Obtaining durability for a known service life in chloride exposures requires knowledge of the concrete properties, relevant transport processes, depth of cover as well as minimization of cracking and construction defects. For example, imperfect curing can result in depth-dependent effects on resistance to chloride ingress. Several service life models with various levels of sophistication exist for prediction of time-to-corrosion of concrete structures exposed to chlorides. The model inputs have uncertainty associated with them such as boundary conditions (level of saturation and temperature), cover depths, diffusion coefficients, time-dependent changes, and rates of buildup of chlorides at the surface. The performance test methods used to obtain predictive model inputs as well as how models handle these properties have a dramatic impact on predicted service lives. Very few models deal with the influence of cracks or the fact that concrete in the cover zone will almost certainly have a higher diffusion coefficient than the bulk concrete as the result of imperfect curing or compaction. While many models account for variability in input properties, they will never be able to account for extremes in construction defects. Therefore, to ensure the reliability of service life predictions and to attain a concrete structure that achieves its predicted potential, designers, contractors and suppliers need to work together to ensure proper detailing, minimize defects, and adopt adequate, yet achievable, curing procedures. As well, concrete structures are often exposed to other destructive elements in addition to chlorides (eg frost, ASR) and this adds another level of complexity since regardless of cause, cracks will accelerate the ingress of chlorides. These issues are discussed along with the need to use performance-based specifications together with predictive models.