We propose a tree-structured heterogeneous autoregressive (tree-HAR) process as a simple and parsimonious model for the estimation and prediction of tick-by-tick realized correlations. The model can account for different time and other relevant predictors ’ dependent regime shifts in the conditional mean dynamics of the realized correlation series. Testing the model on S&P 500 and 30-year treasury bond futures realized correlations, we provide empirical evidence that the tree-HAR model reaches a good compromise between simplicity and flexibility, and yields accurate single- and multi-step out-of-sample forecasts. Such forecasts are also better then those obtained from other standard approaches, in particular when the final goal is multi-period forecasting.