Understanding Software Quality: From Usability Layers to Testing Practices and Cost Management

In the context of cost reduction and the rise of large‑model technologies like ChatGPT, software quality and testing face significant challenges, prompting a need for clearer direction and better decision‑making.

The article defines quality as the degree to which an object's inherent characteristics meet requirements and presents a three‑layer understanding: usable (ensuring service stability and functionality), useful (enhancing user experience and business value), and lovable (driving user advocacy).

A quality loss model is introduced, quantifying loss as a function of change count, problem density, development and test leak rates, and handling level, highlighting the importance of controlling business loss.

Quality cost is discussed, encompassing resources spent on design, testing, issue resolution, and robustness, with emphasis on avoiding ineffective or low‑efficiency investments.

The testing section defines testing, distinguishes verification (VE) and validation (VA), and outlines the relationship between quality and testing, emphasizing that testing provides feedback rather than directly improving quality.

Testing is broken down into five stages: input, execution, analysis, localization, and evaluation, each described with concrete activities and challenges, especially for functions, modules, subsystems, and client apps.

Various classifications of testing are presented, including by problem type, object level, technique, and responsibility (new‑feature vs. regression testing), stressing the importance of regression testing for maintaining stability.

Cost of defect recall is analyzed, showing how recall level (white‑box, module, system) impacts resource consumption, and proposing methods to reduce cost by eliminating ineffective tests and adjusting defect distribution.

Risk‑based testing technology is introduced as a way to prioritize testing based on potential impact, improving ROI and reducing waste.

The article also explores the potential of large‑model AI (TestGPT) to automate test case generation, decision‑making, and execution, suggesting a future where QA relies on creativity rather than experience.

Finally, the article calls for stronger research in quality and testing technologies, better understanding of test objects, and continuous investment in automation and AI‑driven testing to achieve high‑ROI regression testing and proactive defect prevention.