Yield models that gave farmers plot-level advice and contributed to an estimated 15% bigger harvest on pilot farms.
Smallholder farms record planting, inputs, weather and yields, but rarely have anyone to turn that data into decisions. Could data science give farmers advice they could actually act on?
I built predictive models in Python to find which inputs, timing and conditions best explained yield differences between similar plots. The models produced plot-level recommendations ranked by expected impact, which we tested on pilot farms and refined with feedback from the field.
The recommendations contributed to an estimated 15% increase in crop yield across pilot farms, about seven months after the first dataset.