a b s t r a c tSoil fertility is a major constraint to agricultural development in most of the Sahel, with P being the most limiting nutrient for millet production on acid sandy soils. To address this issue, microdose applications of P fertilizer have been widely advocated in recent years. However, little is known regarding the effect of farmer management practices and environmental factors on millet's yield response to this technique. For this purpose, 276 farmer demonstrations were setup across a 3-year period in the Fakara region, western Niger. Five strata were considered based on antecedent organic manure management (corralling or transported manure). At each demo site, conventional management was compared to basal microdose fertilizer application of DAP (2 g hill −1 ), NPK (6 g hill −1 ), or DAP (2 g hill −1 ) with urea (1 g hill −1 ) applied at tillering. Millet grain yields on control plots were low (84% < 400 kg ha −1 ), reflecting the unfavorable environmental conditions of the area. On average, the application of DAP, NPK and DAP + urea increased grain yields by 43, 46 and 69 kg ha −1 (2001)(2002). A positive response to microdose fertilization was observed for 92% of the sites where yields on control plots were <100 kg ha −1 but only for 32% of the sites where yields on control plots were >500 kg ha −1 . In particular, the positive response to microdosing increased with later sowing given that late sowing tended to reduce yields on control plots. Higher rainfall during the early growing season favored a positive response to microdosing. On average over DAP and DAP + urea, 36% of the demonstrations had value-cost ratios (VCR) < 1. However, for low yielding control plots (<200 kg grain ha −1 ), 26% of the demonstrations had VCR < 1, whereas for high yielding plots (>400 kg grain ha −1 ), 55% had a VCR < 1. Not accounting for labor, DAP and DAP + urea had similar economic returns. The use of NPK could not be recommended as the cost per unit P is 3 times higher than DAP. It appears that, for the Fakara study area, microdosing may best be targeted to areas with low expected yields. In particular, it may serve as a famine mitigation strategy in case of late sowing. Nevertheless, for poorly endowed areas such as the Fakara, the economic risk associated with microdosing (2 g DAP hill −1 ) appears higher than has hitherto been reported and widespread adoption may not be warranted without institutional support.