This paper deals with preliminary test estimation for mean of an inverse Gaussian population. Preliminary test estimator has been shown to provide large gains in efficiency, especially around a neighbourhood of the prior guessed value of the parameter, for many distributions including exponential and normal, however, this has not been explored for the inverse Gaussian family of distributions. Owing to diverse applications of the inverse Gaussian model for non-negative and positively skewed data, the investigation considered here makes an important contribution in the area of preliminary test estimation. We consider both the cases of known and unknown dispersion parameters and demonstrate similar conclusions as obtained in the case of Gaussian populations in terms of the efficiency of the resulting estimator.