Interview question

Nine months into a demand-forecasting rebuild, the team has reached 70% of the accuracy-improvement target. The sponsor says "we have already spent 1.2 million; we must finish". Three more months and 300,000 give a 40% chance of hitting the target, worth 1.5 million in the first year (hypothetical). A narrower rules-plus-simple-model version could be delivered in two months for 150,000, with an 85% chance of capturing 900,000 in the first year. What is your recommendation, and how do you handle the sponsor?