A sketch map is a simple conceptual model, where complexity is reduced to the essential details. There is no need to label all the streets. However, the map is useless if a key signpost is wrong. Groundwater modelling is analogous. We simplify the complexity to the essential details for reliable decision making. Yet, what if the signposts are wrong? Groundwater is driven by rainfall on the land surface, which also drives hydrology. Runoff, streamflow, ET and recharge are the signposts. In groundwater models, these processes are simplified to satisfy boundary constraints. However, recharge is spatially and temporally complex, streamflow is highly dynamic and non-linear, and feedback is substantial. The alternative is to simulate groundwater as part of the hydrology using dynamic, distributed, physics based methods. Such models include overland flow, river hydraulics, unsaturated infiltration, ET and groundwater flow. An integrated model internally partitions rainfall into runoff, infiltration and evapotranspiration – avoiding the double accounting common with separate models. Groundwater-surface water interaction changes dynamically with changing stream levels and flooded areas. Riparian vegetation competes for groundwater discharge to streams. A primary outcome is a full, distributed, dynamic water balance – not just a “groundwater water balance” of the 5 % of rainfall left over after ET. Integrated models are useful when surface processes affect sub-surface conditions – and vice versa. The most common objections to integrated models relate to high data requirements and excess model complexity. Typically, these objections are a strawman for a lack of knowledge and expertise. In reality, most modellers are specialized in either groundwater or surface water - rarely with expertise in both. The real issues affecting model reliability (e.g. a poorly defined recharge boundary) are ignored and effort is spent increasing the fidelity of the model domain. The streets are labelled, but the signposts are wrong.