Poverty is perceived to be vague in terms of (i) making judgement about who is to be considered as poor, (ii) selecting the relevant dimensions and indictors. Any benchmark selection to identify poor remains somewhat arbitrary and the vagueness exists irrespective of whether the conventional or non-conventional poverty measure is used. To address the issue of vagueness, this study employs fuzzy regression as a natural alternative to the conventional approach. The fuzzy logic assigns degree of membership to a set of poor people on a scale from 0 to 1 instead of the rigid dichotomization. To cater the second issue, we measure the welfare level of individuals using Engel curve method as it gives a lot of information regarding the consumption behavior of consumers at different levels of total expenditures and for various family compositions. Pakistan Social and Living Standards Measurement (PSLM) survey 2015-16 is used to estimate poverty. Findings reveal that poverty estimates vary significantly across the provinces and regions. Overall, highest incidence of poverty is observed in Balochistan followed by Sindh, KPK is the least poor province. Poverty is not only a rural but a provincial phenomenon as well in Pakistan.