The mutant and uncertain behaviour of insurance environments does not make advisable to use a wide data-base when calculating claim reserves and so, quantifying provisions with fuzzy numbers becomes suitable. This paper firstly describes the fuzzy least squares regression that will be used in posterior developments. Subsequently we expose a claim reserving method that combines fuzzy regression with the classical statistical scheme based on two ways of ANOVA. Finally we develop a numerical application where we show in detail how to use our method to fit expected claiming costs and their variability and compare its results with those from conventional ANOVA two ways.