Phage research has considerably intensified in the last decade, yet clinical phage therapy still proceeds with mixed success. Therapeutic phages are optimally given as cocktails of two or more phage strains, each targeting different host binding receptors. While refining cocktail design for phase I/II clinical trials, strong biological rationale is required for the inclusion of each phage in formulation. Currently, we cannot ensure that all phages given during phage therapy are active in vivo. Moreover, a major hurdle in cocktail design lies in the fact that phages are selfish elements that interfere with infection by other strains. This suggests that phages in combination are unlikely to exhibit synergism and can only interact additively or antagonistically. We determined the lytic activities of Pseudomonas aeruginosa 2-phage cocktails in vitro, in silico, and in patient. We measured the activities of three phages administered as single, 2-phage simultaneous, or 2-phage sequential treatments in well-mixed in vitro time-kill microcultures. We observed the greatest lytic activity (lowest OD₆₀₀) from 2-phage simultaneous treatments. Applying mathematical models to the time-kill kinetics, we predicted that the two phages additively improved bacterial lysis by up to 15%. The increased lytic activity also promoted more robust growth of both phage strains over a longer period of time. Next, we mimicked the spatial complexity and hydraulic pressures of an in vivo circulatory system using an in vitro model. In the hollow fiber infection model (HFIM), 2-phage cocktails additively reduced bacterial numbers for a longer period of time (>24h) compared to single phage (<12h). In contrast, phage numbers suggest antagonism between phage strains. Replication of one of the two phage strains was delayed for up to 12h in the HFIM. Returning to time-kill microcultures, the early timepoints (0-3h) of 2-phage simultaneous treatment showed reduced lytic activity from cocktail treatment compared to singular phage activity. Combined, this suggests that the lytic activity of phage cocktails are additive at the population level but antagonistic at the cellular level. In vivo, however, host factors can interfere with the additive properties of cocktails. In patient, this was observed during intravenous 2-phage cocktail treatment of P. aeruginosa pneumonia. Lung sputum metagenomes collected on days 4, 8, and 13 of treatment showed high relative abundance of only one of the two treatment phages. Despite both phages being administered twice daily at equal ratio, this suggests that only one phage strain was lytically active in vivo. Together this suggests that when designing phage cocktails, it is important to dose phages based on each strain’s lytic activity to provide the greatest additive effect. A strain-centric view of phage cocktails may therefore improve the in vivo success of therapy.