What is contemporaneous group analysis? Teach you to use SQL to handle

Catalog

• One 、 Definition of simultaneous group analysis

• Two 、SQL step

  • 1. View the data

  • 2. according to uid 、 Number of aggregated users per month

  • 3. Calculate the difference between month and year （ Days ）

  • 4. Calculate the difference between month and year （ Monthly number ）

  • 5. perspective （ according to uid 、 The number of users who pay for the first time to see the difference between the months and months ）

  • 6. Calculate the retention rate

One 、 Definition of simultaneous group analysis

「 Simultaneous group analysis 」（Cohort Analysis） It's a kind of passing “ Vertically and horizontally ” Combined with the subdivision type analysis of user clustering ：

• 「 Horizontally 」—— Analyze the changes of the same period group as the period goes by

• 「 Vertically 」—— Analyze the differences between groups at the same stage of the life cycle

「 Contemporaneous group 」 It refers to groups in the same period , It can be a user registered on the same day 、 Users who pay for the first time on the same day .

「 The index change of the cycle 」 It refers to the retention rate of users in a certain period 、 Payment rate and other indicators .

Contemporaneous group analysis consists of three core elements ：

• 「 The first time a customer acts 」： This is the basic point for dividing groups in the same period

• 「 Time period dimension 」： such as N Daily retention rate 、N Of the daily conversion rate N Japan , Generally, it is +N Japan 、+N month

• 「 Indicators of change 」： For example, registration conversion rate 、 Payment conversion rate 、 Retention rate and other indicators

Group analysis of the same period gives more detailed measures , Real user behavior can be monitored in real time 、 Measuring user value , And provide support for the optimization and improvement of marketing plan , avoid “ Be averaged ” Vanity data .

Two 、SQL step

I use PostgreSQL Split step to realize the group report of user retention rate in the same period based on the first order date ,「 Each step is reprocessed on the basis of the previous step 」, This is also reflected in the sub query in the code , After sorting out the idea, you will find that it is actually very simple .

The key points are as follows ：

• Count each user's 「 First order time 」

• Calculate the first order time and the actual order time 「 Date difference 」

• For the number of paying users 「 To heavy statistics 」

• Note the fields 「 Format conversion 」

1. View the data

-- 0.  View the data 

SELECT * FROM " journal " LIMIT 10;

2. according to uid 、 Number of aggregated users per month

-- 1.  according to  uid 、 Number of aggregated users per month 

SELECT
 " journal ".uid,
 to_char( to_date( " journal "." date ", 'YYYY-MM' ), 'YYYY-MM' ) AS  years ,
 min(to_char( to_date( " journal "." date ", 'YYYY-MM' ), 'YYYY-MM' )) OVER(PARTITION BY " journal ".uid) AS  Month and year of first payment 
FROM
 " journal " 
GROUP BY
 " journal ".uid,
 to_char( to_date( " journal "." date ", 'YYYY-MM' ), 'YYYY-MM' )
 ORDER BY " journal ".uid;

3. Calculate the difference between month and year （ Days ）

-- 2.  Calculate the difference between month and year （ Days ）

SELECT *,to_date(t. years ,'YYYY-MM') - to_date(t. Month and year of first payment ,'YYYY-MM') AS  Days difference 
FROM (SELECT
 " journal ".uid,
 to_char( to_date( " journal "." date ", 'YYYY-MM' ), 'YYYY-MM' ) AS  years ,
 min(to_char( to_date( " journal "." date ", 'YYYY-MM' ), 'YYYY-MM' )) OVER(PARTITION BY " journal ".uid) AS  Month and year of first payment 
FROM
 " journal " 
GROUP BY
 " journal ".uid,
 to_char( to_date( " journal "." date ", 'YYYY-MM' ), 'YYYY-MM' )
 ORDER BY " journal ".uid) AS t;

