Each coin accumulates one coin-day per day of dormancy, and spending it destroys the accumulated total. Ten coins untouched for 500 days destroy 5,000 coin-days when moved, while active trading between exchanges destroys almost none.
The purpose is to filter noise. Ordinary transfer volume is dominated by short-term churn, whereas CDD spikes only when genuinely old supply wakes up, which is the behaviour of long-term holders who typically move coins near cycle extremes.
Caveats remain. A custodian migrating cold storage destroys enormous coin-days without anyone selling anything, and the metric cannot see intent. Watch clusters of spikes rather than single ones, and cross-check with exchange-netflow to see whether the old coins went to a venue.
Related: dormancy, hodl-waves, sopr, exchange-netflow