Monolithic multi-junction solar cells are becoming prevalent in concentrator photovoltaic (CPV) systems due to their high demonstrated conversion efficiencies. However, these devices often operate at sub-optimal levels due to current mismatch losses arising between each p-n junction as a result of natural variation in the solar spectrum at the earth's surface. The use of quantum wells in solar cell design affords improved control over the spectral response of the cells via band-gap engineering. This makes it possible to tailor the spectral response of a cell for optimum performance under a given annual spectral resource. Establishing the optimum spectral response for a given location is a major challenge of this approach to cell design, and relies heavily on averaged and modelled data based on combinations of satellite and sparse ground measurements. In addition, there is only limited field experience of such technology to date. The purpose of this work is to investigate through experiment the effect of incorporating multi-quantum-well (MQW) structures into photovoltaic cells to respond to a specific range of annual irradiation spectra and to compare the results obtained with those predicted through modelling. A number of triple-junction cells of different design have been placed side-by-side on an accurate solar tracker, and current-voltage characteristics of each taken at regular intervals over a few months. The outputs are then compared to those predicted by a model of cell performance that includes simulated spectra, generated with the SMARTS program, specific to that location and period of time. Results obtained from outdoor testing indicate that the cells designed to provide improved current matching under the spectral conditions in which they have been tested have performed more consistently than conventional cell designs. This is evident when the short-circuit current is normalised to a fixed direct normal irradiation and temperature, and plotted against a- mospheric depth. Detailed analysis of the spectral resource is expected to reveal additional information on the performance of the cells.