Most probable number estimates are used when one wishes, for example to measure the concentration of parasites or bacteria in a solution. Being able, only to detect presence or absence of the quantity of interest, a maximum likelihood method is used to estimate the unknown concentration. The classical approach for estimating the confidence interval for the most likely concentration is based on asymptotic results. The accuracy of this kind of approximations might be bad especially for small samples. We propose to calculate confidence intervals based on bootstrap percentiles, which are very reliable, as experiments will show.