We revisit the performance and biases of the two-dimensional approach to galaxy–galaxy lensing. This method exploits the information for the actual positions and ellipticities of source galaxies, rather than using only the ensemble properties of statistically equivalent samples. We compare the performance of this method with the traditionally used one-dimensional tangential shear signal on a set of mock data that resemble the current state-of-the-art weak lensing surveys. We find that under idealised circumstances the confidence regions of joint constraints for the amplitude and scale parameters of the NFW model in the two-dimensional analysis can be more than three times tighter than the one-dimensional results. Moreover, this improvement depends on the lens number density and it is larger for higher densities. We compare the method against the results from the hydrodynamical EAGLE simulation in order to test for possible biases that might arise due to lens galaxies being missed, and find that the method is able to return unbiased estimates of halo masses when compared to the true properties of the EAGLE galaxies. Because of its advantage in high galaxy density areas, the method is especially suitable for studying the properties of satellite galaxies in clusters of galaxies.