Several operating parameters for single-droplet analysis by inductively coupled plasma‐atomic emission spectrometry were investigated and optimized. Two plasma observation modes, both of which measure the plasma side-on, were compared. In the “whole-vertical” mode, the entire vertical emission pattern of the center portion of the central channel was spatially integrated, whereas in the “lateral” mode emission from a thin horizontal slice of the vertical plasma image was measured. The limits of detection (LOD) as well as measurement precision attainable by these two observation modes were found to be practically identical. However, the lateral mode is preferred because emission is then more insensitive to a small drift in carrier-gas flow than in the vertical mode. Precision was found to degrade at carrier-gas flows that yield maximum sensitivities in both observation modes. As a result, the best precision and lowest LODs cannot be achieved under the same plasma operating conditions and a compromise is needed. In this study, precision was given a higher priority than LOD because each individual droplet is regarded as a new sample in single-droplet analysis and each such sample can be measured only once. For best precision, the observation region should be 3mm downstream of the atomization site to avoid the adverse local plasma cooling effect of the vaporizing particle. Under optimized conditions, the best precision is about 3–4% and the absolute detection limits for eleven elements (Ag, B, Ca, Cd, Cu, Fe, Mg, Ni, Pb, Sr, and Zn) range from sub-single to hundreds of femtograms, which corresponds to 106 to 109 atoms for single-droplet analysis. In addition, a new synchronization trigger method for droplet analysis was developed. This method is based on Hα emission collected between the first and second lowest turns of the load coil. This trigger signal fires while the droplet is still intact, resides inside the lowest portion of the load coil, and is typically around 1 to 2ms ahead of the atomization event. Jitter between this trigger signal and analyte emission is typically less than 100μs, and is precise enough for correct timing for signal acquisition and integration.