Combinatorial estimations show that, within an unreacted ceramic sample prepared by mixing N different starting materials MxOy with average particle size approximately 1 microm, there are about 10(12) grains per cubic centimeter, sufficient for local reactions to occur that may produce a larger number of product oxides than presently accessible by 2D plate techniques. The "single-sample concept" (SSC) is proposed for performing property-directed syntheses for the preparation of ferri-/ferromagnetic or superconducting compounds. Because of the magnetic properties of the products, libraries of product grains can be sorted by means of magnetic separation techniques. For materials with a large magnetization, the separation efficiency is so high that traces of products can be isolated. The SSC concept was tested experimentally to prepare Fe-based oxides (N=17, 24, 30). The large yields (<75 wt %, N=17) of product grains agree with the literature data, which indicate that 3d metal magnetic oxide phases (Tc>300 K) are most probably Fe oxides. In combination with magnetic separation techniques, SSC seems particularly adapted for exploring the solid-state chemistry of metallic lead elements that form ferri-/ferromagnetic or superconducting oxide phases difficult to detect systematically within the large phase space of theoretically existing compounds.