It has been estimated that between 25% and 40% of people living with dementia suffer from sleep disturbances, and there are currently no known effective treatments. Sleep disturbances may be the direct result of dementia or due to other comorbidities, such as pain and limited mobility. If carers' sleep is also disturbed, carers too can become tired and stressed, and this sometimes results in the breakdown of care in the home. To design an evidence-based manualised non-pharmacological therapy for sleep disturbances and test it for feasibility and acceptability. A single-blind, randomised, parallel-group feasibility trial, with participants randomised 2 : 1 to intervention or treatment as usual (TAU). Five memory services in two London NHS trusts and Join Dementia Research (JDR). The study recruited people with dementia and sleep disturbances (who scored ≥ 4 on at least one question on the Sleep Disorders Inventory) and their primary family carers. All participants were given an Actiwatch (CamNtech Ltd, Cambridge, UK) to wear to record their sleep patterns for 2 weeks before randomisation. The intervention group received Dementia RElAted Manual for Sleep; STrAtegies for RelaTives (DREAMS START). This was designed as a six-session, manual-based intervention for carers of people with dementia, delivered by trained and clinically supervised psychology graduates, based on evidence about managing sleep disturbance in people with dementia. It uses the structure of a previous manual-based treatment, STrAtegies for RelaTives (START). Family carers were consulted about structure, content and design. Sessions were interactive, and each involved techniques, tasks to practise between sessions, relaxation and a recapitulation on the previous session. The sessions covered understanding sleep and dementia, making a plan (incorporating information from Actiwatch read-outs and a light box to increase light), daytime activity and routine, difficult night-time behaviours, taking care of your own (carer's) sleep and using the strategies in the future. Carers kept their own manual, light box and relaxation recordings post intervention. A statistician created an electronic randomisation list, stratified by site, using random permuted blocks. Those assessing the outcome were blinded to allocation; participants were not blinded. Outcomes were assessed at 3 months. (1) Feasibility, defined as the percentage of eligible people who consented to the study recruitment, with an expected value of 50% [95% confidence interval (CI) 41% to 59%]. (2) Acceptability, defined as the percentage of intervention group participants attending ≥ 4 intervention sessions, with an expected value of 75% (95% CI 59% to 87%). The predetermined criterion for progression to the main trial was acceptability of ≥ 70%. Of 95 eligible patients referred, 63 (66%, 95% CI 56% to 76%) consented between 4 August 2016 and 24 March 2017: 61 from memory clinics and two from JDR. Of these, 62 participants (65%, 95% CI 55% to 75%) were randomised: 42 to the intervention arm and 20 to the TAU arm. Thirty-seven out of 42 participants (88%, 95% CI 75% to 96%) adhered to the intervention. The results show that the randomised controlled trial is feasible and that the intervention is acceptable. A higher than expected proportion of eligible patients referred consented to the study and adhered to the intervention. Participants were not blinded and were recruited only in London. The results of this trial indicate that a future efficacy trial is warranted. Current Controlled Trials ISCTRN36983298. This project was funded by the NIHR Health Technology Assessment programme and will be published in full in Health Technology Assessment; Vol. 22, No. 71. See the NIHR Journals Library website for further project information. Funding was also provided by Camden and Islington NHS Foundation Trust and Barnet, Enfield and Haringey Mental Health NHS Trust to pay for excess treatment costs from therapist training and supervision and intervention delivery.