The conventional design approach of soil nails often results in an underestimation of bond stress, which in turn gives an uneconomical design. Therefore this paper aims to propose a methodology to account for the effect of pullout tests on the bond stress. This procedure is formalized in the Bayesian framework. The prior distribution of bond stress is formulated on the basis of the available pullout test data. Three cases of Bayesian updating of bond stress based on pullout tests are studied; namely, all pullout tests not conducted to failure, a part of pullout tests conducted to failure, and all pullout tests conducted to failure. The results indicate that the updated coefficient of variation of bond stress can be significantly smaller than that before the pullout tests. The updated mean bond stress is higher than the prior value when all pullout tests do not conduct to failure. The updated mean bond stress is smaller than the prior value when all pullout tests conduct to failure. With the increase of test loads, the updated mean bond stress increases, but the updated coefficient of variation decreases.
