Effective population size (Ne) is a useful parameter to evaluate the long-term viability of populations. While obtaining enough field data from wild populations to estimate Ne directly is challenging, molecular techniques applied to non-invasive samples provide an appealing alternative. However, Ne estimates may be biased by both ecological and sampling-related factors. In polygamous species, the monopolization of mating tends to reduce the ratio between the effective population and the census population (Ne: Nc) compared to idealized populations. Ne estimates may also be affected by sample size, spatial structure, and relatedness, factors that are often overlooked in studies of wild populations. We investigated how sample size, spatial structure, and relatedness impact Ne estimates in an endangered population of a polygynous forest grouse, the Cantabrian capercaillie, inhabiting the mountains of north-western Spain. We studied  135 capercaillie individuals based on DNA extracted from droppings and shed feathers. We estimated contemporary Ne based on linkage disequilibrium. Ne estimates stabilized at sample sizes of ~20 individuals, with confidence intervals markedly narrowing beyond 30 individuals, i.e., at 23 to 35% of the estimated population size. Samples taken in specific valleys yielded consistently lower Ne values compared to similar-sized random samples from the entire study area, indicating that the spatial structure of the sampling design also affected Ne. We observed a negative correlation between Ne and relatedness, which highlighted the need to account for kinship structure in sampling design. Our estimated Ne: Nc ratio (~15%) was not as low as expected for a polygynous species with male-biased sex ratios and inhabiting a fragmented habitat. However, the absolute Ne values were extremely small. We showed the relevance of integrating field-based observations of demography and spatial ecology into DNA-based estimates of Ne to improve its interpretation and usefulness in conservation efforts.