This article deals with a question in hub-and-spoke network design: ar
e some nodes better off sending their interactions through a local min
i-hub, rather than through a single central facility, and where should
the hubs be located if this form of allocation is allowed? While the
allocations are obtainable by a simple calculation for fixed facility
locations, it turns out that additional model formulation decisions ha
ve to be made to best capture the reality of the way that these intera
ctions are allowed to be routed through the network. In particular, th
e analyst must determine whether the system being represented (a) allo
ws interactions to make more than one stop; (b) requires the interacti
ng nodes to deal with each other through an exclusive mini-hub ''club'
'; or (c) permits each interacting pair to select its own least cost r
oute (consistent with the number of allowed stop overs). In order to r
eflect the ''reality'' of time-windows, the research finds that it is
necessary to constrain the maximum distance of the service region for
the mini-hub. This is accomplished by adding a simple covering constra
int to the location model. Then, given the model design choice, iris r
elatively easy to formulate and serve appropriate location/allocation
models for the problem using integer programming. The article is illus
trated with examples of these formulations, and sample solutions are p
rovided for a well-studied test data set.