Sentinel lymph node (SLN) biopsy has the potential to become a standard operation for most patients with operable breast cancer, but raises a number of issues for the surgeon and institution wishing to start a SLN program. We began to perform SLN biopsy for breast cancer at Memorial Sloan-Kettering Cancer Center in September of 1996. Based on a pilot study of 60 patients, detailed analyses of our first 500 and 1000 cases, and a cumulative experience of 1500 procedures, we report the lessons learned in starting a SLN program, emphasizing study design, case selection, technique and quality control. All patients had clinical T1-2N0 breast cancers and isotope plus blue dye mapping. The combination of blue dye and isotope localization was superior to either method alone, unfiltered technetium sulfur colloid was superior to filtered, and optimum isotope localization was achieved with a low-volume, low-dose intradermal injection. In our first 1000 cases, 14% of SLN were found by isotope alone, and 8% by dye alone. 10% of positive SLN were found by isotope alone, and 11% by dye alone. Failed and false-negative SLN procedures were less frequent as experience increased.