A data scientist reports a 95% CI of [0.001, 0.003] for a model improvement metric. A stakeholder concludes this is strong evidence the improvement is meaningful. What is the flaw in this reasoning?