Background: MRD assessment is a known surrogate marker for survival in multiple myeloma (MM). Most data come from patients enrolled in clinical trials. We present a single institution’s experience assessing MRD in patients receiving frontline therapy and therapy for relapsed disease. We describe the impact of depth, duration, and direction of response on prognosis. Methods: 181 MM patients at University of California, San Francisco (UCSF) from 2008 to 2016. 126 were newly diagnosed and 55 in ≥ 2nd line. MRD was assessed in patients achieving VGPR or better by IMWG criteria. MRD assessment was performed by NGS (Adaptive Biotechnologies, Seattle, WA). PFS curves were plotted by the Kaplan-Meier method, and the log-rank test was used to estimate statistical significance. Results: 398 MRD samples were analyzed. MRD was available at 3 time points for 59 patients and 2 time points for 36 patients. Median follow up was 26m. Overall, 66 of 181 patients (36%)achieved MRD- ( < 10^-6) on one or more samples. In the newly diagnosed group, 43 of 126 (34%), achieved MRD- at least once. These patients had a prolonged PFS versus patients who were persistently MRD+ (NR vs 49m, p = 0.006). Of the 55 patients who received therapy for relapsed disease, 21 achieved MRD- (38%) and PFS was also prolonged versus patients who remained MRD+ (53m vs 23m, p = 0.03). We analyzed the effects of depth of response. Patients who were MRD- or who were MRD+, at a very low level (between 10^-5and 10^-6), had a better prognosis than those with higher disease burdens ( > 10^-5) (p = 0.001). Finally, we analyzed the effect of repeated MRD monitoring on PFS. Three categories were identified in newly diagnosed patients: (A) patients with ≥3 MRD- samples, (B) patients with continuously declining detectable clones, and (C) patients with a stable number of clones. Groups A and B had a more prolonged PFS than group C (NR vs 31m, p < 0.0001). Conclusions: MRD assessment in a real-world setting has the same predictive power as that seen in clinical trials. MRD dynamics can accurately predict disease evolution and drive clinical decision-making. This study lends support to the concept of MRD-driven decision-making and helps validate the relevance of MRD.