A major barrier to the use of web data collection methods for probability-based surveys of the general population is the difficulty of achieving good population coverage. This paper provides an overview of the issues to be faced and of possible solutions, some of which are as yet untested. The nature of the issues to be faced, and the possible solutions, are dependent on the main motivation for wanting to consider web data collection, which can be related either to cost reduction or to improving coverage and participation. We therefore outline the implications of each type of motivation. We also emphasise that the issues and solutions are somewhat different for longitudinal surveys and cross-sectional surveys. And there is an important distinction between single-mode web surveys and mixed-mode surveys that include web as one of the modes. These two different types of surveys present rather different challenges.