Travel time reliability is an important topic in transport policy and transport research. However, the value of improving reliability is often not included in cost-benefit analysis, or computed rather crudely. The value of reliability is more or less known, but the amount of unreiliability is often not included in empirical transport models. The optimal solution is to use state-of the-art models to predict reliability. However, the development, estimation and application of such models takes several years. As a first step towards improvement, we develop a practical method to estimate the volume of unreliability, by analysing the relation between mean travel times and the standard deviation of travel times on trunk roads. This relation appears to be approximately linear, with different slopes for different types of roads, but relatively stable within these categories. Using these results, we devise a typology of road types and of policies for improvement, such as road expansion, road pricing and traveler information. The results are used to recommend improved rules-of-thumb for estimating the benefits of travel time reliability in cost-benefit analysis.