Three months in a row - sql

Question

Please help. How to find users who have refueled 50 or more liters every month for 3 months in a row? Customers are counted on the last month of the three-month period.
I'm new to SQL and only guessed to filter out customers who refuel more than 50 liters per month.
SELECT [Client]
,SUM([Litr]) AS Volume
,DATEPART(mm, [Date_transaction]) AS Month_num
,DATEPART(yyyy, [Date_transaction]) AS Year_num
FROM [dbo].[Транзакции]
GROUP BY [Client], DATEPART(mm, [Date_transaction]), DATEPART(yyyy, [Date_transaction])
HAVING SUM([Litr]) >= 50
I received this table:
|Client|Volume |Month_num|Year_num|
|:-----|:---------------|:--------|:-------|
|33 |52,7497194163861|8 |2019 |
|33 |58,1573308846036|9 |2019 |
|33 |148,852157943067|10 |2019 |
|33 |61,2182430364249|12 |2019 |
|55 |73,0741761044791|1 |2019 |
|55 |136,802367105397|3 |2019 |
|58 |88,0522395673911|7 |2019 |
|58 |140,965207631874|8 |2019 |
|58 |130,20099989797 |9 |2019 |
|181 |507,009488827671|6 |2019 |
Desired result:
|Month_num|COUNT(Client)|
|:--------|-------------|
|9 | 1 | (Client 58)
|10 | 1 | (Client 33)

Answer 1

We can group by Client and the start of each month. Then filter out all rows with SUM(tr.Litr) < 50.
Now we can use the LAG window function to check if the previous two rows were the previous two months.
The final result is simply the count of clients who satisfy that condition
SELECT
MonthStart,
NumOfClients = COUNT(*)
FROM (
SELECT
tr.Client,
v.MonthStart,
PrevMonth1 = LAG(v.MonthStart, 1) OVER (PARTITION BY tr.Client ORDER v.MonthStart),
PrevMonth2 = LAG(v.MonthStart, 2) OVER (PARTITION BY tr.Client ORDER v.MonthStart)
FROM dbo.Транзакции tr
CROSS APPLY (VALUES (
DATEFROMPARTS(YEAR(tr.Date_transaction), MONTH(tr.Date_transaction), 1)
) ) v(MonthStart)
GROUP BY tr.Client, v.MonthStart
HAVING SUM(tr.Litr) >= 50
) trGrouped
WHERE DATEDIFF(month, PrevMonth1, MonthStart) = 1
AND DATEDIFF(month, PrevMonth2, PrevMonth1) = 1
GROUP BY MonthStart;

Answer 2

Try this
scripts for table and insert data. Although It was expected from #Andrey
create table #test(client int, litre decimal(10, 5), Date_transaction date)
insert into #test values (33 ,52.7497194163861,'8/1/2019')
insert into #test values (33 ,58.1573308846036,'9/1/2019')
insert into #test values (33 ,148.852157943067,'10/1/2019')
insert into #test values (33 ,61.2182430364249,'12/1/2019')
insert into #test values (55 ,73.0741761044791,'1/1/2019')
insert into #test values (55 ,136.802367105397,'3/1/2019')
insert into #test values (58 ,88.0522395673911,'7/1/2019')
insert into #test values (58 ,140.965207631874,'8/1/2019')
insert into #test values (58 ,130.20099989797 ,'9/1/2019')
insert into #test values (181 ,507.009488827671,'6/1/2019')
Solution Script
SELECT COUNT(distinct client)client, max(month)month
FROM
(
select t1.client, t1.litre, t1.Date_transaction,(DATEPART(mm, t1.Date_transaction)+2)month
from
#test t1
inner join #test t2 ON (t1.client = t2.client and (DATEPART(mm, t1.Date_transaction)+1) = DATEPART(mm, t2.Date_transaction))
inner join #test t3 ON (t1.client = t3.client and (DATEPART(mm, t1.Date_transaction)+2) = DATEPART(mm, t3.Date_transaction))
where t1.litre>50
)T
GROUP BY client

Answer 3

How to find users who have refueled 50 or more liters every month for 3 months in a row?
I would use aggregation and lag(). To get the last month where this occurs:
SELECT t.*
FROM (SELECT t.Client, v.yyyymm,
SUM(t.Litr) AS Volume,
LAG(v.yyyymm, 2) OVER (PARTITION BY t.Client ORDER BY v.yyyymm) as prev2_yyyymm
FROM [dbo].[Транзакции] t CROSS APPLY
(VALUES (DATEFROMPARTS(YEAR(t.Date_transaction), MONTH(t.Date_transaction), 1))
) v(yyyymm)
GROUP BY t.Client, v.yyyymm
HAVING SUM(Litr) >= 50
) t
WHERE prev2_yyyymm = DATEADD(MONTH, -2, yyyymm);
If you want to aggregate by month:
SELECT yyyymm, COUNT(*)
FROM (SELECT t.Client, v.yyyymm,
SUM(t.Litr) AS Volume,
LAG(v.yyyymm, 2) OVER (PARTITION BY t.Client ORDER BY v.yyyymm) as prev2_yyyymm
FROM [dbo].[Транзакции] t CROSS APPLY
(VALUES (DATEFROMPARTS(YEAR(t.Date_transaction), MONTH(t.Date_transaction), 1))
) v(yyyymm)
GROUP BY t.Client, v.yyyymm
HAVING SUM(Litr) >= 50
) t
WHERE prev2_yyyymm = DATEADD(MONTH, -2, yyyymm)
GROUP BY yyyymm;
What is this doing? The inner query is basically your query with some refinements:
yyyymm is a date, more convenient than storing the year and month in the same column.
The CROSS APPLY is just a convenient way of defining the yyyymm alias.
prev2_yyyymm gets the second time previously when the client met the condition.
The outer query implements the logic: If prev2_yyyymm is exactly two months ago, then we have three months in a row that meet the condition.
Here is a db<>fiddle.