In martensitic phase transformation, the formation of microstructure is due to minimization of elastic plus surface energy. There is by now a lot of work on macroscopically-uniform microstructures. In the presence of stress or a thermal gradient, however, one expects microstructures whose volume fractions and other features vary macroscopically. I'll discuss some model problems of this kind, involving non-uniform twinning of two martensite variants. In these problems the loads or boundary conditions require the volume fractions to vary; this forces the twin boundaries to tilt away from their stress-free orientations, and leads to what might be called a "zig-zag microstructure." The goals for mathematical analysis are (a) to show that such a microstructure achieves the optimal energy scaling law, (b) to explore the local properties of energy-minimizing patterns, and (c) to understand the robust periodicity observed in experimental realizations of such microstructures. I'll discuss some recent progress toward (a) and (b) (joint work with Alex Misiats and Stefan Mueller).