Oil recovery factors in cold heavy oil production with sand (CHOPS) are typically lower than 15 %. Solvent-aided processes, such as cyclic solvent injection (CSI) are common post-CHOPS approaches, where wormhole networks could offer increased reservoir contact. However, grid block sizes in field-scale simulations are much larger than the wormhole scale and large-scale dispersivity values are assigned arbitrarily based on history matching. This work implements a statistical scale-up workflow that facilitates the construction of coarse-scale models for CSI simulation, whose relevant parameters are calibrated against simulation results using high-resolution wormhole networks. The formulated workflow can be integrated with commercial reservoir simulators to effectively simulate solvent processes at multiple scales. Multiple injection scenarios are analyzed. Extended soaking periods may positively impact the ultimate recovery with a slower decline at later times, while a lower initial rate is observed. Interestingly, when an economic limit is imposed, the optimal soaking time is not necessarily the longest one. It depends on the trade-off between extracting additional oil recovery at late times versus producing at a higher rate at early times. The analysis also reveals that the initial cycles contribute the most to the final recovery. In addition, when the amount of solvent available is limited, the results would support the strategy of injecting all the solvent in 1 single cycle, with an extended soaking period, rather than performing shorter consecutive cycles.