Thermal response functions are commonly used to describe the heat exchange capability of ground heat exchangers. When convolved with the ground thermal load, they allow for a quick and efficient prediction of the temperature evolution. Typically, thermal response functions are obtained through physical models of the heat transfer process. However, they can also be reconstructed directly from real experimental data, such as temperature measurements of the fluid leaving the boreholes. Here we reconstruct experimental response functions from monitoring data of 40 borehole heat exchangers. Each borehole is connected through a horizontal pipe to one of three underground vaults, in which sensors for fluid temperature and flow rate are installed. A data period of 27 days is used to deconvolve response functions for each borehole individually. The resulting functions reproduce the measured data with a root mean square error of less than 0.06 K. Moreover, they reveal subtle variations in the heat exchange potential of each individual borehole. Statistical analysis attributes those differences mainly to the length of the horizontal connecting pipes and the distance to neighboring boreholes. However, discrepancies for some boreholes suggest that not all variables affecting the heat transfer were taken into account.