Many composite failure criteria suffer from serious flaws related to prediction accuracy and the number of tests required to apply them, as found for example during the World Wide Failure Exercises. In 2001, Jon Gosse of Boeing Phantom Works developed a new criterion for matrix failure. Initially known as Strain Invariant Failure Theory (SIFT), it was soon extended through collaboration with John Hart-Smith to also include fiber failure. During this process, it was also renamed to Onset Theory. Onset Theory predicts failure using micromechanically obtained strain invariants in the constituents of a composite. As a physics-based criterion, it is claimed to be applicable to all types of loading, boundary conditions, geometries, and layups. However, it has yet to see widespread application. In this thesis, a consistent approach to Onset Theory has been developed. This addresses and resolves the contradictions found in literature regarding the exact details of the application of Onset Theory. In addition to that, failure envelopes are generated for comparison with laminate test data. Previously, finite element models were required for each specimen and state of strain. In this thesis, the entire process has been automated for ease and speed of applicability. Using the consistent, automated approach, Onset Theory is shown to quite accurately predict failure in composite laminates under biaxial strains. This includes non-catastrophic matrix cracking preceding ultimate failure, without the necessity of performing a progressive failure analysis. Onset Theory also correctly captures trends in the difference between lamina and laminate failure. Most importantly, it only requires a single set of critical strain invariants (three in total), regardless of loading condition or stacking sequence. This means that extensive and expensive testing can be reduced significantly.