This paper presents the results of an experimental evaluation of a change in assessment practice in one comprehensive secondary school. The school divided 104 Year 7 pupils into four mixed-ability teaching groups. One of these was given enhanced formative feedback on their work for one year, but no marks or grades. The other three groups were given marks and grades with minimal comments, which was the usual prior practice in this school (and many others). Using data derived from assessment, prior attainment, pupil attitudes and background information, we conducted a contextualized analysis of progress in the four teaching groups for all subjects. This showed that progress in the treatment group (formative feedback only) was substantially inferior to that of the other three groups. In this paper, we also use data from observation of the process and from group interviews with the students involved, to help explain these results. Our findings are relevant to a consideration of the often lessened impact of research findings when 'rolled' out into wider practice, and what may be done about this.