In response to the increasing loss of native vegetation and biodiversity, a growing number of countries have adopted ''offsetting'' policies that seek to balance local habitat destruction by restoring, enhancing and/or protecting similar but separate habitat. Although these policies often have a stated aim of producing a ''net gain'' or ''no net loss'' in environmental benefits, it is challenging to determine the potential impacts of a policy and if, or when, it will achieve its objectives. In this paper we address these questions with a general approach that uses predictive modelling under uncertainty to quantify the ecological impacts of different offset policies. This is demonstrated with a case study to the west of Melbourne, Australia where a proposed expansion of Melbourne's urban growth boundary would result in a loss of endangered native grassland, requiring offsets to be implemented as compensation. Three different offset policies were modelled: i) no restrictions on offset location, ii) offset locations spatially restricted to a strategically defined area and iii) offset locations spatially and temporally restricted, requiring all offsets to be implemented before commencing development. The ecological impact of the policies was determined with a system model that predicts future changes in the extent and condition of native grassland. The case study demonstrates how relative and absolute policy performance can be quantified in relation to best and worst-case scenarios. The study also shows that the ecological benefits of being temporally and spatially strategic in choosing offsets locations are substantially greater than being spatially strategic alone. We also show that even with considerable uncertainties in the system model predicting future grassland condition, the performance of the three offset policies can still be differentiated. Finally, we show the extent to which a policy achieves a ''net gain'' is dependent on the baseline against which policy performance is measured. The quantitative framework presented here can also be used to evaluate other offset policies or extended to deal with different types of environmental policies.