Ground engineering in urban areas faces the great challenge of balancing the increasing demand for underground space against safety and asset protection while avoiding high construction costs. This study shows savings can be achieved on embedded retaining wall design for deep excavation in London Clay through the observational method, without compromising safety. Despite the inherent benefits of the method and its acceptance by design codes, the application of the observational method for excavation design has been slow and inconsistent due to the lack of guidance, in addition to other difficulties. This research aims to promote the application of the observational method in excavation design by proposing a new framework. The framework with four design approaches is established based on the review of historical excavation case histories and four Crossrail station excavations using the observational method. The term Ab initio is used for excavation design from the beginning of construction, covering Optimistic Approach A and Cautious Approach B. The term Ipso-tempore is introduced for excavation redesign after construction starts, comprising a newly defined Pro-active Approach C and the ‘best-way-out’ or Reactive Approach D. Back-analysis is critical in the observational method. The whole process of back analysis is examining monitoring systems (observations) and predictions by the numerical analysis with the adopted soil constitutive models. Three Crossrail excavation cases are back-analysed by the Mohr-Coulomb model using FEMs from Pseudo-FEM to 2D and 3D FEMs. The Crossrail Tottenham Court Road Station, Western Ticket Hall deep box excavation is also back-analysed by the BRICK model using the 3D FEM, representing the advanced soil model. The above back-analysis results are presented including the calibrated most probable soil model parameters. The different results indicate the back-analysis is subject to the type of numerical analysis, the adopted soil constitutive model, also it needs to be tailored to the monitoring data used for comparison. A reassessment of the TCR-WTH excavation design by Approach A is carried out using the semi-FEM with the Mohr-Coulomb model. The optimistic design with the calibrated most probable Mohr-Coulomb parameters shows over 30% saving in construction materials, which is supported by the contingency plan with the characteristic Mohr-Coulomb parameters if the excavation does not encounter the expected conditions. Improvements of specifications for the instruments and monitoring data are recommended to provide more reliable monitoring data.