Installation, operations and maintenance costs represent one of the major barriers to the growth of wave and tidal-stream energy projects and a cost that can be significantly reduced in offshore wind. Time-domain Monte Carlo simulations used for developing and optimising offshore operation strategies can help minimise these costs. To achieve realistic and comprehensive estimates of project duration using these tools, it is essential to represent accurately the durations and variability of the input operation data. This paper attempts to quantify this stochastic nature of operation durations through the analysis of recorded data for an offshore wind farm installation project. Here we present evidence of learning for the majority of analysed operations and describe a stochastic learning curve model that is well-suited for implementation within time-domain Monte Carlo simulations. Additionally, the lognormal probability distribution is shown to be the most applicable for representing the random nature of operation durations for the recorded data-set. The proposed methods can be used during the operation phase of a project, where iterative analysis of recorded operational data is continually performed to improve predictions of project duration. Previously determined learning and distribution characteristics can also be used to obtain estimates for comparable future projects.