Insulators and conductors with periodic structures can be readily distinguished, because they have different band structures, but the differences between insulators and conductors with random structures are more subtle(1,2). In 1958, Anderson provided a straightforward criterion for distinguishing between random insulators and conductors, based on the 'diffusion' distance zeta for electrons at 0 K ( ref.3). Insulators have a finite zeta, but conductors have an infinite zeta. Aided by a scaling argument, this concept can explain many phenomena in disordered electronic systems, such as the fact that the electrical resistivity of 'dirty' metals always increases as the temperature approaches 0 K (refs 4-6). Further verification for this model has come from experiments that measure how the properties of macroscopic samples vary with changes in temperature, pressure, impurity concentration and applied magnetic field(4,5), but, surprisingly, there have been no attempts to engineer a metal-insulator transition by making the sample size less than or more than zeta. Here, we report such an engineered transition using six different thin-film systems: two are glasses that contain dispersed platinum atoms, and four are single crystals of perovskite that contain minor conducting components. With a sample size comparable to zeta, transitions can be triggered by using an electric field or ultraviolet radiation to tune zeta through the injection and extraction of electrons. It would seem possible to take advantage of this nanometallicity in applications.