• metabase sql统计留存率


     1 -- 上周注册了100人,上周以前注册(200人),
     2 -- 本周登录了50人(上周注册的100人里面有50人本周登录),
     3 -- 本周登录了20人(上周之前注册注册的200人里面有80人在上周登录过并且本周登录了20人)
     4 -- 新用户留存率 50/100
     5 -- 老用户留存率 20/80
     6 
     7 使用sql统计出新旧留存率:
     8 
     9 SELECT CONCAT(ROUND((MA.CNT3/MA.CNT1)*100,2),'%')  AS "新用户留存率",
    10        CONCAT(ROUND((MA.CNT5/MA.CNT4)*100,2),'%') AS "老用户留存率"
    11 FROM (SELECT
    12         SUM(CASE WHEN A.CREATE_TIME BETWEEN DATE_SUB({{S_TIME}},INTERVAL WEEKDAY({{S_TIME}}) + 8 DAY) AND DATE_SUB({{S_TIME}},INTERVAL WEEKDAY({{S_TIME}}) + 2 DAY) -- 上周日至上周六
    13                  THEN 1 ELSE 0 END ) AS CNT1, -- 上周注册数(新用户留存分母),
    14         SUM(CASE WHEN A.CREATE_TIME < DATE_SUB({{S_TIME}},INTERVAL WEEKDAY({{S_TIME}}) + 8 DAY) -- 上周日
    15                  THEN 1 ELSE 0 END ) AS CNT2,  -- 上周以前注册数
    16         SUM(CASE WHEN A.CREATE_TIME BETWEEN DATE_SUB({{S_TIME}},INTERVAL WEEKDAY({{S_TIME}}) + 8 DAY) AND DATE_SUB({{S_TIME}},INTERVAL WEEKDAY({{S_TIME}}) + 2 DAY) -- 上周日至上周六
    17                   AND EXISTS (SELECT 1 FROM FCOIN.SETTLEDETAILS B 
    18                                WHERE B.SETTLEDATE BETWEEN DATE_SUB({{S_TIME}},INTERVAL WEEKDAY({{S_TIME}}) +1 DAY) AND DATE_SUB({{S_TIME}},INTERVAL WEEKDAY({{S_TIME}}) - 5 DAY) -- 本周日至本周六
    19                                  AND B.USERID = A.ID) 
    20                  THEN 1 ELSE 0 END) AS CNT3,  -- 上周注册的在本周登录(新用户留存分子)
    21         SUM(CASE WHEN A.CREATE_TIME < DATE_SUB({{S_TIME}},INTERVAL WEEKDAY({{S_TIME}}) + 8 DAY) -- 上周日
    22                   AND EXISTS (SELECT 1 FROM FCOIN.SETTLEDETAILS B  
    23                                WHERE B.SETTLEDATE BETWEEN DATE_SUB({{S_TIME}},INTERVAL WEEKDAY({{S_TIME}}) + 8 DAY) AND DATE_SUB({{S_TIME}},INTERVAL WEEKDAY({{S_TIME}}) + 2 DAY) -- 本周日至本周六
    24                                  AND B.USERID = A.ID)
    25                  THEN 1 ELSE 0 END) AS CNT4,  -- 上周以前注册,上周登录过,本周也登录过(旧用户留存分母)
    26         SUM(CASE WHEN A.CREATE_TIME < DATE_SUB({{S_TIME}},INTERVAL WEEKDAY({{S_TIME}}) + 8 DAY) -- 上周日
    27                   AND EXISTS (SELECT 1 FROM FCOIN.SETTLEDETAILS B  
    28                                WHERE B.SETTLEDATE BETWEEN DATE_SUB({{S_TIME}},INTERVAL WEEKDAY({{S_TIME}}) + 8 DAY) AND DATE_SUB({{S_TIME}},INTERVAL WEEKDAY({{S_TIME}}) + 2 DAY) -- 上周日至上周六
    29                                  AND B.USERID = A.ID)
    30                   AND EXISTS (SELECT 1 FROM FCOIN.SETTLEDETAILS B 
    31                                WHERE B.SETTLEDATE BETWEEN DATE_SUB({{S_TIME}},INTERVAL WEEKDAY({{S_TIME}}) +1 DAY) AND DATE_SUB({{S_TIME}},INTERVAL WEEKDAY({{S_TIME}}) - 5 DAY) -- 本周日至本周六
    32                                  AND B.USERID = A.ID) 
    33                  THEN 1 ELSE 0 END) AS CNT5  -- 上周以前注册,上周登录过,本周也登录过(旧用户留存分子)
    34    FROM ACCOUNT.USERS A
    35   WHERE A.NATION<>'86' and A.account_role<50 and service_area<>'cn') MA
  • 相关阅读:
    1. Hello UWP
    ASP.NET MVC SignalR(1):背景
    ASP.NET MVC SignalR
    nekohtml转换html时标签变大写的问题
    nohup启动java命令导致dubbo无法注册
    SOA架构改造简单记录
    [转]BloomFilter——大规模数据处理利器
    IOS行货自动打包
    Kruskal算法java版
    prim算法java版
  • 原文地址:https://www.cnblogs.com/5sdba-notes/p/13955930.html
Copyright © 2020-2023  润新知