell me you do. You know the cost of replacing it, but not the cost of dissatisfied customer.” -W.E. Deming


Lean and Quality

Deming is a key influence behind Lean principles. His statement about the importance of customer satisfaction is crucial: by doing things right the first time, a company can optimize its operations to harness the power of recommendations to its advantage. While exploring willingness to recommend (NPS) and customer satisfaction is a common topic on Zeffi's blog, this time we will focus on quality as a production factor and the drivers that influence it.

As mentioned in previous parts of the Lean blog series, Lean management aims to meet customer expectations while keeping costs in check, with waste avoidance being a key strategy. It’s often oversimplified to say that Lean aims for "high quality," but this misses the mark—producing a quality higher than customer expectations would be value-destroying, i.e., waste.

Ensuring Quality with Quality Function Deployment


Quality Function Deployment (QFD) is a method that helps Lean practitioners focus on what the customer expects. The essential question is, what conditions must be met to satisfy the typical customer's actual needs? This shifts us from the beloved realm of canvassing customer needs closer to the calculable world, where we objectively determine which factors are significant value sources for the customer, contributing to the perceived quality of a product or service.

QFD is also known as the "House of Quality." There is no standard definition for the House of Quality, so it can be applied organizationally. This blog discusses the general components of QFD and implements a digital tool for quality assurance.

Generally, QFD consists of the following elements:

  • Customer insights (Voice of the Customer) on the importance of various product features (weight, e.g., 1-5)
  • Controllable drivers (measurable factors within your company's control)
  • The relationship between the above (weak, medium, strong; scale e.g., 0-10)
  • Competitors' estimated performance regarding product features
  • Relationships between drivers (very negative, negative, positive, very positive)

Practical Exercise - Truly Listening to the Customer with QFD

In this blog's practical section, we will determine the weights of the Voice of Customer (VOC) quality factors ("what the customer values") and identify the factors with the most significant impact using an alternative question ("what can influence this"). This is done using a unidimensional graphical question (scale) to determine driver weights and deepening the analysis by asking about the relationship between drivers and quality factors. This way, we can objectively determine customer views on the core of QFD.

Practically, collecting VOC weights can be done for large customer groups by asking about the importance of quality factors (e.g., ease of use) through an electronic survey. There are other ways to collect this data, but surveys are an efficient way to gather insights from large groups. Listening to large groups is crucial to avoid overemphasizing the views of the most vocal customers. However, there are reasons to do this, such as if the customers represent significant business value. Collecting the data this way allows us to score the weights, say on a 1-5 scale, reflecting customers' true opinions.

Figure 1: Voice of Customer can be collected en masse via electronic surveys.

Voice of Customer survey

Another practical QFD tool implemented in this blog is gathering the views of relevant, usually internal, stakeholders on the impact of drivers on quality factors essential from the customer's perspective. This differs from the previous example by using image questions to select the most suitable controllable drivers, though this could also be done with scale questions, which academic research suggests are more user-friendly. A scale-based implementation would increase the number of questions since each quality factor would need to be assessed for each controllable driver, resulting in, for example, 49 (7*7) items to evaluate—a feasible but laborious method even with an electronic tool.

The initial thought might be to ask customers about this part, too, but this analysis likely includes factors you cannot or do not want to disclose to a broader audience. When asking internal stakeholders (or other productive stakeholders in a subcontractor model), you can more openly discuss controllable value drivers. Figure 2 below shows a practical example of assessing individual factors by asking which drivers most affect each quality factor. In this example, the number of drivers is limited to three, as the exercise should focus on the essentials, allowing respondents to choose less than half of the available options. You can test the responses here.

Figure 2: In identifying QFD value drivers, the evaluation team gives their opinions on the most significant drivers affecting quality factors, one by one.

