SQL 计算跳出率

Question

我有以下 SQL 表（Postgres）：

+-----------+------------+-------+
| SessionID |  Received  | Event |
+-----------+------------+-------+
|         1 | 1595207019 | visit |
|         1 | 1595207020 | play  |
|         2 | 1595207040 | visit |
|         1 | 1595207050 | click |
+-----------+------------+-------+

我想计算跳出率，其中跳出被定义为访问事件，之后没有任何其他具有相同 session id 的事件。

Answer 1

嗯。 . . 我认为您想通过session_id进行总结并计算访问类型。 然后聚合。 您的问题不是 100% 清楚，但我认为：

select (count(*) filter (where num_notvisits = 0) * 1.0 / count(*)) as bounce_rate
from (select session_id,
             count(*) filter (where event = 'visit') as num_visits,
             count(*) filter (where event <> 'visit') as num_notvisits
      from t
      group by session_id
     ) s
where num_visits > 0;

这是访问和非访问事件的会话数除以访问的会话数的比率。

实际上，您可以更简单地将外部 select 表述为：

select avg( (num_notvisits = 0)::int ) as bounce_rate

Answer 2

您可以像这样显式查询“反弹”的数量：

select count(*)
from t as t1
where t1.event = 'visit'
  and not exists (select * from t as t2 where t1.received < t2.received and t1.sessionid = t2.sessionid)

不确定“跳出率”的分母具体是什么？ 每个 session 的反弹？ # 反弹/# 事件？