Conditions experienced early in development can affect the future performance of individuals and populations. Demographic theories predict persistent population impacts of past resources, but few studies have experimentally tested such carry-over effects across generations or cohorts. We used bumble bees to test whether resource timing had persistent effects on within-colony dynamics over sequential cohorts of workers. We simulated a resource pulse for field colonies either early or late in colony development and estimated colony growth rates during pulse- and non-pulse periods. During periods when resources were not supplemented, early-pulse colonies grew faster than late-pulse colonies; early-pulse colonies grew larger as a result. These results reveal persistent effects of past resources on current growth and support the importance of transient dynamics in natural ecological systems. Early-pulse colonies also produced more queen offspring, highlighting the critical nature of resource timing for population, as well as colony, dynamics of a key pollinator.