Abstract By integrating multiple datasets with relevant theory, covering fluid injection and fracturing, a conceptual model has been developed for the fracture development and induced seismicity associated with the fracking in 2011 of the Carboniferous Bowland Shale in the Preese Hall-1 well in Lancashire, NW England. Key features of this model include the steep fault that has been recognized adjoining this well, which slipped in the largest induced earthquakes, and the presence of a weak subhorizontal ‘flat’ within the depth range of the fluid injection, which was ‘opened’ by this injection. Taking account of the geometry of the fault and the orientation of the local stress field, the model predicts that the induced seismicity was concentrated approximately 700 m SSE of the Preese Hall-1 wellhead, in roughly the place where microseismic investigations have established that this activity was located. A further key observation, critical to explaining the subsequent sequence of events, is the recognition that the fluid injection during stage 2 of this fracking took place at a high net pressure, approximately 17 MPa larger than necessary. As a result, the fluid injection ‘opened’ a patch of the ‘flat’, making a hydraulic connection with the fracture network already created during stage 1. Continued fluid injection thus enlarged the latter fracture network, which ultimately extended southwards far enough to intersect the steep part of the fault and induce the largest earthquake of the sequence there. Subsequent fluid injection during fracking stages 3 and 4 added to the complexity of this interconnected fracture network, in part due to the injection during stage 4 being again under high net pressure. This model can account for many aspects of the Preese Hall record, notably how it was possible for the induced fracture network to intersect the seismogenic fault so far from the injection point: the interconnection between fractures meant that the stage 1 fracture continued to grow during stage 2, rather than two separate smaller fractures, isolated from each other, being created. Calculations indicate that, despite the high net pressure, the project only ‘went wrong’ by a narrow margin: had the net pressure been approximately 15 MPa rather than approximately 17 MPa the induced seismicity would not have occurred. The model also predicts that some of the smaller induced earthquakes had tensile or ‘hybrid’ focal mechanisms; this would have been testable had any seismographs been deployed locally to monitor the activity. The analysis emphasizes the undesirability of injecting fracking fluid under high net pressure in this region, where flat patches of fault and/or subhorizontal structural discontinuities are present. Recommendations follow for future ‘best practice’ or regulatory guidelines. Supplementary material: Background information on the stratigraphy, structural geology, rock-mechanical properties of the study region and its state of stress, as well as theory for fluid injection, hydraulic fracturing and Coulomb failure analysis, is available at https://doi.org/10.6084/m9.figshare.c.3781121