Mean daily traffic (MDT) is the expected traffic volume at some site on a typical day, and it is usually estimated from short-count data by computing average daily traffic (ADT) and then correcting this ADT for the season or day of week of the count. Although considerable guidance exists on how to construct seasonal factor groups from automatic traffic recorder counts, less guidance is available on how to select the appropriate factors for correcting a particular short count or for estimating MDT when the appropriate factors are uncertain. A data-driven approach to assigning highway sites to factor groups using arbitrary samples of daily traffic counts, a method for designing traffic count samples to minimize the likelihood of assigning the site to the wrong factor group, and a Bayes estimator of MDT are described. A likelihood function describing the sample count is combined with prior estimates of the probabilities that a site belongs to each factor group to produce posterior classification probabilities. The site is then assigned to that factor group showing the highest posterior classification probability. The classification method is evaluated by using actual traffic data and appears to perform reliably with 14-day samples. A Bayes estimator of MDT is then developed, which is applicable even when it is unclear to which factor group a short-count site ought to be assigned. This estimator is evaluated by using actual data, and it also performs creditably with 14-day samples.