Name Change SQL Query - Positive Attrition - sql

Question

Using SQL Server Management Studio and I've got an Employee table which has records for every time something changes with an employee, be it manager, pay-scale, etc. etc. etc. Whenever a change is made, and EffectiveEndDateKey will be marked for the last time that complete field was relevant and the next record will have the next day as and EffectiveBeginDateKey.
What I'm trying to do is extract the last record an employee had BEFORE they changed their job title in the last month. The goal is IF an employee changes their job title within a given month that would count as "Positive Attrition" and we're trying to figure out how much positive attrition we get in a given month. (It's always for the previous month so in the where statement I'm pulling just changes for the previous month).
Take a look below:
In this example, on July 4th, John Doe went from being an Apple Store Clerk to the next day being a manager, so there was a job title change. What I want is to pull the record in the red box (Jon's last day - in July - when her was an Apple Store Clerk before he became a manager) b/c that tells me that an EffectiveEndDateKey had a change that resulted in a job title change.
So the where statement is going to have a Cast in it to convert the EffectiveEndDateKey to a date and then look at last months data, pulling only records that have EffectiveEndDateKeys from last month (July) and what I need help with is the part where those records must ALSO have a different job title.
if say someone changed job titles on July 31st (so their new job title/EffectiveBeginDateKey was 20130801), that would still count as 1 positive attrition and we'd want to pull the last record from July 31st.
Any thoughts?

Answer 1

You can do this with a self-join:
select eprev.*
from Employee e
Employee eprev
on e.EmployeeId = eprev.EmployeeId and
cast(cast(e.EffectiveBeginDateKey as varchar(255)) as datetime) =
cast(cast(eprev.EffectiveEndDateKey as varchar(255)) as datetime) + 1
where cast(cast(e.EffectiveBeginDateKey as varchar(255)) as datetime) >= dateadd(mm, -1, getdate()) and
eprev.JobTitle <> e.JobTitle;
The key here is the conversion of the number to a datetime. The format YYYYMMDD is easily convertible, when it is a string. So, convert the number of a string first, then to a datetime. The rest is just the mechanics of the join.

Answer 2

Since your strings are effectively just dates in ISO format (yyyymmdd), you don't even have to convert it to datetime, you could just get previous row:
select E.EmployeeID, E.JobTitle, ENEXT.JobTitle as NextJobTitle
from Employee as E
outer apply (
select top 1 T.JobTitle
from Employee as T
where
T.EmployeeID = E.EmployeeID and
T.EffectiveEndDateKey > E.EffectiveEndDateKey
order by T.EffectiveEndDateKey asc
) as ENEXT
where
E.JobTitle <> ENEXT.JobTitle and
E.EffectiveEndDateKey >= convert(nvarchar(8), dateadd(mm, datediff(mm, 0, getdate()) - 1, 0), 112)
see SQL FIDDLE example