The article shows that the use of the normal distribution for statistical processing of the results of mechanical testing of soils, in particular, when determining the bearing capacity of piles in terms of soil, can lead, in case of a small number of tests and the presence of outliers in the experimental data, to an illegal underestimation of the calculated indicator to values less than the minimum values obtained in the tests performed. It is proposed to replace this distribution with a lognormal one that allows avoiding "implausible" conclusions.