Three-day moving average of daily orders

An operations dashboard smooths daily order volume with a 3-day moving average.

From daily_orders, return order_day, orders_count and moving_avg_3d (average of the current day and the two previous days, rounded to 2 decimals). Only return days for which three full days of data exist, so the first two days are excluded. Order by order_day.

1. Step 1: `ROWS BETWEEN 2 PRECEDING AND CURRENT ROW` defines a frame of the current row plus the two before it.
2. Step 2: For the first two days the frame is smaller than 3 rows; `COUNT(*)` over the same frame tells you how many rows were actually available.
3. Step 3: Filtering on `days_in_window = 3` in the outer query keeps only full windows.

WITH w AS (
    SELECT order_day,
           orders_count,
           ROUND(AVG(orders_count) OVER (
                   ORDER BY order_day
                   ROWS BETWEEN 2 PRECEDING AND CURRENT ROW), 2) AS moving_avg_3d,
           COUNT(*) OVER (
                   ORDER BY order_day
                   ROWS BETWEEN 2 PRECEDING AND CURRENT ROW)     AS days_in_window
    FROM daily_orders
)
SELECT order_day, orders_count, moving_avg_3d
FROM w
WHERE days_in_window = 3
ORDER BY order_day;