There are concerns that non-anatomical resection (NAR) worsens perioperative and oncological outcomes compared with those following anatomical resection (AR) for colorectal liver metastases (CRLM). Most previous studies have been biased by the effect of tumour size. The aim of this study was to compare oncological outcomes after NAR versus AR.This was a retrospective study of consecutive patients who underwent CRLM resection with curative intent from 1999 to 2016. Data were retrieved from a prospectively developed database. Survival and perioperative outcomes for NAR and AR were compared using propensity score analyses.Some 358 patients were included in the study. Median follow-up was 34 (i.q.r. 16-68) months. NAR was associated with significantly less morbidity compared with AR (31·1 versus 44·4 per cent respectively; P = 0·037). Larger (hazard ratio (HR) for lesions 5 cm or greater 1·81, 95 per cent c.i. 1·13 to 2·90; P = 0·035) or multiple (HR 1·48, 1·03 to 2·12; P = 0·035) metastases were associated with poor overall survival (OS). Synchronous (HR 1·33, 1·01 to 1·77; P = 0·045) and multiple (HR 1·51, 1·14 to 2·00; P = 0·004) liver metastases, major complications after liver resection (HR 1·49, 1·05 to 2·11; P = 0·026) or complications after resection of the primary colorectal tumour (HR 1·51, 1·01 to 2·26; P = 0·045) were associated with poor disease-free survival (DFS). AR was prognostic for poor OS only in tumours smaller than 30 mm, and R1 margin status was not prognostic for either OS or DFS. NAR was associated with a higher rate of salvage resection than AR following intrahepatic recurrence.NAR has at least equivalent oncological outcomes to AR while proving to be safer. NAR should therefore be the primary surgical approach to CRLM, especially for lesions smaller than 30 mm.