Following marginality approach developed at ZEF, Bonn we identified five marginal sub-districts in Bangladesh i.e. underperforming areas since in such areas yield gaps (potential minus actual yields) are high and productivity gains (of main staple crops) are likely to be achieved. Then we conducted qualitative and quantitative sample survey of 313 poor Smallholders and used livelihood assets and need assessment, Principal component analysis (PCA), logit regression model and cluster analysis to draw the conclusion. Results suggest that only cereal based growth productivity program could not improve food and livelihood security of the poor SHs in the study areas and thus we find that intensive crop system, hybrid seeds, water management technologies, non-crop farming, non-farm enterprise/business, etc. are the suggested potential technology innovations. However, there is very limited availability of extension services among the poor SHs in the study areas and thus despite being poor, the poor SHs have high willingness to pay for extension services, say, awareness and motivation building for increasing agricultural intensification, knowledge service for crop related agricultural production and agriculture related business. Regression results show that household head education, length of the permanent residency in the locality, land ownership, farm size, availability of seeds, having fair price, geographical locations are important determinants for willingness to pay for agricultural extension services. We finally suggest that creating an agricultural technology cum business promoter at the village level address the generalized barriers for the poor SHs for adopting those technology innovations, that is, low level of motivation for the poor SHs, lack of appropriate information, technical knowledge and extension/rural business services/networking, lack of credit and liquid money, etc.