4. Calculate the difference between month and year （ Monthly number ）

-- 3.  Calculate the difference between month and year （ Monthly number ）

SELECT t.*,
(case when t." Days difference " <= 30 then ' First month ' 
    when t." Days difference " <= 60 then '+1 month '
   when t." Days difference " <= 90 then '+2 month ' 
   when t." Days difference " <= 120 then '+3 month '
   when t." Days difference " <= 150 then '+4 month '
   else NULL
   END)  AS  Monthly difference 
FROM (SELECT *,to_date(t. years ,'YYYY-MM') - to_date(t. Month and year of first payment ,'YYYY-MM') AS  Days difference 
FROM (SELECT
 " journal ".uid,
 to_char( to_date( " journal "." date ", 'YYYY-MM' ), 'YYYY-MM' ) AS  years ,
 min(to_char( to_date( " journal "." date ", 'YYYY-MM' ), 'YYYY-MM' )) OVER(PARTITION BY " journal ".uid) AS  Month and year of first payment 
FROM
 " journal " 
GROUP BY
 " journal ".uid,
 to_char( to_date( " journal "." date ", 'YYYY-MM' ), 'YYYY-MM' )
 ORDER BY " journal ".uid) AS t) AS t;

5. perspective （ according to uid 、 The number of users who pay for the first time to see the difference between the months and months ）

-- 4.  perspective （ according to  uid 、 The number of users who pay for the first time to see the difference between the months and months ）

SELECT t. Month and year of first payment ,
count(distinct case when t. Month month difference  = 0 then t.uid else NULL end) AS  First month ,
count(distinct case when t. Month month difference  = 1 then t.uid else NULL end) AS "+1 month ",
count(distinct case when t. Month month difference  = 2 then t.uid else NULL end) AS "+2 month ",
count(distinct case when t. Month month difference  = 3 then t.uid else NULL end) AS "+3 month ",
count(distinct case when t. Month month difference  = 4 then t.uid else NULL end) AS "+4 month "
FROM (SELECT * FROM (SELECT *,round((to_date(t. years ,'YYYY-MM') - to_date(t. Month and year of first payment ,'YYYY-MM')) / 30,0) AS  Month month difference 
FROM (SELECT
 " journal ".uid:: text,
 to_char( to_date( " journal "." date ", 'YYYY-MM' ), 'YYYY-MM' ) AS  years ,
 min(to_char( to_date( " journal "." date ", 'YYYY-MM' ), 'YYYY-MM' )) OVER(PARTITION BY " journal ".uid) AS  Month and year of first payment 
FROM
 " journal " 
GROUP BY
 " journal ".uid,
 to_char( to_date( " journal "." date ", 'YYYY-MM' ), 'YYYY-MM' )
 ORDER BY " journal ".uid) AS t) AS t) AS t
 GROUP BY t. Month and year of first payment ;

6. Calculate the retention rate

-- 5.  Calculate the retention rate 

SELECT t. Month and year of first payment ,t. First month ,
round((t."+1 month "::numeric / t. First month ::numeric) * 100,2)::text || '%' AS "1 After month ",
round((t."+2 month "::numeric / t. First month ::numeric) * 100,2)::text || '%' AS "2 After month ",
round((t."+3 month "::numeric / t. First month ::numeric) * 100,2)::text || '%' AS "3 After month ",
round((t."+4 month "::numeric / t. First month ::numeric) * 100,2)::text || '%' AS "4 After month "
FROM(SELECT t. Month and year of first payment ,
count(distinct case when t. Month month difference  = 0 then t.uid else NULL end) AS  First month ,
count(distinct case when t. Month month difference  = 1 then t.uid else NULL end) AS "+1 month ",
count(distinct case when t. Month month difference  = 2 then t.uid else NULL end) AS "+2 month ",
count(distinct case when t. Month month difference  = 3 then t.uid else NULL end) AS "+3 month ",
count(distinct case when t. Month month difference  = 4 then t.uid else NULL end) AS "+4 month "
FROM (SELECT * FROM (SELECT *,round((to_date(t. years ,'YYYY-MM') - to_date(t. Month and year of first payment ,'YYYY-MM')) / 30,0) AS  Month month difference 
FROM (SELECT
 " journal ".uid:: text,
 to_char( to_date( " journal "." date ", 'YYYY-MM' ), 'YYYY-MM' ) AS  years ,
 min(to_char( to_date( " journal "." date ", 'YYYY-MM' ), 'YYYY-MM' )) OVER(PARTITION BY " journal ".uid) AS  Month and year of first payment 
FROM
 " journal " 
GROUP BY
 " journal ".uid,
 to_char( to_date( " journal "." date ", 'YYYY-MM' ), 'YYYY-MM' )
 ORDER BY " journal ".uid) AS t) AS t) AS t
 GROUP BY t. Month and year of first payment ) AS t;

- END -

 contrast Excel The cumulative sales of the series of books reached 15w book , Make it easy for you to master data analysis skills , You can click on the link below to learn about purchasing ：