Dry-land legumes, well adapted to drought-prone areas, have largely been neglected in the past despite the good opportunities they offer for income growth and food (and nutritional) security for the poor. This study evaluated the adoption and impact of two farmer and market-preferred and disease-resistant pigeonpea varieties that were developed and promoted in semi-arid Tanzania. The new varieties were resistant to fusarium wilt, a fungal disease devastating the crop. However, farmers wanting to adopt new varieties did not adopt due to seed access constraints and under-developed seed delivery systems. Adoption of new varieties is therefore analyzed using an augmented double hurdle model that allows estimating variety adoption conditional on seed access thresholds accounting for the additional information on sample separation. The study identifies the crucial role of seed access (local supply), extension, education, participatory decision making, capital, and household assets in determining adoption. The social economic benefits of the technology and policies for improved seed access were further analyzed using the extended economic surplus method (DREAM model). Even under restrictive assumptions, overall discounted benefits were found to be quite attractive, indicating the need for additional efforts to scale-up the success story. Analysis of changes in research benefits from relaxing the seed access constraint showed that net gains would increase by up to 30% if farmer access to improved seeds can be assured. Smallholder farmers are the major beneficiaries along with consumers and rural net-buyers who gain from productivity-induced lower market prices.