Development of underground transportation systems often involve twin tunnels, which may encounter existing pile groups during construction. Since many previous studies mainly focus on the effects of single tunnelling on single piles, settlement and load transfer mechanism of a pile group subjected to twin tunnelling are not well investigated and understood. To address these two issues, two three-dimensional centrifuge tests were carried out in this study to simulate side-by-side twin tunnels (excavated one after the other on both sides of the pile group) at two critical locations relative to the pile group, namely next to (Test TT) and below the toe of the pile group (Test BB). Moreover, numerical back-analyses of the centrifuge tests are conducted by using a hypoplastic model, which takes small-strain stiffness into account. Both measured and computed results show that the induced tilting of the pile group in Test TT is significantly larger than that in Test BB, with a maximum percentage difference of 120%. On the other hand, a slightly smaller (about 13%) settlement of the pile group is induced in Test TT, as compared to that in Test BB. This is because the pile group in Test TT is partially located within the major influence zone of tunnelling-induced ground settlement while the entire pile group in Test BB is bounded by the major influence zone of ground settlement. Two distinct load transfer mechanisms due to twin tunnelling are identified, i.e., the load in the pile group in Test TT transfers downwards from the pile shaft to the pile toe while the load in the pile group in Test BB transfers upwards from the pile toe to the pile shaft. Apart from load transfer along each pile, load re-distribution also occurs among piles during twin tunnelling. In both Tests TT and BB, axial load at pile head only reduces at a pile closet to the advancing tunnel face and the reduction is re-distributed to the other three piles. The load re-distribution among piles results in a maximum increase of axial force of 10% in Test TT.