The sources of variance and errors in transcriptomics based on the electrophoretic separation of amplified cDNA fragments were investigated using cDNA-amplified fragment length polymorphism (AFLP). Transcriptome profiles of the plant-pathogenic fungus Verticillium longisporum were generated by a standard cDNA-AFLP protocol followed by electrophoretic separation of amplified DNA fragments in flatbed polyacrylamide gels with fluorescence detection as well as by capillary electrophoresis (DNA sequencer). The total variance was partitioned into contributions of cDNA synthesis, adapter ligation, preamplification, amplification, and electrophoresis. Parameters of computer-aided peak recognition and matching were investigated and strategies improving matching success based on double passage with different signal intensity thresholds were developed. The overall quality of data was similar for cDNA-AFLP and microarray hybridization. Variance of cDNA-AFLP was independent of signal intensity, whereas microarray data showed higher variance for low-intensity signals. Capillary electrophoresis significantly reduced the number of wrongly matched and unmatched signals as compared with flatbed gels. These results are also likely to apply to related electrophoresis-based transcriptome analysis techniques such as mRNA differential display.