Construction projects, even exemplarily planned and organized, bear a risk of unforeseen events and problems which can result in completion of the works after the deadline, that is delays. The construction of bridges is an inseparable part of road and rail projects and construction and expansion of the transport network. The paper aims at finding a relationship between the independent variables characterizing bridge projects and the delays during their implementation. Two alternative models were proposed to solve the problem: logit and probit. The data set comprising road and rail bridges built in Poland in the last 12 years (2005–2017) was used to build the models. The evaluation, quality and accuracy parameters of proposed models were determined in the final part of the paper.

The article presents briefly several methods of working time estimation. However, three methods of task duration assessment have been selected to investigate working time in a real construction project using the data collected from observing workers laying terrazzo flooring in staircases. The first estimation has been done by calculating a normal and a triangular function. The next method, which is the focus of greatest attention here, is PERT. The article presents a way to standardize the results and the procedure algorithm allowing determination of the characteristic values for the method. Times to perform every singular component sub-task as well as the whole task have been defined for the collected data with the reliability level of 85%. The completion time of the same works has also been calculated with the use of the KNR. The obtained result is much higher than the actual time needed for execution of the task calculated with the use of the previous method. The authors argue that PERT is the best method of all three, because it takes into account the randomness of the entire task duration and it can be based on the actual execution time known from research.