Animal welfare is usually excluded from life cycle assessments (LCAs) of farming systems because of limited consensus on how to measure it. Here, we constructed several LCA-compatible animal-welfare metrics and applied them to data we collected from 74 diverse breed-to-finish systems responsible for 5% of UK pig production. Some aspects of metric construction will always be subjective, such as how different aspects of welfare are aggregated, and what determines poor versus good welfare. We tested the sensitivity of individual farm rankings, and rankings of those same farms grouped by label type (memberships of quality-assurance schemes or product labelling), to a broad range of approaches to metric construction. We found farms with the same label types clustered together in rankings regardless of metric choice, and there was broad agreement across metrics on the rankings of individual farms. We found woodland and Organic systems typically perform better than those with no labelling and Red tractor labelling, and that outdoor-bred and outdoor-finished systems perform better than indoor-bred and slatted-finished systems, respectively. We conclude that if our goal is to identify relatively better and worse farming systems for animal welfare, exactly how LCA welfare metrics are constructed may be less important than commonly perceived.