SQL/Redshift : Querying user behaviour before first purchase

Question

我有下表:

Table:  users
id | created_ts | username_str
---------------------------------

Table:  activity
user_id | created_ts | activity_descr_str
---------------------------------

Table:  purchase_orders
id | user_id | created_ts | price

users.id 是外键 user_id 的主键。 我想知道有多少用户在首次购买之前进行过 <= 3 次事件。但仅适用于日期 X(例如“2016-08-01”)之后新加入的用户。

非常感谢。

编辑，2016-10-06:

这是我尝试过的:

SELECT COUNT(DISTINCT user_id)
FROM (SELECT po.user_id
             FROM purchase_orders po
        LEFT JOIN activity a USING (user_id)
        JOIN users u ON u.id = po.user_id
    WHERE   DATE (po.created_ts) >= '2016-08-01'
      AND   a.created_ts < po.created_ts
      GROUP BY 1
      HAVING COUNT(a.created_ts) <= 3)

Answer 1

您只需将逻辑分成几个部分:

• 获取每个用户的首次订购日期。
• 过滤掉年龄过大的用户。
• 参加事件，注意日期。
• 计算每个用户的事件数量。
• 过滤并再次计数。

因此，一种方法是这样的查询:

select count(*)
from (select user_id, count(a.user_id) as num_activities
      from (select po.user_id, min(created_ts) as mints
            from purchase_orders po
            group by po.user_id
            having min(created_ts) >= '2016-08-01'
           ) o left join
           activity a
           on a.user_id = o.user_id and a.created_ts < o.created_ts
      group by user_id
     ) u
where num_activities <= 3;