In the risk assessment of the ability of a substance or preparation to cause skin irritation, a patch test involving human volunteers is often seen as providing definitive information. However, the skin exposure conditions found in an occluded patch may be far removed from those in a particular use situation, not least for such reasons as the high dose per unit area, the duration of exposure and the maceration of the skin. Whilst these factors may enhance sensitivity in the identification of intrinsic hazard, they may do little to ensure accurate safety evaluation for real use. In this paper, we report data from a series of studies with an unmarketed facial skin cosmetic product. Whilst the product was unexpectedly highly irritating in a standard patch test in 30 volunteers, subsequent use tests, including a 6× daily open application to the elbow for 3 weeks and 2× daily application in a half-face test lasting 3 to 4 weeks and involving 52 volunteers failed to show any evidence of skin irritation. It is concluded that the most meaningful results for skin irritation risk assessment are likely to come from studies which involve relevant patterns of exposure.