Background: Considerable resources are spent on research to establish what works to improve the nation’s health. If the findings from this research are used, better health outcomes can follow, but we know that these findings are not always used. In public health, evidence of what works may not ‘fit’ everywhere, making it difficult to know what to do locally. Research suggests that evidence use is a social and dynamic process, not a simple application of research findings. It is unclear whether it is easier to get evidence used via a legal contracting process or within unified organisational arrangements with shared responsibilities. Objective: To work in cocreation with research participants to investigate how research is utilised and knowledge mobilised in the commissioning and planning of public health services to reduce alcohol-related harms. Design, setting and participants: Two in-depth, largely qualitative, cross-comparison case studies were undertaken to compare real-time research utilisation in commissioning across a purchaser–provider split (England) and in joint planning under unified organisational arrangements (Scotland) to reduce alcohol-related harms. Using an overarching realist approach and working in cocreation, case study partners (stakeholders in the process) picked the topic and helped to interpret the findings. In Scotland, the topic picked was licensing; in England, it was reducing maternal alcohol consumption. Methods: Sixty-nine interviews, two focus groups, 14 observations of decision-making meetings, two local feedback workshops ( n = 23 and n = 15) and one national workshop ( n = 10) were undertaken. A questionnaire ( n = 73) using a Behaviourally Anchored Rating Scale was issued to test the transferability of the 10 main findings. Given the small numbers, care must be taken in interpreting the findings. Findings: Not all practitioners have the time, skills or interest to work in cocreation, but when there was collaboration, much was learned. Evidence included professional and tacit knowledge, and anecdotes, as well as findings from rigorous research designs. It was difficult to identify evidence in use and decisions were sometimes progressed in informal ways and in places we did not get to see. There are few formal evidence entry points. Evidence (prevalence and trends in public health issues) enters the process and is embedded in strategic documents to set priorities, but local data were collected in both sites to provide actionable messages (sometimes replicating the evidence base). Conclusions: Two mid-range theories explain the findings. If evidence has saliency (relates to ‘here and now’ as opposed to ‘there and then’) and immediacy (short, presented verbally or visually and with emotional appeal) it is more likely to be used in both settings. A second mid-range theory explains how differing tensions pull and compete as feasible and acceptable local solutions are pursued across stakeholders. Answering what works depends on answering for whom and where simultaneously to find workable (if temporary) ‘blends’. Gaining this agreement across stakeholders appeared more difficult across the purchaser–provider split, because opportunities to interact were curtailed; however, more research is needed. Funding: This study was funded by the Health Services and Delivery Research programme of the National Institute for Health Research.