Digital PCR (dPCR) offers absolute quantification through the limiting dilution of template nucleic acid molecules and has the potential to offer high reproducibility. However, the robustness of dPCR has yet to be evaluated using complex genomes to compare different dPCR methods and platforms. We used DNA templates from the pathogen Mycobacterium tuberculosis to evaluate the impact of template type, master mixes, primer pairs and, crucially, extraction methods on dPCR performance. Performance was compared between the chip (BioMark) and droplet (QX100) formats. In the absence of any external calibration, dPCR measurements were generally consistent within ∼2-fold between different master mixes and primers. Template DNA integrity could influence dPCR performance: high molecular weight gDNA resulted in underperformance of one master mix, while restriction digestion of a low molecular weight sample also caused underestimation. Good concordance (≤1.5-fold difference) was observed between chip and droplet formats. Platform precision was in agreement with predicted Poisson error based on partition number, but this was a minor component (<10%) of the total variance when extraction was included. dPCR offers a robust reproducible method for DNA measurement; however, as a predominant source of error, the process of DNA extraction will need to be controlled with suitable calibrators to maximize agreement between laboratories.