Objective: Understanding how treatments work is a goal of psychotherapy research, however the strength of relationships between therapy processes and outcomes is inconsistent. DeRubeis, Cohen, et al. (2014) proposed that process-outcome relationships are moderated by patient characteristics. These “patient response patterns” (PRPs) indicate individuals’ responsiveness to the active ingredients of treatment. Given the same quality of therapy, one individual may receive more benefit than another depending on their PRP. The “prognosis moderation hypothesis” states that PRPs can be defined by pretreatment prognostic indicators. Medium prognosis groups (“pliant-like”) will have stronger process-outcome relationships than good (“easy-like”) or poor (“challenging-like”) groups. Method: N = 190 individuals received unguided computerized CBT. They were 58% women, aged 44.7 years. Engagement with the cCBT program was the process variable. PRPs were defined by predicted scores from a prognostic regression model. Outcomes were BDI scores at 3, 6, and 12 months. “Easy-like,” “pliant-like” and “challenging-like” groups were created and the engagement-outcome relationship was assessed as a function of group. Results: Engagement-outcome correlations by PRP were: easy-like, r = −.27 (p < .05); pliant-like, r = −.36 (p < .01); and challenging-like, r = .05 (p = .70). The pliant-like group was found to be the only moderator of the engagement-outcome relationship. Results were similar at 6 months but faded at 12. Conclusions: The engagement-outcome relationship varied as a function of prognosis, providing support for the prognosis moderation hypothesis. The “pliant-like” group appeared most sensitive to treatment procedures. Future research is needed to refine the methods for identifying PRPs.