Many unconventional therapies (e.g. dietary, phytotherapy, acupuncture, homeopathy) are well known and often applied, but their efficacy has hardly been proven. New trial designs and study components must be found to meet the specific demands of the particular unconventional therapy on one hand and keep the high methodological standard of controlled clinical trials on the other hand. Biometricians and unconventional therapists are challenged to develop such designs. Typical problems in designing studies of unconventional therapies include that placebo is not possible, therapies cannot be masked, outcome variables are not reliable, therapy is highly individualized, and studies on the efficacy of soft therapies require many patients and long treatment periods. Studies with unconventional therapies should be performed by practitioners (because they use these therapies), but this leads to further problems. Some solutions are given in examples: A study is described investigating the herbal remedy Kava-Kava for patients in the state of anxiety, tension and restlessness; a study on classical homeopathy for chronical headaches is specified; some designs for dietary studies in patients with rheumatoid arthritis are compared. A design called "cross-allocation of patients to two treatments with randomization option" and the "N-of-1 design", also called "single case design" are described and discussed. The "change-to-open-label design" could be useful to investigate soft and natural therapies which require studies with many patients and long-term treatment.