When developing a complex system, it can be impractical and prohibitively expensive to apply the ultimate degree of rigour to safety assurance of every part of the design. Some way of measuring the relative importance of different artefacts is needed, to allow effort to be applied proportionately. The forthcoming system safety standard, IEC 63187, defines the concept of a ‘measure of importance scheme’ (MoI scheme) to achieve this aim. IEC 63187 is being drafted to supplement systems engineering processes with requirements to make them appropriate for complex safety-critical systems such as those found in defence. However, IEC 63187 applies to the abstract domain of systems engineering and does not tell one how to actually realise concrete system elements using technologies like software, programmable electronics, machine learning or traditional mechanical engineering. Understanding how to move from the abstract to the concrete, considering the usage scenarios in the operating environment, will be vital to putting the new standard into practice. This paper reviews what needs to be taken into account when transforming from an abstract MoI to the importance metrics used in existing realisation standards and presents an example of how this might be achieved for a ma-chine learning-based system element, using ED-324/ARP6983 (EUROCAE and SAE 2025) to support IEC 63187.