Evolutionary biologists increasingly use pedigree-based quantitative genetic methods to address questions about the evolutionary dynamics of traits in wild populations. In many cases, phenotypic data may have been collected only for recent parts of the study. How does this influence the performance of the models used to analyse these data? Here we explore how data depth (number of years) and completeness (number of observations) influence estimates of genetic variance and covariance within the context of an existing pedigree. Using long-term data from the great tit Parus major and the mute swan Cygnus olor, species with different life-histories, we examined the effect of manipulating the amount of data included on quantitative genetic parameter estimates. Manipulating data depth and completeness had little influence on estimated genetic variances, heritabilities, or genetic correlations, but (as expected) did influence confidence in these estimates. Estimated breeding values in the great tit were not influenced by data depth but were in the mute swan, probably because of differences in pedigree structure. Our analyses suggest the ‘rule of thumb’ that data from 3 years and a minimum of 100 individuals per year are needed to estimate genetic parameters with acceptable confidence, and that using pedigree data is worthwhile, even if phenotypes are only available toward the tips of the pedigree.