The need for technological forecasting as an input to life studies is discussed, with the telecommunications industry as a particular case study;Twenty-two historical multi-industry cases of technological growth and ten cases of the substitution of stored program control for electromechanical switching from ten telephone companies are analyzed with six growth models. The models are statistically compared for fitting and forecasting ability. It is shown that the relationship between the fitting ability of a model and its forecasting ability is weak so that fitting alone should not be used a priori to select a forecasting model;The Normal, Fisher-Pry and the Gompertz growth models are shown to forecast significantly better than the Weibull and the lognormal, especially at lower penetration levels. At higher penetration levels, all the models studied perform equally well. Nonlinear estimation is shown to give better forecasts than linear estimation especially at higher penetration levels;It is recommended that the Normal and the Gompertz be considered along with Fisher-Pry by those industries presently considering the implementation of substitution analysis in their life estimations. In the early stage of growth, linear estimation should suffice to give reasonable forecasts. In the latter stage, however, as more data become available, nonlinear estimation should be used;It is suggested that when substitution analysis is used in life estimation, the life cycle forecasts derived from the analysis be used as an additional constraint to future experience before service life indications are developed. An upper limit to the life forecast using a life cycle is derived. A routine for transposing the life cycle information into vintage analysis is proposed. More research is necessary especially in regard to the question of addition and retirement patterns which substitution and life cycle analyses do not specifically address.