BACKGROUND: Contingency Management (CM) is being piloted as a treatment for stimulant use disorder in several US states, highlighting the need for treatment optimization. One important goal of optimization is decreasing drug use during the early stages of treatment, which has predicted success in other interventions. However, this "critical period" has not been reported in CM trials. The purpose of this analysis was to determine if, after accounting for baseline abstinence and incentive condition, abstinence in a CM trial for people with Cocaine Use Disorder (CUD) could be predicted by cocaine use during a first-week critical period. METHODS: Eighty-seven participants with CUD were randomized to receive contingent high or low value incentives for cocaine abstinence or were in a non-contingent control group. Generalized estimating equations (GEE) were used to analyze urine test results over 36 timepoints during the 12-week intervention. To assess for a critical period, the first three visits were included in the GEE as a covariate for remaining urine test results. RESULTS: Participants who provided more negative samples during the critical period were significantly more likely to produce a negative urine sample during the remainder of the trial, though some effects of group remained after controlling for the critical period. CONCLUSIONS: These results indicate that a critical period exists for CM trials, and it can explain a substantial amount of future performance. Early contact with an abstinence-contingent high magnitude alternative reinforcer may explain additional performance beyond the critical period, further justifying the use of high magnitude alternative reinforcers.