There is lots of discussion on shear banding in the literature but for different materials and systems. Clearly the same mechanisms can't explain banding in colloidal suspensions and also in polymer solutions.
Although I'm not expert in shear banding I can give a few hints: 1.wide gaps 2. increased packing or concentration 3. unsheared material. All these parameters either in solid/liquid systems, or in gels (certain polymers and/or biopolymers) are crucial to observe shear banding (at least, based on my experience)
While shear banding is often indeed associated with an intrinsically unstable flow curve (minimum in shearstress vs shearrate, typically resulting in a plateau in the measured flow curve), this is not a necessary condition for SB to occur. Examples e.g. in concentrated colloids and emulsions show that a coupling between shear and concentration ('dilation') or collective effects during yielding of such concentrated disordered systems can also lead to shear banding (well known for granular systems). In these cases no minimum or plateau in the stress vs strain curve is required. See for example attached paper 'Shear banding and flow-concentration coupling in colloidal glasses. R Besseling et al Physical Review Letters 12/2010; 105(26):268301)
I would like to add something to what is written by Stéphane Costeux. Favorable circumstance for plastic deformation to be localized in narrow bands is the presence of two branches of the nonlinear material behavior. One branch corresponds to active loading of the material (with relatively small plastic module), while the other branch - to unloading (with relatively large elastic module). However, in any case, the deformation of the material should take place under constrained conditions. Otherwise, the instability can be distributed, rather than to occur in the localized manner.
Many of the above answers are correct but apply only to one type of material, e.g. fluid vs. solid. Generally, shear-banding is a reflection of a "soft mode" instability. It can happen because viscous stress decreases with shear rate in steady shear in a fluid, or because elastic stress decreases with strain in an elastic sold, or because plastic stress decreases with plastic strain in a plastic solid such as a granular soil. In the case of an elastic solid, it is the analog of a phase change, e.g. of the Van der Waals model of a fluid which has three coexistent states of pressure vs. density, with density being the analog of strain.