Quality improvement collaboratives are increasingly being used in many countries to achieve rapid improvements in health care. However, there is little independent evidence that they are more cost effective than other methods, and little knowledge about how they could be made more effective. A number of systematic evaluations are being performed by researchers in North America, the UK, and Sweden. This paper presents the shared ideas from two meetings of these researchers. The evidence to date is that some collaboratives have stimulated improvements in patient care and organisational performance, but there are significant differences between collaboratives and teams. The researchers agreed on the possible reasons why some were less successful than others, and identified 10 challenges which organisers and teams need to address to achieve improvement. In the absence of more conclusive evidence, these guidelines are likely to be useful for collaborative organisers, teams and their managers and may also contribute to further research into collaboratives and the spread of innovations in health care.