Mobile apps are often developed and then evolved on more than one mobile operating system. For the publisher of such apps, the problem is how to ensure equivalence of the product on the various platforms, in the sense of acting equivalently with respect to a test set. In this paper we present an approach that tackles this problem from two directions: architecture and testing. First we will explain the role and pitfalls of reference architectures. Then we will present our equivalence testing framework. It is based on mocking components of the target implementation. Instead of implementing mock components for all platforms, our approach is to implement these mocks only once and run them on a central server. On the tested target device, stub components are injected that forward to their server-based counterparts. Thus the exactly same test code is applied to all platform specific implementations.