Faecal microbiota transplantation (FMT) has proved to be an effective treatment strategy for recurrent infections with Clostridium difficile. The success of FMT requires donor bacteria to engraft in the gut of the patient, but the factors that promote engraftment of individual strains has remained elusive. Alm and colleagues used high-resolution deep shotgun metagenomics sequencing to profile the faecal microbiota of patients with recurrent C. difficile after FMT. They also developed Strain Finder, which infers the genotypes and frequencies of strains in complex metagenomics samples, and combined it with machine learning to quantitatively model bacterial engraftment in humans. They report that the most important factors in their model are bacterial abundances, bacterial taxonomy and the amount of elapsed time since the FMT. They validated their findings for metabolic syndrome, which suggests that the models of engraftment apply to other conditions.