In order to set up the precision limit that can be reached with laser-induced breakdown spectrometry microprobe, a steel sample was scanned by using a 6-µm diameter spot. Besides being close to the limit of the technique, such a spot diameter resulted in a small plasma size that minimized self-absorption effects. To minimize shot noise, Cr and Fe were used as test elements because of their high contents. Scan consisted of 25 successive matrices formed by 5×6 shots, i.e. a total of 750 shots. Results were studied as a whole, as well as between matrices and within matrices, to search for inhomogeneity, outliers and drift. Except a few outliers, the main contribution in the experimental RSD was the drift either within a matrix or between matrices. Drift attributed to laser warm up could be compensated for either by using a polynomial fitting or by using the other element. %RSD significantly below 2 were then obtained demonstrating that there is no penalty in terms of precision to perform laser microprobe using a series of single shots.