Inverter-interfaced distributed generation (DG) can provide voltage support in unbalanced distribution feeders; however, sequential, phase-selective siting and sizing under local interconnection constraints have received limited attention. This study proposes a constructive Volt/VAR planning method that allocates DG according to the apparent demand and voltage-deficient phases at each candidate bus. A descending reactive power feasibility test rejects solutions that cause non-convergence, overvoltage, or deterioration of the minimum system voltage. The method was implemented through automated unbalanced three-phase power-flow simulations and tested on the IEEE 13-node feeder and the real 3849-bus CHIRIPA medium-voltage network in Honduras, using active power allocation factors of 0.7 and 0.9. The selected configurations were evaluated over 24 h profiles for representative sunny and rainy days, annualized, and assessed economically over a 25-year horizon. Annual energy losses decreased by 25.8% and 26.4% in the IEEE feeder, although local capacity constraints prevented all phase voltages from reaching the 0.95–1.05 p.u. range. In CHIRIPA, all medium-voltage bus-phase voltages satisfied this range at maximum demand, although this improvement persisted only during solar hours; annual losses fell by 15.1% and 18.7%. The economic feasibility threshold in CHIRIPA depended primarily on installed capacity (approximately 6.5–6.6 MW), rather than unit count. These results provide a reproducible framework for prioritizing phase-specific DG investments under technical and economic constraints.