Numerous methods exist to determine the optimal number of classes when using latent profile analysis (LPA), but none are consistently correct. Recently, the likelihood incremental percentage per parameter (LI3P) was proposed as a model effect-size measure. To evaluate the LI3P more thoroughly, we simulated 50,000 datasets, manipulating factors including sample size, class distance, number of indicators, and number of classes. Results indicate the LI3P performs similarly to established techniques, reflecting class separation and the number of classes. The novel simulation method also yields new findings for the standard approaches. We then demonstrate the LI3P by applying it to a study searching for subgroups of formerly-religious individuals.