Quality Function Deployment (QFD) aligns product features with customer needs through cause-effect matrices, facing challenges in scalability and functional model appeal but showing promise in software testing and automotive design. The podcast contrasts QFDs underutilization with AIs overreliance, while exploring Swiss efforts to integrate causal reasoning with LLMs for transparency in high-stakes applications.

Richard Seidl Software Testing

This is the other podcast on Software Testing by Richard Seidl, the episodes are in spoken German but the show notes and site are written in English. Our summaries are generated from AI transcript translations.

• https://www.richard-seidl.com/en/blog/tag/podcast-software-testing
• https://www.richard-seidl.com/en/

Episode Details

• Show Notes: https://www.richard-seidl.com/de/blog/quality-function-deployment-llm
• Published: 2026-04-28T04:00:00Z
• Duration: 00:25:03
• Author: Richard Seidl - Experte fur Software-Entwicklung und Testautomatisierung

Overview

The podcast discusses Quality Function Deployment (QFD) as a methodology for aligning product functionalities with customer needs through cause-effect analysis and matrices, illustrated by examples like designing a coffee machine to meet brewing preferences. It highlights challenges in applying QFD, such as the unappealing nature of functional models tied to managerial metrics and scalability issues with traditional matrices, though advances since 2014 allow handling larger datasets. QFD is also explored in software testing, where it links user interactions to customer-driven outcomes, prioritizing tests based on real-world usage and emphasizing security/privacy as baseline requirements. The podcast contrasts QFDs underutilization with the widespread but sometimes naive adoption of AI, suggesting potential synergies between the two approaches.

Limitations of large language models (LLMs) are examined, particularly their struggles with causal reasoning and explainability despite strengths in pattern recognition. Efforts in Switzerland, such as Cioto AI in Lausanne and research at ETH Zurich, focus on integrating causal models with LLMs to improve transparency and decision-making, with applications in critical fields like automotive manufacturing. The discussion also addresses the challenges of applying causal reasoning to AI systems to reduce hallucinations and certify AI for high-stakes industries. In automotive contexts, historical case studies (e.g., Skoda, Volkswagen) demonstrate QFDs role in optimizing design based on customer-centric metrics like space for handbags.

The podcast touches on broader themes, including the Swiss AI landscapes ambitions for causal reasoning leadership, challenges in retaining talent, and institutional struggles with funding and scalability. It explores how transfer functions, used in Six Sigma and fields like exoplanet detection, could aid in managing complexity and variability. Additionally, it envisions personalized testing aligned with individual user needs in Industry 4.0, moving away from mass production models. While customer feedback mechanisms like Net Promoter Scores (NPS) are emphasized for prioritizing testing and development, the discussion underscores persistent barriers to QFD adoption, including limited awareness, technical hurdles, and the treatment of QFD techniques as trade secrets.

What If

• What if you used QFD principles to restructure your software testing process with AI-driven cause-effect analysis?

  • Concrete move: Build a lightweight QFD matrix that maps customer feedback (e.g., from NPS) directly to test cases using an AI tool to prioritize high-impact scenarios.
  • Why now: LLMs excel at pattern recognition, but testing requires cause-effect logic (e.g., “If a user reports slow login, which modules need stress-testing?”). QFD bridges this gap, and recent AI advancements in Switzerland (e.g., Cioto AI) suggest tools for integration.
  • Expected upside: Reduce redundant testing by 30% while aligning with customer priorities, improving product quality and reducing time-to-market.

• What if you integrated a causal model into your LLM-based feature generation workflow to prevent hallucinations?

  • Concrete move: Prototype a system that pairs your LLM with a simplified causal graph (e.g., from transfer functions in Six Sigma) to validate feature suggestions against known dependencies (e.g., “A payment gateway change affects user data encryption”).
  • Why now: LLMs struggle with explainable AI, but causality checks are critical for software safety (e.g., automotive manufacturing where QFD is already used). Research hubs like ETH Zurich are actively exploring this synergy.
  • Expected upside: Create a more reliable feature roadmap, reducing rework costs and increasing stakeholder trust in AI-generated outputs.

• What if you developed a QFD-inspired tool to optimize your softwares security and privacy testing based on customer feedback?

  • Concrete move: Use customer surveys and NPS data to build a cause-effect matrix that identifies security risks (e.g., “Users report data leaks when using API X”) and automatically generates targeted test cases.
  • Why now: The text emphasizes security/privacy as non-negotiable in testing, yet many solo developers lack tools to prioritize this. QFDs structured approach, validated in automotive sectors (e.g., Volkswagen), offers a scalable framework.
  • Expected upside: Cut security-related bugs by 25% while ensuring compliance with evolving privacy regulations (e.g., GDPR), improving customer retention and reducing legal risks.

Takeaway

• Implement QFD matrices to prioritize testing by mapping customer feedback and cause-effect relationships directly to test cases, reducing redundant testing efforts and aligning software features with measurable user needs.

• Integrate causal reasoning tools with AI models to enhance transparency, such as using frameworks or research from Swiss institutions (e.g., ETH Zurich) to mitigate hallucinations and enable explainable AI for critical software features.

• Leverage Net Promoter Score (NPS) and feedback analytics to prioritize development and testing tasks, ensuring software updates directly address the most impactful customer pain points identified through real-time or periodic user surveys.

• Explore open-source causal AI projects or collaborate with Swiss research hubs (e.g., Cioto AI in Lausanne) to adopt nascent techniques for combining LLMs with causal models, improving decision-making in software design and testing workflows.

• Adopt scalable QFD tools or AI-driven methods to handle large datasets (e.g., 1000+ user stories) by utilizing post-2014 advancements, ensuring systematic alignment between customer needs and technical implementation without manual matrix limitations.