如何根据事件在 SQL 中计算客户保留率?
Posted
技术标签:
【中文标题】如何根据事件在 SQL 中计算客户保留率?【英文标题】:How to calculate customer retention in SQL based on events? 【发布时间】:2020-08-25 13:56:56 【问题描述】:我正在尝试创建一个 SQL 语句来找出哪些客户没有连续参加三个活动
表 1 - 客户: 客户 ID、客户名称
+-------------+---------------+
| Customer ID | Customer Name |
+-------------+---------------+
| 01 | Customer 01 |
| 02 | Customer 02 |
| 03 | Customer 03 |
+-------------+---------------+
表 2 - 事件 事件 ID、事件日期、事件名称
+----------------------------------+
| Event ID Event Date Event Name |
+----------------------------------+
| 01 01/01/2020 Event 01 |
| 02 01/15/2020 Event 02 |
| 03 02/15/2020 Event 03 |
| 04 03/13/2020 Event 04 |
| 05 05/17/2020 Event 05 |
| 06 06/20/2020 Event 06 |
+----------------------------------+
表 3 - 事件活动 事件 ID、客户 ID
+----------+-------------+----+
| Event ID | Customer ID | |
+----------+-------------+----+
| 01 | | 01 |
| 01 | | 02 |
| 01 | | 03 |
| 02 | | 01 |
| 03 | | 01 |
| 03 | | 02 |
| 04 | | 01 |
| 05 | | 01 |
| 06 | | 01 |
| 06 | | 03 |
+----------+-------------+----+
现在我正在寻找那些连续没有参加 3 场活动的客户。
所以在给定的示例中,客户 2 和客户 3。
我使用了史蒂夫的建议。这里是更新的 SQL 语句:
drop table if exists dbo.customer;
create table dbo.customer(
CustID int not null,
CustName varchar(20) not null);
insert dbo.customer(CustID, CustName) values
(1,'Cust 1'),
(2,'Cust 2'),
(3,'Cust 3'),
(4,'Cust 4'),
(5,'Cust 5')
;
drop table if exists dbo.events;
create table dbo.events(
EventID int not null,
EventDate date not null,
EventName varchar(20) not null);
insert dbo.events(EventId, EventDate, EventName) values
(1,'2020-01-01','Event 1'),
(2,'2020-01-15','Event 2'),
(3,'2020-02-15','Event 3'),
(4,'2020-03-13','Event 4'),
(5,'2020-05-17','Event 5'),
(6,'2020-06-20','Event 6');
drop table if exists dbo.eventactivity;
create table dbo.eventactivity(
EventID int not null,
CustID int not null);
insert dbo.eventactivity(EventID, CustID) values
(1,1),
(1,2),
(1,3),
(1,4),
(1,5),
(2,1),
(2,2),
(2,4),
(2,5),
(3,1),
(3,5),
(4,1),
(4,5),
(5,1),
(5,2),
(5,3),
(5,5),
(6,1),
(6,2),
(6,3);
(6,5);
这里:
;with
events_sorted as (
select e.*, row_number() over (order by EventDate) seq from dbo.events e),
activity_lag as
(
select
a.*, e.seq,
lag(e.seq, 1, 0) over (partition by CustId order by e.seq) lag_seq,
iif(lag(e.seq, 1, 0) over (partition by CustId order by e.seq)=0, 1,
iif((e.seq-lag(e.seq, 1, 0) over (partition by CustId order by e.seq))>1, 1, 0)) seq_break
from dbo.eventactivity a
join events_sorted e on a.EventID=e.EventID
),
activity_lag_sum as (
select
alag.*, sum(seq_break) over (partition by CustId order by alag.seq) seq_grp
from
activity_lag alag
),
three_in_a_row_cte as (
select distinct CustId
from activity_lag_sum
group by CustID, seq_grp
having count(*)>=3
)
select *
from customer c
where not exists(select 1
from three_in_a_row_cte r
where c.CustID=r.CustID);
问题是,这会返回客户 2、客户 3、客户 4 - 客户 2 确实参加了 2 个活动,跳过了 2 个,参加了 2 个,所以客户 2 不应该在列表中。
有什么建议吗?
