Fecal microbiota transplantation (FMT) is a promising treatment for microbiota dysbiosis associated diseases, such as infection (CDI) and inflammatory bowel disease (IBD). The engraftment of donor bacteria is essential for the effectiveness of FMT, which to some extent depends on the matching of donors and recipients. However, how different types of donor-derived bacteria affect FMT efficacy has not been fully dissected. We recruited two longitudinal IBD cohorts of 103 FMT recipients and further analyzed 1,280 microbiota datasets from 14 public CDI and IBD studies to uncover the effect of donor-derived microbiota in recipients. We found that two enterotypes, RCPT/E and RCPT/B (dominated by Enterobacteriaceae and , respectively), consistently exist in both CDI and IBD patients. Based on a time-course-based multi-cohort analysis of FMT fecal samples, we observed the interplay between recipient and donor-derived microbiota during FMT, in which the FMT outcome was significantly associated with the enterotype and microbiota distance between donor and recipient after FMT. We proposed a new measurement, the ratio of colonizers to residents after FMT (C2R), to quantify the engraftment of donor-derived bacteria in the recipients, and then constructed an enterotype-based statistical model for donor-recipient matching, which was validated by both cross-validation and an additional IBD FMT cohort (n = 42). We believe that with the accumulation of FMT multi-omics datasets, machine learning-based methods will be helpful for rational donor selection for improving efficacy and precision FMT practices.