The roll-out phase of the next generation of mobile networks (5G) has started and operators are required to devise deployment solutions while pursuing localization accuracy maximization. Enabling location-based services is expected to be a unique selling point for service providers now able to deliver critical mobile services, e.g., autonomous driving, public safety, remote operations. In this paper, we propose a novel roll-out base station placement solution that, given a Throughput-Positioning Ratio (TPR) target, selects the location of new-generation base stations (among available candidate sites) such that the throughput and localization accuracy are jointly maximized. Moving away from the canonical position error bound (PEB) analysis, we develop a realistic framework in which each positioning measurement is affected by errors depending upon the actual wireless channel between the measuring base station and the target device. Our solution, referred to as LOKO, is a fast-converging algorithm that can be readily applied to current 5G (or future) roll-out processes. LOKO is validated by means of an exhaustive simulation campaign considering real existing deployments of a major European network operator as well as synthetic scenarios.