Model-Based Testing - is Now (Finally) a Good Time?

Model-Based Testing (MBT) is a powerful quality assurance methodology. When using MBT, QA teams create a computer-actionable model of intended system behavior. Then, they use automatic tools to derive optimized test suites and scenarios from that model. These scenarios can be automated, manual, or scripts to be used by other test platforms (from BDD to Python scripts). Additionally, the models can be visualized, validated with non-technical stakeholders, and verified for compliance with company policies and regulations. The latter allows finding bugs in requirements prior to system implementation.

The automation level enabled by MBT can revolutionize QA and its role in the SDLC. For example, being able to clarify, validate, and verify requirements early in the development process means that QA can work with the product team before passing requirements to developers. The radical reduction in time required for test script maintenance and creation makes QA teams very agile, and makes test pyramids outdated.

First suggested in the late 1990s, MBT has yet to gain wide adoption. However, with new modeling paradigms and GenAI, this might finally change. This talk will introduce a modern approach to MBT, and discuss our experience applying it in the software industry.

Paper | Presentation

Michael Bar-Sinai

Michael is a software engineer by training, with a mid-career PhD in formal methods and requirement modeling. He has created software of various qualities since about 1998, including production systems, online information systems, code libraries, and open-source tools. Former data science toolsmith at Harvard's IQSS. CTO and Co-Founder at Provengo, a start-up creating model-driven software engineering tools. Married, 3 kids, sadly no dogs.