[PDD interview] analyze the activity of applications (cameras) in mobile phones

Big factory SQL Interview questions sql Data analysis

（1） Without considering the business data , The same user may log in multiple times , Or the same customer can place multiple orders in one day ：

-- select sum(TRUE); →  by  1 ;

select  a. Landing time  as dt
       ,count(distinct a.user_id) as  Number of active users 
       ,sum(datediff(b. Landing time ,a. Landing time ) = 1) as  Number of users retained in the next day 
       ,sum(datediff(b. Landing time ,a. Landing time ) = 3) as  Number of users retained in three days 
       ,sum(datediff(b. Landing time ,a. Landing time ) = 7) as  Number of users retained in seven days 
       ,concat(round(sum(datediff(b. Landing time ,a. Landing time ) = 1)/ count(distinct a.user_id)*100,0),'%') as  Next day retention rate 
       ,concat(round(sum(datediff(b. Landing time ,a. Landing time ) = 3)/ count(distinct a.user_id)*100,1),'%') as  Three day retention rate 
       ,concat(round(sum(datediff(b. Landing time ,a. Landing time ) = 7)/ count(distinct a.user_id)*100,2),'%') as  Seven day retention rate 
from ( SELECT  MemberID AS user_id, date(STime) AS  Landing time 
	   FROM OrderList
	   WHERE MemberID IS NOT NULL and shopid = 'WDGC'
	             AND (date(STime) < '2016-01-10') 
	  ) as a  
left join( SELECT  MemberID AS user_id, date(STime) AS  Landing time 
	   	   FROM OrderList
	       WHERE MemberID IS NOT NULL and shopid = 'WDGC'
	             AND (date(STime) < '2016-01-10') 
		  ) as b
on  a.user_id = b.user_id
group by a. Landing time ;

Finally, the next day 、 Three days 、 There is too much retention in seven days ：

Correct insurance practices ： I have considered weight removal ：

select  d.a_t,
		count(distinct d. user id) as  Number of active users ,
		count(distinct case when d. The time interval =1 then d. user id     
		               else null
		               end) as   The amount retained on the next day , 
		count(distinct case when  The time interval =1 then d. user id
		               else null
		               end) /count(distinct d. user id) as  Next day retention rate ,
		count(distinct case when d. The time interval =3 then d. user id     
		               else null
		               end) as  3 Daily retention  ,
		count(distinct case when  The time interval =3 then d. user id
		               else null
		               end) /count(distinct d. user id) as 3 Daily retention rate ,
		count(distinct case when d. The time interval =7 then d. user id     
		               else null
		               end) as  7 Daily retention  ,
		count(distinct case when  The time interval =7 then d. user id
		               else null
		               end) /count(distinct d. user id) as 7 Daily retention rate 
from
(
select    *
         ,timestampdiff(day,a_t,b_t) as  The time interval 
from (  select  a.` user id`
               ,a. Landing time  as a_t 
               ,b. Landing time  as b_t
	    from ( SELECT  MemberID AS  user id, date(STime) AS  Landing time 
			   FROM OrderList
			   WHERE MemberID IS NOT NULL and shopid = 'WDGC'
				     AND (date(STime) < '2016-01-10') 
			  ) as a  
		left join( SELECT  MemberID AS  user id, date(STime) AS  Landing time 
			   	   FROM OrderList
			       WHERE MemberID IS NOT NULL and shopid = 'WDGC'
				         AND (date(STime) < '2016-01-10') 
				  )as b
		on  a.` user id`=b.` user id`
	 ) as c
) as d
group by d.a_t;