<p>Children’s number-based procedures for mental computation have been widely studied but few studies have focused on children’s flexibility with mental computation procedures. In this study, we investigated children’s capacity for procedural flexibility with mental addition by developing and testing a novel assessment that required them to recognize an efficient strategy used by another (fictitious) child. Testing was conducted in eight metropolitan primary schools in Australia and involved administering the assessment to children in Years 3 and 4 (<i>N</i> = 853). The final version of the assessment comprised 23 items and produced data that were closely aligned to the Rasch model, reflected a unidimensional construct and showed good precision and excellent reliability. Findings from item analyses confirmed predictions about the problem features that evoke thinking about particular strategies and the comparative difficulty of strategy types. These findings contribute insights into how procedural flexibility may be operationalised, making it possible to compare instructional approaches for promoting procedural flexibility on a large scale.</p>