Identifying and resolving a defect early in the software development lifecycle is undeniably going to improve the quality but it is even better when defects can be prevented in the first place. Preventing defects from occurring requires a structured disciplined methodology; including gathering and analysing data and conducting root cause analysis, determining and implementing the corrective actions and sharing the lessons learned between projects to avoid future defects.

With ValidatAI we have fully automated the defect analysis and prediction process, leveraging machine learning algorithms and AI techniques to predict defects in real-time and determine the ‘next best action’.

This post, the first of a series , describes how ValidatAI, leveraging advanced cognitive approaches, can uniquely transform you’re your testing and development processes towards a defect prevention experience.

Predicting defects with speed, accuracy and intelligence

Quite often, identified defects are assigned to the wrong person or team based on incorrect assumptions on their root cause. This is because the person that assigns the defect, use his own insights, experience and sometimes intuition. A wrong defect assignment can delay its fixing by hours or even days.

Consider the case of a tester working at a global bank, who is raising and assigning defects for Temenos AA products to developers, BAs or environment managers based on his personal interpretation and intuition. After spending 2 days to analyse the nature of the defect, the developer suggests that the issue is environment-related and reassigns the defect to the environment manager who then spends another couple of hours to fix the issue.

This is a very common situation and if the tester was able to know the exact root cause of the defect from the beginning, the issue would have been fixed within 2 hours rather than 2 or 3 days.

ValidatAI is able to automatically identify the root cause of the defects and provide the right justifications to ensure confidence in the recommendation.

Problem Definition and Context Understanding

The system aims to assist the human tester, based on the data that it extracts from the core banking system, providing Real-Time Recommendation on workflow paths (creation or extension) that achieve the highest Test Data Coverage, and identifying common failure patterns across data to prioritize paths that will detect bugs utilizing all available context.

The deep learning model is able to understand the hidden dependencies beyond pair-wise entities in a more abstract way. This provides an estimation for the time to fix and time to resolve for each issue. It executes data mining algorithms on the defect characteristics, mainly on real-time structured or unstructured data, such as the defect reports, which provides detailed characteristics about the nature of the bug.

With this approach we have the ability to create semantic enrichment on the actual wording with accuracy and completeness.

In the next post we will describe how defects are getting prioritized and analysed to automatically identify their root cause.