We conduct a multi-local, multivariate analysis of be like in comparable datasets from three discontinuous geographic settings (the U.S.A., England, and New Zealand). Previously, comparative cross-variety analysis of this form has been fundamentally hampered by key methodological differences. A methodologically coherent analysis reveals that the 'classic factors' (Tagliamonte and D'Arcy 2007: 203) grammatical person and content of the quote, as well as the effect of mimesis, are transferred to the receptor variety, albeit with varying degrees of completeness. Other conditioning factors are particularized to the local system into which be like is adopted, which leads us to define its spread as a case of weak transfer. We suggest that there are at least two explanations for this finding: (1) global innovations must be considered in light of the local systems into which they are adopted; and (2) the form and amount of contact must be correlated with respect to the knowledge transfer they allow.