Standardized protocols for surveying secretive marsh birds have been implemented across North America, but the efficacy of surveys to detect population trends has not been evaluated. We used survey data collected from populations of marsh birds across North America and simulations to explore how characteristics of bird populations (proportion of survey stations occupied, abundance at occupied stations, and detection probability) and aspects of sampling effort (numbers of survey routes, stations/route, and surveys/station/year) affect statistical power to detect trends in abundance of marsh bird populations. In general, the proportion of survey stations along a route occupied by a species had a greater relative effect on power to detect trends than did the number of birds detected per survey at occupied stations. Uncertainty introduced by imperfect detection during surveys reduced power to detect trends considerably, but across the range of detection probabilities for most species of marsh birds, variation in detection probability had only a minor influence on power. For species that occupy a relatively high proportion of survey stations (0.20), have relatively high abundances at occupied stations (2.0 birds/station), and have high detection probability (0.50), ≥40 routes with 10 survey stations per route surveyed 3 times per year would provide an 80% chance of detecting a 3% annual decrease in abundance after 20 years of surveys. Under the same assumptions but for species that are less common, ≥100 routes would be needed to achieve the same power. Our results can help inform the design of programs to monitor trends in abundance of marsh bird populations, especially with regards to the amount of sampling effort necessary to meet programmatic goals.