We introduce and study the production routing problem with time windows and customized commodities (PRPTWCC) that generalizes the basic production routing problem (PRP) to multiple commodities as follows. The single supplier produces multiple commodities utilizing a joint production capacity. Each commodity may be demanded by several customers, but each customer demands a single commodity only. We consider the PRPTWCC an archetype for a PRP with an interesting multi-commodity lot-sizing component. To solve the PRPTWCC exactly, we develop and compare several branch-price-and-cut (BPC) algorithms based on path-based formulations with an exponential number of routing variables. The formulations differ in how production decisions are represented: one models production decisions explicitly, whereas the other two feature variables associated with feasible production plans. BPCs based on the latter formulations require the dynamic generation of two different types of variables (routes and production plans) through structurally different pricing problems. Furthermore, we study different variations of the BPCs with production-plan variables resulting from the inclusion or not of additional bounds on the production quantities in the associated pricing problems. An extensive computational study analyzes the performance of all