PCR amplification, a key step in next-generation sequencing (NGS) library construction, can generate an unlimited amount of product from limited input; however, it cannot create more information than was present in the original template. Thus, NGS libraries can be made from very little DNA, but reducing the input may compromise assay sensitivity in ways that are difficult to ascertain unless library complexity (ie, the number of unique DNA molecules represented in the library) and depth of coverage with unique sequence reads (those derived from input DNA molecules) versus duplicate sequence reads (those resulting from overamplification of particular molecules) are discretely measured. A series of experiments was performed to explore the impact of low DNA input on an amplicon-based NGS assay using unique molecular identifiers to track unique versus duplicate reads. At high sequencing depths, unique and total (unique plus duplicate) read coverage are not well correlated, so increasing the number of sequenced reads does not necessarily improve sensitivity. Unique coverage depth tends to improve with more input, but improvements are not consistent. Fluctuations in library complexity complicated variant detection using both standardized and clinical specimens, often resulting in technical replicates with vastly different estimates of variant allelic fraction. In conclusion, depth of coverage with unique reads must be tracked in clinical NGS to ensure that sensitivity and accuracy are maintained.