Hi Ashwin,
Hooray...someone who believes that experience, real experience, is a solid foundation for doing estimates..... ;-)
When I was young I used to think 'experience' was a word the 'old guys'
bandied about so us 'young guys' could not claim to be able to be 'comparable' in some ways to the old guys....
Now I am an 'old guy' I can see what the 'old guys before me' were talking about....
You cannot put an old head on young shoulders....experience is to be earned...either through personal experience or though learning for 'masters'......
Fortunately I read 'How to Win Friends and Influence People' when I was young and I took the advice of 'learn from the masters' early on...and those who mentored me I owe a debt I cannot pay....but I have said Thank You.. ;-)
Example....In 1988 I wrote an expert system to calculate the development effort for what we used to call mid sized projects....approx 200 work months....I talked to a lot of people to build it.....and in the end...the way to figure out how long a project was going to take was.......
'Ask Chris Nicholson'.
No-one here will know Chris.....but he was, by a large measure, the most brilliant Project Manager I ever met, anywhere....ever!!! He is probably retired now.....it is no joke...you could go to Chris and say 'How long is this going to take' and he would ask a few questions and give you an amazingly insightful assessment of what to consider and how long it might take to do the project.
I find it amazing that in 25 years of IT I have never met another Project Manager, anywhere, on 4 continents, of his caliber.
If you want good estimates, ETL or otherwise, seek someone with experience.....real experience...not 'paper experience'....
And look for ways of reducing the costs of ETL sub-systems.......good ETL subsystems for large complex projects are very expensive to write and more expensive to maintain....and that amount can be surprising.....I am committed to 'disappearing' the cost of ETL such that all companies can afford BI...
Best Regards
Peter
>
> Hi,
> Well we also had tried to estimate the ETL effort using methods like
FP
> etc but we were proved that we were terribly off the mark.
Unfortunately
> in this case my experience has been to best rely on expert judgements
> and experience. It has worked with me time and again and continues to
do
> so still ..........fortunately............
> Regards,
> Ashwin
> -----Original Message-----
> From: Nic Harvard via dw-select
[
> Sent: Saturday, 7 October 2006 9:02 AM
> To: Ashwin Deokar
> Subject: RE: [dw-select] RE: How to best estimate the number of ELT
jobs
> needed in a project
> Errm.. Was not me that spoke about these tools. Perhaps a thread got
> truncated?
> I've never seen a tool that could do this, nor could I conceive of
one.
> There are a bunch of methods to estimate function points etc in SLDC
> processes, but I have never found they apply more than 30% to an ETL
> data-transfer / migration project.
> First, get your ERD of the source system.
> Then, get the business rules (either in front-end logic or business
> protocal/customs) by time period that put stuff there Then, get your
> target system import templates/logic/ERD/design/schema/adaptors
> (then prepare for lots of rounds of discussion, as you expose how crap
> the data in source/legacy systems is/are)
> Then get out pencil and paper, and consider what human skill resources
> you can access.
> Sounds stone-age, but best advice I can give...
> Nic
> From: MarkM via dw-select [
> Sent: Friday, October 06, 2006 8:58 PM
> To: Harvard, Nicolaus
> Subject: RE:[dw-select] RE: How to best estimate the number of ELT
> Hi, Interesting discussion. In addition to hearing any further
answers
> to
> your question, Nicolaus, I'd also like to request that you (or anyone
> else
> !)
> share what tools you have found that do the estimation once you
> determine these classifications? And what you think of their
> efficacy, if you have the time? Thanks
> Mark
> >Sent: Wednesday, August 09, 2006 1:14 AM
> >To: Harvard, Nicolaus
> >Subject: [dw-select] How to best estimate the number of ELT jobs
> needed in a
> >project
> >
> >I have discovered a few estimation tools that compute the amount
> of ETL
> >effort required to deliver a DW project based on the number of
> simple,
> >medium and complex ETL jobs that have to be designed, developed
> and
> >tested.
> >However, I have not found an estimation tool, paper, or rational
> for how
> >to estimate the number of simple, medium and complex ETL jobs in
> the
> >first place.