Models of discrete counts are popular in many application fields, ranging from the incidence data in epidemiology, the number of stock market transactions in finance, traffic intensities in networks and transportation, the number of particle arrivals in physics, to phenomena in social networks. High counts can generally be approximated by continuous data models, but those can fail if the counts are low and include many zeros. Inspired by the popular poissonian generalized linear models in frequency statistics, we propose a method for their sequential estimation that is rooted in the Bayesian paradigm. Its details will be presented.