This paper provides fresh empirical evidence on the adaptation process to face climate changes through the analysis of original cross-sectional data collected at household-level in Niger merged with detailed geo-referenced climatic information. In particular, we identify the main drivers and barriers of crop and labour diversification, which constitute two livelihood strategies in mitigating the adaptation deficit by employing a Seemingly-Unrelated Regression (SUR) model, which accounts for potential interdependence among different diversification practices. Secondly, the effectiveness of diversification practices is assessed by means of three complementary welfare measures, namely income changes, food security and the poverty gap using quantile regression and instrumental variable strategy. We find that, aside from climate shocks, the diversification level varies in response to the educational level of household members and spatial location as well as the adoption of ICTs. The impacts of diversification appear differentiated. While labour diversification is always positively associated with all the three welfare measures, positive coefficients of crop diversification are significant only when associated to food security. Robust causal inference confirms that anomalies in rainfall patterns and droughts in particular, induce adaptation responses, which result in welfare gains limited by a richer calorie intake, while the effects on income and severity of poverty appear detrimental.