Today’s retail environment is characterized by an increasingly complex and interrelated set of channels across which consumers move freely. These movements include visits to one channel and sales in another, referred to as research shopping, as well as post-purchase experience effects that create feedback within the omni-channel system. As a result, channels become causally interrelated, depending endogenously on each other. The emerging omni-channel system may exhibit non-linear and state-dependent behavior rather than linear behavior in equilibrium. To assess these channel interrelations based on the movements of consumers, often only aggregate time series data are available (e.g., online visits per time period). The authors introduce empirical dynamic models (EDM), a nonlinear methodology used in research on biological eco-systems, to capture different types of channel interrelations and non-linear behaviors. In particular, EDMs allow for empirically testing all pairs of channel time series variables for an uni-directional or bi-directional, possibly non-linear relationship within a common omni-channel system. The resulting interrelation network can be exploited to assess the state-dependent and interacting within-channel and cross-channel effects. EDM are applied to examine daily visits and sales time series data from a three-channel system (brick-and-mortar, online store, and mobile store) of a fashion retailer. We find that not all possible relationships between channel variables are relevant. Specifically, the online and mobile visits remain independent of each other. Furthermore, not all channels interact synergistically such that the retailer should focus on increasing the number of visits in mobile and brick-and-mortar stores, but not the online store.