Although DNA barcoding has become a useful tool for species identification and biodiversity surveys in plant sciences, there remains little consensus concerning appropriate sampling strategies and the treatment of indels. To address these two issues, we sampled 39 populations for nine <em>Taxus</em> species across their entire ranges, with two to three individuals per population randomly sampled. We sequenced one core DNA barcode (<em>mat</em>K) and three supplementary regions (<em>trn</em>H-<em>psb</em>A, <em>trn</em>L-<em>trn</em>F and ITS) for all samples to test the effects of sampling design and the utility of indels. Our results suggested that increasing sampling within-population did not change the clustering of individuals, and that meant within-population <em>P</em>-distances were zero for most populations in all regions. Based on the markers tested here, comparison of methods either including or excluding indels indicated that discrimination and nodal support of monophyletic groups were significantly increased when indels were included. Thus we concluded that one individual per population was adequate to represent the within-population variation in these species for DNA barcoding, and that intra-specific sampling was best focused on representing the entire ranges of certain taxa. We also found that indels occurring in the chloroplast <em>trn</em>L<em>-trn</em>F and <em>trn</em>H-<em>psb</em>A regions were informative to differentiate among for closely related taxa barcoding, and we proposed that indel-coding methods should be considered for use in future for closed related plant species DNA barcoding projects on or below generic level.