•The comet assay is widely used in human biomonitoring to measure DNA damage.•Variation in comet assay results between laboratories has been a problem.•Specific steps in the assay have been identified as causes of this variation.•Including reference standards in experiments helps to control variation.•We give recommendations for improving the reliability of the assay. The comet assay (single cell gel electrophoresis) is widely used as a biomonitoring tool to assess DNA damage – strand breaks, as well as oxidised bases; it can also be adapted to measure DNA repair. It is based on the ability of breaks in the DNA to relax supercoiling, allowing DNA loops to extend from the nuclear core (nucleoid) under an electric field to form a comet-like tail. Most commonly, it is applied to white blood cells. The range of detection is between a few hundred breaks per cell and a few thousand, encompassing levels of damage that can be repaired and tolerated by human cells. Its applications include monitoring various diseases, studying the influence of nutrition on DNA stability, and investigating effects of environmental and occupational mutagens. Here we address the issue of inter-laboratory variation in comet assay results. This variation is largely due to differences in methods. Imposing a standard protocol is not practical, but users should be aware of the crucial parameters that affect performance of the assay. These include the concentration of agarose in which the cells are embedded; the duration of cell lysis, and of enzyme incubation when oxidised bases are being measured; the duration of alkaline unwinding; the duration of electrophoresis and the voltage gradient applied; and the method used to score the comets. Including reference standards in each experiment allows experimental variability to be monitored – and if variation is not extreme, results can be normalised using reference standard values. Reference standards are also essential for inter-laboratory comparison. Finally, we offer recommendations which, we believe, will limit variability and increase the usefulness of this assay in molecular epidemiology.