Policymakers such as English healthcare commissioners are encouraged to adopt ‘evidence-based policy-making’, with ‘evidence’ defined by researchers as academic research. To learn how academic research can influence policy, researchers need to know more about commissioning, commissioners’ information seeking behaviour and the role of research in their decisions.

In case studies of four commissioning organisations, we interviewed 52 people including clinical and managerial commissioners, observed 14 commissioning meetings and collected documentation e.g. meeting minutes and reports. Using constant comparison, data were coded, summarised and analysed to facilitate cross case comparison.

The ‘art of commissioning’ entails juggling competing agendas, priorities, power relationships, demands and personal inclinations to build a persuasive, compelling case. Policymakers sought information to identify options, navigate ways through, justify decisions and convince others to approve and/or follow the suggested course. ‘Evidence-based policy-making’ usually meant pragmatic selection of ‘evidence’ such as best practice guidance, clinicians’ and users’ views of services and innovations from elsewhere. Inconclusive or negative research was unhelpful in developing policymaking plans and did not inform disinvestment decisions. Information was exchanged through conversations and stories, which were fast, flexible and suited the rapidly changing world of policymaking. Local data often trumped national or research-based evidence. Local evaluations were more useful than academic research.

Commissioners are highly pragmatic and will only use information that helps them create a compelling case for action.Therefore, researchers need to start producing more useful information.

To influence policymakers’ decisions, researchers need to 1) learn more about local policymakers’ priorities 2) develop relationships of mutual benefit 3) use verbal instead of writtencommunication 4) work with intermediaries such as public health consultants and 5) co-produce local evaluations.

The online version of this article (doi:10.1186/s12913-015-1091-x) contains supplementary material, which is available to authorized users.