A comprehensive joint orientation data collection program is necessary to ensure a sufficient degree of confidence in constructing a structural model. This paper investigates the influence of the number of joints sampled on the characterisation of the joint dip and dip direction. In a series of numerical experiments, Discrete Fracture Networks (DFNs) were used to model joint set populations. The average joint properties and the variation of the orientation cluster are obtained by simulating vertical and inclined boreholes in three different DFN models. The confidence level in the joint orientation data is calculated with the confidence limit method for different drilling densities. For comparable joint orientation data variability, a similar trend is observed regarding the confidence level and the number of joint sampled. Based on that trend, a series of recommendations were developed to estimate the number of joint to sample for a single joint set in order to reach a targeted level of confidence in the dip and dip direction data. Depending on the project requirements, the use of a range of levels of confidence and degrees of precision in the recommendations, can provide greater flexibility in design decisions. The proposed recommendations can be used to optimize the planning of geotechnical drilling campaigns for new mining projects or to review an existing structural database for an undergoing project identifying gaps in the data.