OBJECTIVES Uncertain decisions made using a cost-effectiveness threshold applied to each decision problem separately fail to identify the true opportunity costs of displacing other unrelated programmes. We show that the allocation problem can be characterised to provide a more general and comprehensive approach to informing adoption and research decisions. METHODS A stochastic mathematical programming approach is used to solve the allocation problem. The formulation allows the characterisation of actual budgetary policies, including a strict budgetary rule where deficits are not possible and constraints must always be met. The opportunity costs (health forgone due to curtailing some programmes and treatments) of violating the budget constraint are incorporated directly. In addition, the value of acquiring new evidence to inform the allocation problem in light of its current uncertainty is considered simultaneously and consistently. RESULTS The allocation and research decision problem depends on a number of considerations: 1) size of overall budget; 2) budgetary policy in place; 3) information that is revealed and its timing; 4) subsequent actions available to decision makers; and 5) costs of effectively monitoring ex-ante plans. Standard decision rules in cost-effectiveness analysis are only optimal under very special circumstances, which require budget constraints to be soft in addition to assumptions of perfect divisibility, constant returns and all costs and benefits occurring within the budgetary period. However, if the budget constraint is hard then technologies will need to be more cost-effective (an incremental cost-effectiveness ratio substantially below the threshold) before the decision maker should take the risk of an ex-ante decision to adopt them. CONCLUSIONS Standard decision rules and measures of value are proxies for an uncertain and complex process. There are no simple ex-ante decision rules in most common circumstances and the value of information cannot be established for one programme independently of the rest of the allocation problem.