New emerging and re-emerging threats, the weight of public opinion and new technology for surveillance and treatment are likely to impact on how, and if, effective surveillance can be performed in the future. If surveillance fails to address the needs of practitioners and policy-makers, it is likely that there will be loss of confidence. Current surveillance systems are reasonably effective at detecting significant events that are localised in time and space. It is more difficult to detect diffuse and progressive events with a slow increase over time or sporadic and widespread events without obvious links to time, place or person. Detection of these events relies on good data collection, comparative background data and sophisticated analytical tools. To improve surveillance systems, we need methods with the appropriate sensitivity and specificity for the outputs desired. Targeted surveillance should enable better ascertainment of those cases which must be considered and those which can be dismissed. New methods, such as mathematical modelling and geographical information systems, support conventional surveillance in moving events into the known and predictable category. It is important to integrate surveillance across local, regional and international levels and to base surveillance on local public health structures. The purpose and value of data aggregation at each level and the amount of detail needed at each level must be carefully evaluated. The key to all these improvements is developing the workforce. Surveillance needs individuals with a broad range of skills: clinical, epidemiological, anthropological, and mathematical; in particular, people who can think laterally. These individuals must be encouraged through effective training courses, good mentorship, networking and clear career structures.