AbstractTwo separate reliability models are proposed which one may wish to use to assess product performance. By combining stresses associated with traditional high temperature tests, temperature cycling tests and power cycling tests, we believe more representative and comprehensive reliability tests can be created. Although this understandably makes the models employed slightly more complex, the introduction of a step‐stress methodology compensates for this, thereby eliminating the requirement for additional test cells to determine the particular model's parameters. Estimates for the model parameters, or their associated acceleration factors relative to the normal use environment, will enable us to make accurate reliability prediction