Objectives The United Kingdom Frozen Shoulder Trial (UK FROST) compares stand-alone physiotherapy and two operative procedures, both with post operative rehabilitation, for primary frozen shoulder in secondary care. We developed physiotherapy protocols for UK FROST, incorporating best evidence but recognizing uncertainty and allowing flexibility. Methods We screened a UK Department of Health systematic review and UK evidence-based guidelines (Hanchard et al., 2012; Maund et al., 2012) for recommendations, and previous surveys of UK physiotherapists (Hanchard et al., 2011, 2013) for strong consensus. We conducted a two-stage, questionnaire-based, modified Delphi survey of shoulder specialist physiotherapists in the UK National Health Service. This required positive, negative or neutral ratings of possible interventions in four clinical contexts (stand-alone physiotherapy for, respectively, predominantly painful and predominantly stiff frozen shoulder; and post operative physiotherapy for, respectively, predominantly painful and predominantly stiff frozen shoulder). We proposed respectively mandating or recommending interventions with 100% and 90% positive consensus, and respectively disallowing or discouraging interventions with 90% and 80% negative consensus. Other interventions would be optional. Results The systematic review and guideline recommended including steroid injection and manual mobilizations in non-operative care, and we mandated these for stand-alone physiotherapy. Consensus in the pre-existing surveys strongly favoured advice, education and home exercises, which we mandated across contexts. The Delphi survey led to recommendation of some supervised exercise modalities, plus the disallowing or discouragement-in various contexts-of immobilization and some 'higher-tech' electrotherapies and alternative therapies. Conclusions We developed physiotherapy protocols despite incomplete empirical evidence. Their clear structure enabled implementation in data-sheets designed to facilitate recording, monitoring of fidelity and reporting of interventions. Other trials involving physiotherapy may benefit from this approach.