In a large-scale online experiment, 1% of users are assigned to treatment to limit exposure risk. The test runs for four weeks and detects no significant effect. A colleague concludes the treatment has no effect. What is the concern?