数据分析项目 --SQL 的用户消费分析

1、导入数据
方法 1 直接导入 CSV 数据，不用建表做关联，如下图

方法 2 建表做关联后导入数据
建表，订单表（订单编号，用户 ID，付款情况，价格，订单付款时间）

create table orderform(  
orderid varchar(20) null,  userid  varchar(10) null, paid varchar(10) null,  
price varchar(10) null, payment_time datetime null)  
​

建表，客户信息表

create table userlist(  
userid varchar(10) null,gender varchar(5) null,birth datetime null)
​

1.1 需要更新下列名，更新列名的语句如下，不过还是表设计里面改起来最快
exec sp_rename 'order_info_utf.column1','order_info_utf.orderid'

2、用户消费行为分析相关
有关用户消费行为的统计指标，主要的分析内容主要是：
销量和消费金额分析、消费的时间节点、复购率回购率分析、用户分层、用户生命周期和消费周期等，具体可以看看

2.1 可以统计不同月份订单数、用户数、下单总金额

select convert(varchar(6),payment_time,112) as year_month, count(orderid) as order_num ,
count(distinct userid) as user_num,sum(price) as sum from order_info_utf  where paid='已支付'
group by convert(varchar(6),payment_time,112)
order by year_month
​

2.2 统计用户复购率、回购率（这两个不一样的）
a. 复购率 当月消费一次以上(叫复购)人数的占比情况（消费次数 >1 的人数 ÷ 消费总人数）

这里遇到了一个难点，就是 SQL 转换成带 % 号的比率，折磨人，计算完毕的比率是 float 型，% 是字符型，所以拼接是要将 float 转换后再加 %,然后是 char 和 varchar 的区别导致的，也可以用 convert 转换

select t.*,concat(cast(round((cast(t.repeat_user as float)/t.user_num)*100,2)as varchar),'%') as repeat_rate from 
(select  year_month,count(distinct userid)as user_num,
sum(case when b.frequency>1 then 1 else 0 end) as repeat_user from 
(select convert(varchar(6),payment_time,112) as year_month , userid,count(userID) as frequency from order_info_utf 
where paid='已支付' group by convert(varchar(6),payment_time,112), userID) b
group by year_month) t
order by year_month
​

b.回购率，回购率一般以月为单位区分，上月购买过的人在本月又买了,先找出本月购买的

select a.year_month1 年月,a.购买人数,b.回购人数  from 
 (select convert(varchar(6),payment_time,112) as year_month1,count( DISTINCT userid) 购买人数
 from order_info_utf where paid='已支付' group by convert(varchar(6),payment_time,112))a
 left join 
 (select convert(varchar(6),payment_time,112) year_month2,count (DISTINCT a.userid )回购人数 
 from order_info_utf a where a.paid='已支付' and  a.userid in 
 (select userid from order_info_utf B where paid='已支付' 
 and convert(varchar(6),B.payment_time,112)=convert(varchar(6),a.payment_time,112)-1)
 group by convert(varchar(6),payment_time,112))b on a.year_month1=b.year_month2
 order by a.year_month1
​

2.3.统计男女的消费频次是否有差异，一般用户分层可以从性别、年龄段来着手
a.从性别分层找差异

 select u.gender,sum(b.frenquency) as sum_frequency,count(distinct u.userid),
round(convert(float,sum(b.frenquency))/convert(float,count(distinct u.userid)),2)
 as avg_frequency from user_info_utf u left join 
 (select userid,count(userid) as  frenquency from order_info_utf  where paid='已支付'
 group by userid) b on u.userid=b.userid  where gender is not null group by gender
​

结论： 可见在性别方面，平均频次相差不大，那么可以顺便分析下不同性别的消费总额与平均额

select u.gender,sum(b.expenditure) as sum_expenditure,count(distinct u.userid), round(convert(float,sum(b.expenditure))/convert(float,count(distinct u.userid)),2)
as avg_expenditure from user_info_utf u left join 
(select userid,sum(price) as expenditure from order_info_utf  where paid='已支付' group by userid)b
on u.userid=b.userid where u.gender is not  null
group by u.gender
​

b.年龄分层，从年龄段来分析，先将年龄分层，最小的 3 岁，最大的 1947，显然大于 119 的基本属于异常值，可以将其剔除

 select c.age_leval,count(c.userid) as sum_frequency ,sum(a.expenditure) as sum_expenditure,
 round(sum(a.expenditure)/count(c.userid),4) as avg_expenditure from 
 (select b.userid,age_leval=(case when age <=17 then '0-17岁' when age between 18 and 25 then '18-25岁'when age between 26 and 35 then '26-35岁'
                        when age between 36 and 45 then '36-45岁' when age between 46 and 55 then '46-55岁' else '56岁以上' end)
 from (select userid,age=datediff(yy,convert(varchar(10),birth,23),GETDATE()) from user_info_utf where birth is not null
 and datediff(yy,convert(varchar(10),birth,23),GETDATE())<119) b )c
 left join 
(select userid,sum(price) as expenditure from order_info_utf  where paid='success' group by userid) a on c.userid=a.userid
 group by c.age_leval
​

可以看出 26-35 和 36-45 年龄段的消费热刺和消费额都遥遥领先，但是在人均方面优势并不是十分明显，只有 36-45 年龄段的人均比较高，总消费额和总消费频次最多的 26-35 年龄段，人均消费和 56 岁以上年龄段的人均消费相差无几。

2.4.统计多次消费的用户，第一次和最后一次消费时间的间隔

select distinct userid,count(userid)as frequency,
convert(varchar(10),min(payment_time),120) as first_payment_time,
convert(varchar(10),max(payment_time),120) as last_payment_time,
datediff(dd,min(payment_time),max(payment_time)) as time_lag
from order_info_utf where paid='已支付' 
group by userid having count(userid)>=2
order by frequency desc
​

2.5.消费黄金法则，二八原则,订单数量最多的 20% 客户，创造了多少的消费额，占总消费额的比例
以及订单额最多的客户，占到 80% 的订单额的用户的数占比总用户数的比率

select count(userid),sum(user_price),c.sum_user,c.sum_price,round(sum(b.user_price)/c.sum_price,2) as rate
from (select top 20 percent  userid,count(userid) as user_frequency,sum(price) as user_price  from order_info_utf 
where paid='success' group by userid order by count(userid) desc) b ,
(select count(distinct userid) as sum_user ,sum(price) as sum_price from  order_info_utf where paid='success') c
group by c.sum_price,c.sum_user
​

以上，符合二八法则，关键用户满足，如果比率小于 0.8，可以着重发展先大客户