We analyzed the behavior of convergent flow tracer tests performed in a 3‐D heterogeneous sandbox in presence of connected gravel channels under laboratory‐controlled conditions. We focused on the evaluation of connectivity metrics based on characteristic times calculated from experimental breakthrough curves (BTCs), and the selection of upscaling model parameters related to connectivity. A conservative compound was injected from several piezometers in the box, and depth‐integrated BTCs were measured at the central pumping well. Results show that transport was largely affected by the presence of gravel channels, which generate anomalous transport behavior such as BTC tailing and double peaks. Connectivity indicators based on BTC peak times provided better information about the presence of connected gravel channels in the box. One of these indicators, β, was defined as the relative temporal separation of the BTCs peaks from the BTCs centers of mass. The mathematical equivalence between β and the capacity coefficient adopted in mass transfer‐based formulations suggests how connectivity metrics could be directly embedded in mass transfer formulations. This finding is in line with previous theoretical studies and was corroborated by reproducing a few representative experimental BTCs using a 1‐D semianalytical bimodal solution embedding a mass transfer term. Model results show a good agreement with experimental BTCs when the capacity coefficient was constrained by measured β. Models that do not embed adequate connectivity metrics or do not adequately reproduce connectivity showed poor matching with observed BTCs.