Although demographic models have become increasingly important tools in plant conservation, few models have considered the implications of seed banks for population persistence. Based on a 15-year study of the threatened herb, Helenium virginicum, we created a stage-class transition matrix to model the population dynamics of the plant. Our goal was to determine the role of the seed bank in population persistence and in the design of monitoring programs for the plant. Using data from marked plants, a long-term study of seed viability, and a seed bank census, we created a deterministic model and three stochastic models. The stochastic models were a model in which yearly correlations among parameters were retained, a model in which parameters were uncorrelated, and a model in which parameters were derived from log-normal distribution. We also constructed a fourth model in which we assumed a minimum seed lifetime consistent with the seed viability data. Both elasticity and perturbation analysis suggested that seed survival within the seed bank had by far the largest effects on the population growth rate (λ), with 47-64% of the change in λ being controlled by seed survival. Correlations among life history parameters had little effect on λ in the original models, but substantially reduced λ (from 0.97 to 0.86) when seed survival was limited. Given the importance of the seed bank and the high yearly variability in adult plant density, we used simulations to compare power to detect declines with seed bank samples versus censuses of adult plants. The power of adult plant censuses was extremely low (13-22%). The power of seed bank censuses was substantially higher (48-59%), but was limited by large pulses of recruitment to the seed bank in good years. Power was only moderately reduced when seeds were sampled every two or four years instead of yearly (from 59% to 48%). Together, our results suggest that seed survival is crucial to persistence of Helenium populations and that future empirical studies should focus on understanding the factors that affect seed survival. In addition, managers should consider seed bank censuses rather than above-ground plants when designing monitoring programs for plants in variable environments where above-ground plants vary greatly from year to year.