The success rate of a squeeze cementing operation is generally low due to complexities in the range of potential leakage pathways and enormous uncertainty in characterizing these pathways. This work aims first to develop a physically based numerical model of the key part of the squeeze cementing operation, i.e. invasion of cement slurry into a realistic microannuli geometry under operational conditions. Secondly, it aims to account for uncertainties by using statistical tools combined with the deterministic model to deliver probabilistic information regarding outcomes. This paper develops this novel risk-based approach. The first objective of this work is to build upon our previous model, which focused on a single-perforation injection scenario. We reconstruct a multi-perforation injection scenario by collocating radial (single-perforation) invasion flows. We use this model to investigate the invasion of a viscoplastic fluid, representing a cement slurry, into a randomized varying width microannulus channel. The second objective is to investigate the effect of relevant parameters on leakage reduction. This is broken into two parts. First, how effectively the squeeze operation fills the microannulus around a perforation, explored through metrics that quantify penetration/filling. Second, what effect the penetration/filling has on the leakage of the entire well, i.e. given that the squeeze operation is local. We compare the effect of different perforation patterns and rheological parameters on penetration/filling metrics and on the reduction of leakage. We generate a probability distribution of leakage rates before and after the operation. In this way can estimate both the mean reduction in leakage for different perforation patterns, and associated confidence intervals. Such predictions have not been made before. Our results demonstrate the inherent uncertainty of a squeeze operation, which arises from the geometrical complexity. Practically, this suggests that higher perforation density and lower yield stress slurry will lead to higher success rates, as is intuitive. However, the spread of uncertainty in outcomes is not reduced much by such practices, meaning that one can still be unlucky in perforating the wrong part of the annulus.