The standard activity sampling procedure is reviewed. It relies on the ‘binomial proportion confidence interval’ in order to assign statistical credibility to its findings. In particular, the logistics and operation management community typically uses the so-called ‘Wald Interval’ within the activity sampling procedure. Unfortunately, the Wald Interval is now well known to be a rather poor approximation. Herein we evaluate the performance of the Wald Interval for activity sampling and propose a set of performance measures for activity sampling confidence intervals. We then review a range of alternative confidence intervals before making a recommendation for a new interval, the Clopper-Pearson Interval. A new activity sampling procedure that exploits this research is presented.