【问题讨论】:
您使用的是哪种 DBMS 产品? “SQL”只是所有关系数据库都使用的一种查询语言,而不是特定数据库产品的名称。请为您使用的数据库产品添加tag。 Why should I tag my DBMS 所以您正在寻找没有连续参加过三场活动的客户?或者您是在寻找连续三场活动中至少失踪一次的客户? 您好 Jere,第一部分 - “连续未参加过 3 场活动的客户” - 或者换句话说 - 未连续参加 3 场活动的客户。 你试过什么了吗??? 【参考方案1】:以下查询返回的 CustId 具有:1) 跳过 3 个或更多活动,或 2) 参加的活动总数少于 3 个。
;with
events_sorted as (
select e.*, row_number() over (order by EventDate) seq from #events e),
activity_lag as
(
select
a.*, e.seq,
lag(e.seq, 1, 0) over (partition by CustId order by e.seq) lag_seq,
iif(lag(e.seq, 1, 0) over (partition by CustId order by e.seq)=0, 1,
iif((e.seq-lag(e.seq, 1, 0) over (partition by CustId order by e.seq))>1, 1, 0)) seq_break
from #eventactivity a
join events_sorted e on a.EventID=e.EventID
)
select distinct CustId
from activity_lag
where seq-lag_seq>3
union all
select CustId
from activity_lag
group by CustId
having count(*)<3;
结果
CustId
3
4
【讨论】:
太棒了(如果我能用更大的字母写的话)...非常感谢这里的帮助。 如果还有什么问题请告诉我【参考方案2】:您只需要连续跳过 3 个或更多事件的客户,您可以通过从事件活动表本身查询来获得。请在下面找到查询和查询结果:-
创建表格
create table event_activity ("event_id" varchar(2),"customer_id" varchar(2))
insert into event_activity
values ('01','01'),('01','02'),('01','03'),('02','01'),('02','03'),('03','01'),
('03','02'),('04','01'),('04','02'),('05','02'),('06','01'),('06','03'),
('07','03'),('08','04'),('12','04'),('13','05')
以上查询将产生下表:-
event_id | customerid
---------------------
01 | 01
01 | 02
01 | 03
02 | 01
02 | 03
03 | 01
03 | 02
04 | 01
04 | 02
05 | 02
06 | 01
06 | 03
07 | 03
08 | 04
12 | 04
13 | 05
从上表我们可以观察到除了客户 4 和 5 之外的所有客户连续跳过少于 3 个事件。根据您的问题,我们只需要 4 和 5,因为 4 已连续跳过 3 场活动,但 5 只参加了 1 场活动。
PS : - 在这里您可以发现客户 3 也跳过了 3 个事件,但在此之前他有 参加了一些活动没有跳过任何所以,它必须被淘汰。
最终查询
select c.customer_id
from
(
select customer_id,
skipped_count,
lag(skipped_count,1) over (partition by customer_id order by event_id)
as ref
from
(
select customer_id,
event_id,
LAG(event_id,1) over (partition by customer_id order by event_id)
as previous_event,(event_id - LAG(event_id,1) over (partition by
customer_id order by event_id)-1) as skipped_count
from
(
select CONVERT(int,event_id) as event_id,
CONVERT(int, customer_id) as customer_id
from event_activity
)a
)b
)c
join
(
select convert(int,customer_id) as customer_id,
count(event_id) as count_event
from event_activity
group by customer_id
)d
on c.customer_id=d.customer_id
where (skipped_count >=3 and ref is null)
or count_event = 1
or (skipped_count >=3 and ref > 2)
输出
4
5
【讨论】:
以上是关于如何根据事件在 SQL 中计算客户保留率?的主要内容,如果未能解决你的问题,请参考以下文章
SQL如何计算订单购买之间的平均时间? (根据下一行和上一行进行sql计算)
SQL sever 2008 如何根据一个表中的两个时间计算出时间差