We would like to thank the discussers for the interesting points raised in their interpretation of the undrained shear strength (s u) of Finnish and Scandinavian soft clays (Mesri and Wang 2017). The objective of this reply is to further examine the observations made by the discussers on the correlations for s u of Finnish soft clays from the multivariate database reported in the paper under discussion (D'Ignazio et al. 2016). In this reply, the databases discussed are labelledfor example, F-CLAY/10/216-to describe their contents in terms of country 2-soil type / number of parameters of interest / number of data points. The aim of our study (D'Ignazio et al. 2016) was to establish a global correlation model to estimate s u of Finnish soft clays and its transformation uncertainty explicitly, when only information on OCR and index parameters was available. The idea was that this model could supplement site-specific s u data when they are available through, for example, Bayesian updating (Ching and Phoon 2014; Ching et al. 2014). The data points from 24 different sites in Finland were treated as one population to establish a global description and correlation for the undrained behaviour of Finnish clays. During the course of the work, we screened out some data points based on the considerations described in our paper (D'Ignazio et al. 2016). The database with the entire data population was F-CLAY/ 10/216. The database resulting from the screening in D'Ignazio et al. (2016) was F-CLAY/10/173. The conclusions in D'Ignazio et al. (2016) on the global OCR-(s u FV / vo)-Y j model used the latter more refined set of data points. The secondary input parameters Y j analysed were natural water content (w), liquid limit (LL), plastic limit (PL), plasticity index (PI), liquidity index (LI), and sensitivity (S t). The s u data contained in F-CLAY/10/216 were measured in situ from field vane (FV) tests, while preconsolidation stresses (p) were inferred from incrementally loaded (IL) and constant-rate-of-strain (CRS) oedometer tests. Hence, the OCR-(s u FV / vo)-Y j model derived