This paper presents the discovery of a range of high-performance organic non-linear optical (NLO) materials, that arises from ‘smart material’ design and systematic search strategies. This systematization circumvents the previous use of iterative discovery methods, which can only ever afford incremental improvements to currently known NLO materials, and they have no capacity to reveal entirely new classes of suitable NLO materials. This new approach employs data-mining, using the world’s repository of all published organic crystal structures as a representative set of chemical space. Two independent search strategies are implemented, each predicting the best organic NLO materials. The first search method relies on the concept of ‘molecular lego’, taking particular types of molecular fragments that are known to be important constituents of an NLO active material (the ‘lego’), and searching for these through chemical space, with the assistance of graph theory algorithms and systematic enumeration and classification. The second search method uses quantum- mechanical calculations to evaluate the molecular hyperpolarizability, β, of every organic molecule in the aforementioned database. Since β affords the intrinsic measure of NLO output, all organic molecules listed in descending order of values reflects a ranked list of their NLO potential. The NLO properties of selected materials that are highly-ranked in these two lists were then tested experimentally, using Hyper-Rayleigh Scattering (HRS). The predictions are shown to be borne out by such experiments: HRS results show β0 (static hyperpolarizability) values that are up to 10 x greater than those for the industrial reference Disperse Red 1. Due to the commercial potential of these results, four new classes of NLO materials identified by this study have recently been patented.