Soluble formazan assays are widely used for cell number assessment. However, in our hands, we observed frequent occasions in which the actual cell number was at odds with the assay reading. In this study, we have determined that (i) a large proportion of the reading obtained in commonly used culture media can be caused by media component amplification of formazan production in a way that cannot be corrected for by media-only controls; (ii) the albumin present in 10% serum can reduce the assay absorbance by 50% so that an actual doubling of cell number can be obscured; and (iii) this latter effect is dependent on the concentration of fatty acids. To counter these problems, we have developed a protocol that gives consistent readings that are fully representative of cell number while retaining some of the original advantages of soluble formazan assays.