Consumer exposure to cosmetic ingredients is estimated in a tiered manner. Simple Tier1 deterministic aggregate exposure modelling generates a worst case estimate of exposure. Tier1 assumes that a consumer uses all cosmetic products concomitantly daily, at maximum frequency, and products always contain the ingredient at the maximum allowed % w/w concentration. Refining exposure assessment from worst case to more realistic estimates uses evidence from surveys of actual use levels of ingredients and Tier2 probabilistic models, where distributions of consumer use data can be applied. In Tier2+ modelling, occurrence data provides evidence of products on the market actually containing the ingredient. Three case studies are presented using this tiered approach to illustrate progressive refinement. The scale of refinements from Tier1 to Tier2+ modelling for the ingredients, propyl paraben, benzoic acid and DMDM hydantoin were: 0.492 to 0.026; 1.93 to 0.042 and 1.61 to 0.027 mg/kg/day exposure dose. For propyl paraben, moving from Tier1 to Tier2+ represents a refinement from 49-fold to 3-fold overestimate of exposure when compared to a maximum estimate of 0.01 mg/kg/day exposure seen in human studies. Such refinements from worst case to realistic levels of exposure estimation can be critical in the demonstration of consumer safety.