For over a decade now, High Throughput sequencing (HTS) approaches have revolutionized phylogenetics, both in terms of data production and methodology. While transcriptomes and (reduced) genomes are increasingly used, generating and analyzing HTS datasets remain expensive, time consuming and complex for most non-model taxa. Indeed, a literature survey revealed that 74% of the molecular phylogenetics trees published in 2018 are based on data obtained through Sanger sequencing. In this context, our goal was to identify the strategy that would represent the best compromise among costs, time and robustness of the resulting tree. We sequenced and assembled 32 transcriptomes of the marine mollusk family Turridae, considered as a typical non-model animal taxon. From these data, we extracted the loci most commonly used in gastropod phylogenies (cox1, 12S, 16S, 28S, h3 and 18S), full mitogenomes, and a reduced nuclear transcriptome representation. With each dataset, we reconstructed phylogenies and compared their robustness and accuracy. We discuss the impact of missing data and the use of statistical tests, tree metrics, and supertree and supermatrix methods to further improve phylogenetic data acquisition pipelines. We evaluated the overall costs (time and money) in order to identify the best compromise for phylogenetic data sampling in non-model animal taxa. Although sequencing full mitogenomes seems to constitute the best compromise both in terms of costs and node support, they are known to induce biases in phylogenetic reconstructions. Rather, we recommend to systematically include loci commonly used for phylogenetics and taxonomy (i.e. DNA barcodes, rRNA genes, full mitogenomes, etc.) among the other loci when designing baits for capture.