A scratch test in the intermediate load range between the nanoscratch test and the microscratch test is used to determine the failure properties of thin coatings and their interface with the substrate on which the coating was deposited. The technique consists in drawing a stylus with a rounded conical tip along a coated material, while the stylus is simultaneously lowered into the material. A new piece of equipment is introduced for these experiments, that allows simultaneous measurement of forces and displacements in three orthogonal directions, with respectively micro-Newton and nanometre scale resolutions. It is shown that this equipment restrains the degrees of freedom of the moving system and the indenter, to suppress all moments in the system, thereby minimising the uncertainty of the measurement. A full description of the measurement system is given, together with its control electronics and calibration details. Among the many materials systems tested, three systems showed several interfacial failures in the load range considered, and are examined in detail. A simple model of the scratch test is used to derive the work of adhesion of these coatings, and it is further shown that: the first failure event is a combination of the coating and the interface failures, where the interface certainly fails in mixed-mode; and all subsequent failure events being characterised by similar values of the work of adhesion, they are attributed to a mode I failure of the interface. Works of adhesion of 3·06±0·41 J m−2 for the ZnO–Ag/glass system, and 2·36±0·12 J m−2 for the TiB2–Hf/steel system are determined. These values are well within the range of possible adhesion values for these systems, while the dispersion of the results is exceptionally low for this type of experiment, and found comparable to the experimental uncertainty.