Drug policy needs reliable measures of the prevalence and extent of different forms of illicit drug use. Drug agency statistics shed only a little light on the size of the powder cocaine market. There is very little data on levels of consumption, with little or no information being provided by the main household surveys in the United Kingdom. Traditionally the size of illicit drug markets has been estimated by reference to measures of supply, drawing on statistics of production or seizure. The alternative is to build up a picture of the size of illicit drug markets from information about consumption. There are two methods to estimating the volume of drugs consumed: (1) the direct method, which involves asking representative samples of the general population about their drug use through household surveys; and (2) the indirect method which relies on inferring about drug use in the general population from subsets of known drug users for whom reliable consumption data exist. Direct methods are unlikely to work well for all categories of users. Indirect methods are likely to be viable only for problem drug users, for whom drug services have accumulated a fair amount of information. The first step in trying to improve on this approach was to develop a new user classification. Powder cocaine users were divided into four groups, according to levels of use: rare users, occasional users, frequent users, and problem users. The next step was to derive methods for estimating prevalence and consumption that were appropriate to each of the four groups. Direct estimation methods were likely to be best for the first two groups -- rare and occasional users. For frequent users the best bet still appears to be the British Crime Survey, though analysis suggests that there may be scope for some corrections to adjust for both sample bias and response bias. Indirect estimation methods seem more viable than direct means for problem drug users. It is essential that epidemiological research into powder cocaine use is accompanied by qualitative research on cocaine use. 10 references