First, please pay attention to how I setup the test data. This is something you should do for us. I found it easier for this solution to use a table variable. You could use a temp table (#TestData for instance) or a permanent table (dbo.TestData for another example).

Second, notice when I wrote the insert statement for the test data I used the ISO standard yyyy-mm-dd for the dates. I also could have used yyyymmdd. Both of these are preferable as everyone should recognize this format.

Here is my shot at a solution:

/*

-- Format of dates is not stated based on anticipated results posted later

-- making the assumption that the following dates are in dd/mm/yyyy format.

Date method

01/05/2013 A

01/05/2013 B

01/05/2013 C

01/05/2013 D

02/05/2013 A

02/05/2013 B

02/05/2013 C

02/05/2013 D

03/05/2013 A

03/05/2013 B

03/05/2013 C

03/05/2013 D

07/08/2014 A

07/08/2014 B

05/09/2014 A

05/09/2014 B

*/

declare @TestData table (

MethodDate date,

Method char(1)

);

insert into @TestData

values ('2013-05-01','A'),

('2013-05-01','B'),

('2013-05-01','C'),

('2013-05-01','D'),

('2013-05-02','A'),

('2013-05-02','B'),

('2013-05-02','C'),

('2013-05-02','B'),

('2013-05-03','A'),

('2013-05-03','B'),

('2013-05-03','C'),

('2013-05-03','D'),

('2013-08-07','A'),

('2013-08-07','B'),

('2013-09-05','A'),

('2013-09-05','B');

select

td.MethodDate,

td.Method

from

@TestData td

cross apply (select max(MethodDate) MethodDate, Method from @TestData where Method = 'B' group by Method) ca

where

(td.Method = ca.Method and td.MethodDate = ca.MethodDate) or

(td.Method = 'A' and td.MethodDate < ca.MethodDate);