我正在尝试在 Clickhouse 中做一个漏斗，但遇到了在步骤中只记录第一个事件的问题。 让我们假设： 客户端应该经历 3 个事件 - '001_event'、'002_event' 和 '003_event'。 问题是一个客户可以无限次地走这条路，我想考虑到每个客户通过这 3 个事件的这个漏斗的所有尝试。
例如我有数据集：
| appmetica_device_id | event_datetime | 事件名称 |
|---|---|---|
| 168 | 2021-03-01 12:01:00 | 001_event |
| 168 | 2021-03-01 12:02:00 | 002_event |
| 168 | 2021-03-01 12:03:00 | 003_event |
| 168 | 2021-03-01 15:00:00 | 001_event |
| 168 | 2021-03-01 18:00:00 | 001_event |
| 168 | 2021-03-01 18:02:00 | 002_event |
| 200 | 2021-03-01 12:01:00 | 001_event |
| 200 | 2021-03-01 12:02:00 | 002_event |
| 200 | 2021-03-15 12:03:00 | 001_event |
最后我想得到一个包含漏斗通过次数的表格，而不是通过漏斗的用户数量：
| 001_event | 002_event | 003_event |
|---|---|---|
| 5 | 3 | 1 |
我的代码：
select
sum(step1_time != 0) as step1_achieve,
sum(step2_time != 0) as step2_achieve,
sum(step3_time != 0) as step3_achieve,
sum(step4_time != 0) as step4_achieve
from (
select groupArray(event_name) as event, groupArray(event_datetime) as event_time,
arrayFilter(time, name -> (name = '001_event'), event_time, event)[1] as step1_time,
arrayFilter(time, name -> (name = '002_event') and (time >= step1_time) and (step1_time != 0), event_time, event)[1] as step2_time,
arrayFilter(time, name -> (name = '003_event') and (time >= step2_time) and (step2_time != 0), event_time, event)[1] as step3_time,
from app_metrica.events_all
prewhere event_date >= '2021-03-01'
where event_name in ('001_event', '002_event', '003_event')
group by appmetrica_device_id
)
但是我明白了：
| 001_event | 002_event | 003_event |
|---|---|---|
| 2 | 2 | 1 |