Σ ABS(drift) per BusinessDay end, one line per account_role. Healthy days sit on the zero baseline; spikes mark when the feed or the parent-roll-up diverged. Use this to spot recurring violations vs one-off events — a role that spikes every Monday is a different problem than a role that spiked once after a deploy. KPIs surface the largest single-day magnitude in the past 7 days.
Max Σ ABS(drift) on any single BusinessDay across leaf accounts in the visible date range. Healthy = $0 (green ✓); red ✗ otherwise. Distinct from the Drift sheet's "Largest Leaf Drift" KPI: that one is the worst single account-day row; this one is the day when total |drift| across all leaves was largest, so it can be smaller (one big drift on a quiet day) or larger (many medium drifts on a busy day) than the row-grain peak.
Max Σ ABS(drift) on any single BusinessDay across parent accounts in the visible date range. Healthy = $0 (green ✓); red ✗ otherwise. Distinct from the Drift sheet's "Largest Parent Drift" — that one is the worst single account-day row, this one is the worst single business day's roll-up. See "Largest Leaf Drift Day" subtitle for the why.
Σ ABS(drift) per BusinessDay end for leaf accounts (cardholder DDAs, merchant DDAs, external counters — every non-aggregate L2 instance in the system). One line per account_role so a single recurring-drift role separates from a one-off spike. Companion: 'Parent Account Drift Over Time' (below) shows the same shape for aggregate accounts. A role hugging zero is healthy; persistent non-zero ⇒ ongoing feed divergence; one-off spike ⇒ isolated event worth drilling into.
Σ ABS(drift) per BusinessDay end for parent accounts (GL controls + concentration masters + funds pools — every aggregate L2 instance that should equal SUM(its leaves)). One line per account_role. Non-zero ⇒ child postings didn't roll up correctly that day. Compare with 'Leaf Account Drift Over Time' (above) to triage whether a drift event lives at the leaf or at the rollup.