Why does joining two child tables inflate a SUM?

Independent one-to-many joins multiply rows: two payments and three line items can produce six joined rows for one order. Summing afterward counts repeated values. Aggregate each child to one row per parent before joining, or calculate independent totals in separate correlated subqueries. Test a parent with two records on each side and compare against manually calculated totals. SUM(DISTINCT amount) is not a general fix because two legitimate payments may have equal amounts. Establish the expected grain of every intermediate relation before adding another join. Minimal example:

-- Aggregate each child before joining; verified with SQLite.
WITH payments(order_id, amount) AS (VALUES (1, 10), (1, 20)),
     items(order_id, quantity) AS (VALUES (1, 1), (1, 2)),
     paid AS (SELECT order_id, SUM(amount) total FROM payments GROUP BY order_id),
     sold AS (SELECT order_id, SUM(quantity) units FROM items GROUP BY order_id)
SELECT paid.total, sold.units FROM paid JOIN sold USING (order_id);
-- Expected: 30, 3