As early as 2001, the need for the 'functional motif theory' was pointed out, to assist the rational design of functional materials. The properties of materials are determined by their functional motifs and how they are arranged in the materials. Uncovering functional motifs and their arrangements is crucial in understanding the properties of materials and rationally designing new materials of desired properties. The functional motifs of materials are the critical microstructural units (e.g. constituent components and building blocks) that play a decisive role in generating certain material functions, and can not be replaced with other structural units without the loss, or significant suppression, of relevant functions. The role of functional motifs and their arrangement in materials, with representative examples, is presented. The microscopic structures of these examples can be classified into six types on a length scale smaller than ∼10 nm with maximum subatomic resolution, i.e. crystal, magnetic, aperiodic, defect, local and electronic structures. Functional motif analysis can be employed in the function-oriented design of materials, as elucidated by taking infrared non-linear optical materials as an example. Machine learning is more efficient in predicting material properties and screening materials with high efficiency than high-throughput experimentation and high-throughput calculations. In order to extract functional motifs and find their quantitative relationships, the development of sufficiently reliable databases for material structures and properties is imperative.