Predicting behavior of soil in the field in response to construction activities and natural events is the reason why so much effort in the past 50 years has been put into developing constitutive models. It is the field responses of soils that are of interest to geotechnical practitioners, and this aspect of the problem has not received nearly enough attention. There are many logical reasons for this, e.g., the need for establishing behavior under well-defined boundary conditions, the difficulty of obtaining high quality samples, the effects of sample disturbance, the effort required to obtain field performance data of sufficient quality and quantity such that it can be correlated with a causative construction activity, to name a few. The author and his colleagues have collected detailed field performance data from a number of supported excavations in the Chicago area and have correlated the responses with construction activities. As part of the effort, block samples have been cut as the excavations have proceeded. Detailed laboratory evaluations of the stress-strain strength responses have been conducted on specimens from these blocks and from thin-wall tube samples. Inverse analyses have been conducted on results of finite element simulations of the excavations using the field observations as a basis of comparison so that best fit parameters for several constitutive models were determined based on the field data. This paper summarizes the techniques used and describes the lessons learned from this effort. Comments are made regarding applicability of laboratory data, the relation between the capabilities of the constitutive model and selection of observations for inverse analyses and numerical modeling.