One of the key reliability limitations of low cost encapsulated optoelectronic devices has been temperature cycling. This is because the clear polymers used to encapsulate the devices have expansion coefficients significantly greater than the devices to be encapsulated. This mismatch causes thermal stresses and cyclical stress causes low cycle fatigue failure.1 In this paper we show how the use of log-normal plots of the failure rate allows ready comparison of different device types and structures, extrapolation to different stress conditions and the examination of the statistical significance of the results. Two types of stress condition are distinguished, polymer motion and polymer force. In the polymer motion mode, the polymer portion is large and stiff compared to the rest of the device. In the polymer force mode, the polymer is small and/or soft compared to the device. In terms of device design, the two things that affect the mode and the temperature cycle reliability are structure and materials. In the structures area, a series of comparisons between elastic stress theory and observed failure rates is done for different device geometries. Stress concentrations around ceramic substrates, bending forces on ceramic substrates, stress concentrations around stiff lead frames, stress in wire loops and kink formation in gold wire are all examined. An integrated model for certain failure modes is proposed. In the materials area, the benefits of soft silicones and so-called magic epoxes are presented.