You are the engineering manager for a mission-critical autonomy feature that has missed two internal milestones and is now 8 weeks behind plan. The system is technically viable, but integration work is consuming more senior engineer time than expected, field feedback is mixed, and the product team believes the feature could still unlock a major customer deal. At the same time, another roadmap item is already slipping because the same engineers are being pulled in two directions, and leadership wants a clear recommendation before the next planning cycle.
How do you decide whether to keep investing in a struggling project or stop it?