Transcriptome-based exon capture approaches, along with next-generation sequencing, are allowing for the rapid and cost-effective production of extensive and informative phylogenomic datasets from non-model organisms for phylogenetics and population genetics research. These approaches generally employ a reference genome to infer the intron-exon structure of targeted loci and preferentially select longer exons. However, in the absence of an existing and well-annotated genome, we applied this exon capture method directly, without initially identifying intron-exon boundaries for bait design, to a group of highly diverse Haloniscus (Philosciidae), paraplatyarthrid and armadillid isopods, and examined the performance of our methods and bait design for phylogenetic inference. Here, we identified an isopod-specific set of single-copy protein-coding loci, and a custom bait design to capture targeted regions from 469 genes, and analysed the resulting sequence data with a mapping approach and newly-created post-processing scripts. We effectively recovered a large and informative dataset comprising both short (<100 bp) and longer (>300 bp) exons, with high uniformity in sequencing depth. We were also able to successfully capture exon data from up to 16-year-old museum specimens along with more distantly related outgroup taxa, and efficiently pool multiple samples prior to capture. Our well-resolved phylogenies highlight the overall utility of this methodological approach and custom bait design, which offer enormous potential for application to future isopod, as well as broader crustacean, molecular studies.