Objective The objective of the study was to evaluate the cost‐effectiveness of implementing the Grip on Challenging Behaviour care programme (GRIP) on dementia special care units in comparison with usual care. Methods A stepped wedge design was used. Challenging behaviour and quality of life were measured using the Cohen Mansfield Agitation Inventory (CMAI) and the QUALIDEM. Quality‐adjusted life years (QALYs) were calculated using the EuroQol‐5D. Psychoactive medication use (range 0–5 per measurement) and sick leave were registered. Costs included medication, time spent on challenging behaviour and education. Costs and effects were analysed using linear multilevel regression. Incremental cost‐effectiveness ratios were calculated. Statistical uncertainty was estimated using bootstrapping. Results Seventeen dementia special care units participated. GRIP led to improvement on the QUALIDEM subscale social relations (1.6; 95% CI 0.18 to 3.4) and on the use of psychoactive medication (−0.73; 95% CI −1.1 to −0.46) and to a decrease in QALYs (−0.02; 95% CI −0.06 to −0.003). No significant effects on CMAI, sick leave and other QUALIDEM subscales were found. The intervention was not cost‐effective in comparison with usual care with regard to CMAI score, QALYs and sick leave. The willingness to pay should be 320€/point improvement on the QUALIDEM subscale social relations and 370€/psychoactive medication less to reach a 0.95 probability of cost‐effectiveness. Conclusion It depends on how much society is willing to pay whether GRIP can be considered cost‐effective. Because the appropriateness of the current methods for analysing cost‐effectiveness in this specific population is uncertain, the positive effects on behaviour, medication and job satisfactions should also be taken in account in the decision making. Copyright © 2015 John Wiley & Sons, Ltd.