Life Cycle Assessment (LCA) is increasingly used as an early-stage design-decision tool to support choices of structural system. However LCA modellers must first make numerous methodological decisions, and the resultant wide variations in approach are often inadequately described by the modellers. This paper identifies, and quantifies, the three major areas of methodological variation. These are: temporal differences in the stages considered; spatial differences in the material boundaries; and physical disparities in the data coefficients. The effects are then demonstrated through a case study of a student residential building in Cambridge. The cross-laminated timber (CLT) structure is compared with concrete frame, steel frame and load-bearing masonry, considering the influence that varying the temporal boundaries, the data coefficients, and the spatial boundaries has on the choice. While for this building CLT is confirmed as the lowest impact material, the paper demonstrates that varying the methodological choices can change the results by an alarming factor of 10 or even more. The findings confirm the need for the utmost clarity and transparency with all LCA calculations. Making wider industry or policy decisions based on LCA results should be undertaken with extreme caution. (C) 2018 Elsevier B.V. All rights reserved.