Pharmaceutical value, a multidimensional factor that includes proven efficacy, is now crucial to market success. A key parameter of value is reliability in use, determined by the several sources of variance in drug response, e.g. the Harter-Peck model, based on linear pharmacometrics of theophylline, whose dose-response has an estimated 80% coefficient of variation, due mainly to variable pharmacokinetics and non-compliance. The main forms of non-compliance are multiday intervals between doses, the impact of which is difficult to assess because too little is known of the time-course of drug actions after dosing stops ('off-responses'). Omeprazole, the best-selling drug, is an exception, and its off-response data reveal markedly non-linear pharmacodynamics that appear to filter most of the variance that a linear model passes, projecting a big gain in reliability. The impact of variable compliance is attenuated by 'forgiveness', the post-dose duration of effective action (4-5 days for omeprazole) minus its recommended dosing interval.