Find biggest order by client in Microsoft SQL Server - sql-server

Question

I am looking for some elegant solution to find preferred channel by client.
As an input we get list of transactions, which contains clientid, date, invoice_id, channel and amount. For every client we need to find preferred channel based on amount.
In case some specific client has 2 channels - outcome should be RANDOM among those channels.
Input data:
Clients ID | Date | Invoice Id | Channel | Amount
-----------+------------+------------+---------+--------
Client #1 | 01-01-2020 | 0000000001 | Retail | 90
Client #1 | 07-01-2020 | 0000000002 | Website | 180
Client #2 | 08-01-2020 | 0000000003 | Retail | 70
Client #2 | 09-01-2020 | 0000000004 | Website | 70
Client #3 | 10-01-2020 | 0000000005 | Retail | 140
Client #4 | 11-01-2020 | 0000000006 | Retail | 70
Client #4 | 13-01-2020 | 0000000007 | Website | 30
Desired output:
Clients ID | Top-Channel
-----------+-----------------
Client #1 | Website >> website 180 > retail 90
Client #2 | Retail >> random choice from Retail and Website
Client #3 | Retail >> retail 140 > website 0
Client #4 | Retail >> retail 70 > website 30
Usually to solve such tasks I do some manipulations with GROUP BY, add a random number which is less than 1, and many other tricks. But most probably, there is a better solution.
This is for Microsoft SQL Server

Answer 1

If you have the totals, then you can use window functions:
select t.*
from (select t.*,
row_number() over (partition by client_id order by amount desc) as seqnum
from t
) t
where seqnum = 1;
If you need to aggregate to get the totals, the same approach works with aggregation:
select t.*
from (select t.client_id, t.channel, sum(amount) as total_amount,
row_number() over (partition by client_id order by sum(amount) desc) as seqnum
from t
group by t.client_id, t.channel
) t
where seqnum = 1;

Answer 2

So keeping your desired output in mind, I wrote the following T-SQL without using group by
declare #clients_id int =1
declare #clients_id_max int = (select max(clients_id) from random)
declare #tab1 table (clients_id int, [Top-Channel] nvarchar(10), amount int)
declare #tab2 table (clients_id int, remarks nvarchar (100))
while #clients_id <= #clients_id_max
begin
if ((select count(*) from random where clients_id =#clients_id) > 1)
begin
insert into #tab2 select top 1 a.clients_id, a.channel +' '+ cast (a.amount as nvarchar(5)) +' ; '+ b.channel +' '+ cast (b.amount as nvarchar(5)) as remarks
from random a, random b where a.clients_id =#clients_id and a.clients_id = b.clients_id and a.channel <> b.channel
order by a.amount desc
end
else
begin
insert into #tab2 select a.clients_id, a.channel +' '+ cast (a.amount as nvarchar(5)) as remarks
from random a, random b where a.clients_id =#clients_id and a.clients_id = b.clients_id
order by a.amount desc
end
insert into #tab1 select top 1 clients_id, Channel as [Top-Channel], amount from random where clients_id = #clients_id order by amount desc
set #clients_id = #clients_id +1
end
select a.clients_id, a.[Top-Channel], b.Remarks from #tab1 a join #tab2 b on a.clients_id = b.clients_id
[Query Output
: https://i.stack.imgur.com/7RzcV.jpg ]

Answer 3

This will work:
select distinct ID,FIRST_VALUE(Channel) over (partition by ID order by amount desc,NEWID())
from Table1