No reliable signal for when to kill vs. persist on an underperforming campaign — budgets burn without learning

Summary

Marketers face a recurring, high-stakes judgment call: results may simply take time, but continuing a dead campaign wastes money, and paid ads make it easy to burn an entire budget while learning almost nothing. With no trustworthy way to separate 'still ramping' from 'never going to work', teams either cut winners early or dump more spend into losers, and the post-mortem produces no reusable learning for the next campaign. Platform reporting is optimized to keep spend flowing, and existing optimization tools flag tactical metrics (CTR, CPC) rather than giving a defensible stop/continue decision rule tied to expected outcome and confidence. This remains essentially unsolved — there is no lightweight 'kill criterion' engine that encodes pacing expectations, minimum learning thresholds, and confidence bounds per channel — and the pain is acute because the money is lost in real time.