You haven't posted your table structure, sample data, column datatypes, or desired results, so it is hard to tell exactly what you are after.

It is best to test something like...

declare @StartDate datetime
set @StartDate = '20070718'

Select
 *
from
 MyTable
where
 -- Greater than or equal start of month
 MyDate >= dateadd(mm,datediff(mm,0,@StartDate),0) and
 -- Less than start of next month
MyDate < dateadd(mm,datediff(mm,0,@StartDate)+1,0)

This wiil produce a random distribution of -0.05 to +0.05

select A = ((5-((abs(convert(bigint,convert(varbinary(8),newid())))%11)))*.01)

If your objective is to increment by a random value between -5 through +5, it does not do that.  Your algorithim does not produce a random distribution, and the range is only...

What more could I really say about them?  They seem fairly self explanatory in the links I posted on SQLTeam.

I did notice that I posted the link...

If the integer you want to convert to is UNIX time, this will do it:

select datediff(ss,'19700101',MySmallDateTime)

The functions in this script can be used to convert to/from SQL Server date time...

If you need a higher number of decimal places in your random number generation, this will give you 12 decimal places in the range of 0.95 to 1.05

I think that backup/restore is a better method to use for database migration from SQL 7.0 to 2005.

I don't see how detach/attach could cause the problem you described.

I posted a much simpler way to get the last day of the current month in my prior post.  It also has the advantage of removing the time part of...

If your condition for the last week of the month is as simple as finding the first day of the week containing the last day of the month, the code...

You can't do an update in a function.

This should do it:

select * from information_schema.columns where DATA_TYPE = 'nvarchar'

This will do it.  You won't have to ask people to leave the DB, because it will kick them out.

use master
alter database MyDatabase set read_write with rollback immediate

Different subject, but it is usually bad idea to use NOLOCK hints, because you can get data that is uncommitted or may eventually be rolled back.

I think this conversion method is a little shorter:

select

 run_date,
 run_time,
 [Job Run Date/Time] =
 -- Convert run_date and run_time to datetime
 dateadd(ss,((run_time/10000)*3600)+(((run_time/100)%100)*60)+(run_time%100),right(run_date,8))
from
 (
 -- Select sample of 20 jobs
 select top 20 * from sysjobhistory order by...