Fibre fragment pollution from textiles has become a significant environmental issue, with fibre fragments now ubiquitous in nature and growing concerns over their potential harm to ecosystems and human health. One proposed solution to mitigate this issue is to modify textile manufacturing processes to create fabrics that shed fewer fibres. To support the textile industry in adopting best practices for creating lower-shedding textiles, this research investigates the influence of fabric characteristics on fibre fragmentation. Using data from The Microfibre Data Portal, which includes over a thousand fabrics tested by 95 textile industry signatories from across the textile industry, this study provides valuable insights into how various factors contribute to fibre fragmentation. A machine learning approach was used to assess how fabric composition, structure, yarn type, and finishing influence shedding. Finishing techniques emerged as a major factor due to their strong effect on fragmentation and their potential for easier modification during production, despite being mostly overlooked in existing research. However, the findings underscore the need to consider all fabric characteristics in efforts to reduce fibre fragmentation. As well as providing valuable insights for the textile industry when addressing fibre fragmentation, the machine learning model generated provides future potential for the prediction of fibre fragmentation for fabrics on known specifications, reducing the need for physical testing.