Define the segment
The core definition is simple - anyone with a completed purchase or active subscription event - but the edge cases matter: does a customer who churned six months ago still count as "existing" for suppression purposes (usually no, they move to the win-back segment instead), and does a customer with only a free-tier account count as a full customer for suppression versus upsell purposes (usually treated separately, since the messages they need differ).
Sync it - twice, for two different purposes
The same underlying customer list gets synced two ways: as a suppression list attached to acquisition campaigns (excluding them from spend meant for new customers), and separately as a targeting list attached to an upsell or cross-sell campaign built specifically for existing customers. Treating these as one audience with one creative wastes the distinction the data already gives you for free.
Why the creative differs
Suppression itself has no creative - it is pure exclusion. The upsell/cross-sell campaign built from the same list needs messaging that assumes existing product knowledge and account context (a plan upgrade, a complementary product, a renewal message) rather than the awareness-stage messaging an acquisition campaign requires. Running acquisition creative at existing customers, even unintentionally through a missing suppression list, is a mismatched message on top of wasted spend.
KPI framing
Measure acquisition campaigns' cost-per-new-customer after suppression is applied (not blended cost-per-conversion including customers who were never actually new), and measure upsell/cross-sell campaigns against incremental revenue per existing customer reached - two different KPIs for two different uses of the same list.
Illustrative, not a measured result: a subscription business adds a real-time suppression sync where none existed before, on an acquisition campaign that had been running to its full customer file for months. The expected effect is a lower reported cost-per-new-customer once spend against already-converted people stops being counted as acquisition spend at all - illustrative of the mechanism, not a measured client result.