The growing trend of distributed generation in Low-voltage (LV) networks requires Electricity Distribution Businesses (EDBs) to consider how their networks will perform with this new technology. Of particular interest is the level of distributed generation that can be supported until power quality issues or overloading of assets results, collectively referred to as network congestion. The concept of Distributed Generation (DG) hosting capacity is introduced which defines how much power can be injected per DG system into the network at a selected penetration level before steady-state voltages at the point of supply and/or the current ratings of equipment are likely to be exceeded. An approximate technique called DGHost is described, which requires only a few basic network parameters to accurately estimate DG hosting capacity. It leverages results from full power-flow simulations of over 20 thousand LV networks in New Zealand by using a k-nearest neighbour algorithm to identify a subset of “similar” simulation states within the results database. The impact of phase imbalance is addressed in the model and the tool has been expanded to incorporate network variables such as undersized neutral conductors and single phase transformers - factors identified by the GREEN Grid Network Analysis Group (NAG) as needing separate consideration. Cross-validation techniques were used to optimise the method and to determine its practical estimation accuracy. Hosting capacities estimated by DGHost resulted in a better than two-fold reduction in error compared to alternative methods that rely on the same simplified network data inputs. The DGHost approximation technique provides EDBs with a short to medium-term solution for managing increasing levels of small-scale DG, without the need for complete asset data collection and power-flow modelling, which may be impractical or cost-prohibitive. It is demonstrated how DGHost can be applied to streamlining the application process for small-scale DG such as rooftop solar photovoltaic systems and how it can identify parts of the distribution network vulnerable to congestion.