This document summarizes the process of operationalizing prediction using partial least squares path modeling (PLS-PM). It outlines how PLS-PM is used to (1) estimate the measurement and structural models on training data, (2) form exogenous latent variable scores to predict endogenous variables, (3) estimate outcome variables, and (4) evaluate predictions on holdout data using various metrics like RMSE and cross-validation. Open issues like incorporating mediation and alternative PLS estimators are also discussed. Implementations of these prediction procedures in R packages are noted.