Unreplicated fractional factorial experiments with response modeled with Generalized linear models (GLM) are found more and more frequently in industrial applications. GLM analysis relies heavily on large sample results. This paper presents a Bayesian method for detecting the active effects in unreplicated factorial experiments analyzed by a GLM that does not require the large sample assumption. The proposed method is based on Bayesian model selection. In the examples shown, the Bayesian method produces more consistent results thaninference based on Wald’s test, and in a simulated example the usual approach brakes down while the Bayesian method identifies the significant effects correctly. The method is presented for the 2 kexperiments, but it can easily be generalized to other designs.