Background All NHS providers collect data on patient experience, although there is limited evidence about what to measure or how to collect and use data to improve services. We studied inpatient mental health services, as these are important, costly and often unpopular services within which serious incidents occur. Aims To identify which approaches to collecting and using patient experience data are most useful for supporting improvements in inpatient mental health care. Design The study comprised five work packages: a systematic review to identify evidence-based patient experience themes relevant to inpatient mental health care (work package 1); a survey of patient experience leads in NHS mental health trusts in England to describe current approaches to collecting and using patient experience data in inpatient mental health services, and to populate the sampling frame for work package 3 (work package 2); in-depth case studies at sites selected using the work package 2 findings, analysed using a realist approach (work package 3); a consensus conference to agree on recommendations about best practice (work package 4); and health economic modelling to estimate resource requirements and potential benefits arising from the adoption of best practice (work package 5). Using a realist methodology, we analysed and presented our findings using a framework based on four stages of the patient experience data pathway, for which we coined the term CRAICh (collecting and giving, receiving and listening, analysing, and quality improvement and change). The project was supported by a patient and public involvement team that contributed to work package 1 and the development of programme theories (work package 3). Two employed survivor researchers worked on work packages 2, 3 and 4. Setting The study was conducted in 57 NHS providers of inpatient mental health care in England. Participants In work package 2, 47 NHS patient experience leads took part and, in work package 3, 62 service users, 19 carers and 101 NHS staff participated, across six trusts. Forty-four individuals attended the work package 4 consensus conference. Results The patient experience feedback cycle was rarely completed and, even when improvements were implemented, these tended to be environmental rather than cultural. There were few examples of triangulation with patient safety or outcomes data. We identified 18 rules for best practice in collecting and using inpatient mental health experience data, and 154 realist context–mechanism–outcome configurations that underpin and explain these. Limitations The study was cross-sectional in design and we relied on examples of historical service improvement. Our health economic models (in work package 5) were therefore limited in the estimation and modelling of prospective benefits associated with the collection and use of patient experience data. Conclusions Patient experience work is insufficiently embedded in most mental health trusts. More attention to analysis and interpretation of patient experience data is needed, particularly to ways of triangulating these with outcomes and safety data. Future work Further evaluative research is needed to develop and evaluate a locally adapted intervention based on the 18 rules for best practice. Study registration The systematic review (work package 1) is registered as PROSPERO CRD42016033556. Funding This project was funded by the National Institute for Health Research (NIHR) Health Services and Delivery Research programme and will be published in full in Health Services and Delivery Research; Vol. 8, No. 21. See the NIHR Journals Library website for further project information.