QFD value driver survey

Figure 3 below presents data on factors affecting the ease of use of a service. The next step in the assessment would be for the QFD implementation team to discuss which factors are ultimately chosen and scored in the QFD diagram's center as coefficients of the relationships between drivers and quality factors. In this example, no single factor emerged as the most significant (weight 10); on the other hand, the group noted the minor importance of training, possibly reasoning that if training is required, the product is not user-friendly enough.

Figure 3: Identifying drivers from collected data.

Summarizing quality factors

As noted, the QFD implementation method is organizational, but there are at least two good tactics here. 1) Score the top 5 factors with a weight of 5, average effect, and leave price points and training with minimal significance without coefficients. Alternatively, 2) score the top 3 factors with a weight of 5. You can test the scoring here.

Next-Generation Diagnostic QFD with AI Assistance

The above process provides a more objective view of the quality delivered to customers than the traditional productization method, where you take the expert role and decide the significance of quality factors on behalf of the customers. However, we still did not completely free the customer to express the most significant factors using traditional Lean methodologies. Where did we go wrong? The answer: we set the factors to be assessed, i.e., quality factors, on behalf of the customer.

Can this be done differently? Yes, it can. Let's do another exercise by first asking customers about their quality perceptions and then letting generative AI find the root cause behind the assessment and summarize the root causes of all respondents. This way, the process becomes genuinely customer-driven in terms of the Voice of Customer section.

Figure 4 below shows an example of a generative AI's conversation with a customer about quality. The discussion is initially undirected, with the AI interview freely exploring the reason why the customer rated the company's product quality a certain way on a scale of 0-10. The interview deepens as a dialogue between the customer and AI, used as the basis for root cause analysis. Root cause analysis is a key Lean method—here, it is supported by AI-generated interview material.

Figure 4: A diagnostic interview supported by generative AI helps uncover unexpected factors in customer preferences.

AI interview to identify quality drivers

AI can help extract the right insights from interviews and complement traditional Lean methods. AI also supports objective classification, identifying the most significant factors customers express about quality. From the researcher's or developer's perspective, we do not dictate which factors are most significant to customers—they provide the input, and AI does the classification. Figure 5 below shows a sample of VOC themes derived from a hypothetical 20 customer responses for Company XYZ.

Figure 5: In an alternative approach, AI-summarized interviews provide VOC themes.

Thematic, diagnostic AI analysis of quality factors

In this example, a panel of 125 customers responded to a quality interview and responses were thematized and analyzed by AI. Viewed this way, the VOC process becomes even more customer-driven. AI performs classification and thematizing more reliably and objectively than its human counterpart. Of course, it is good to use common sense here, but for Lean management tools, it is worth considering whether generative AI could truly be an alternative or complement to traditional methods. Preliminary experiences are highly interesting and suggest that AI-assisted QFD works very well.

Conclusion - Producing Quality, but from Whose Perspective?

This blog post has focused on Lean and quality, particularly QFD. It is good to challenge oneself and the organization—when we talk about quality, from whose perspective is it defined? If not from the customer's perspective, then whose? How can we avoid optimizing product and service development from the viewpoint of small silent minorities?

As a company offering SaaS products, we are constantly faced with this last question, as are others in the industry. Development should follow the views of the significant majority, which can be (judiciously) complemented with customization options and add-ons. This perspective is not limited to the software industry but is universal in business—how to maximize customer satisfaction?

Lean's systematic lessons, like QFD presented here, help crystallize the essential factors—the majority's voice can be sought by crowdsourcing the formation of opinions from a sufficiently large sample of customers instead of optimizing based on received customer feedback alone. A different approach could lead to a strong bias in development and optimization from the margins.

It is also possible and recommended to consider different smaller customer segments (strategically) and study the impact of various background factors on choices. This is also easier with electronic tools and quality reporting.

Hopefully, this post helps you in your path toward even more meaningful quality assurance, which is, after all, at the heart of Lean philosophy.


Janne Vainikainen, Zeffi, COO, Lean Six Sigma Black Belt


ps. Would you like to get a notification when our next blog post is published? Subscribe>>

Blog Series Overview: