Hi everybody,

who can tell me the statistical allocation of costs (percentages) for data
warehousing in a structure like this:
1. Hardware
2. Software
3. Professional Services (external consulting)
4. Internal Labor

Thanks for any help!

Michael

Michael Weimar wrote:

This is very hard to tell. Also, the plain numbers would seem rather
meaningless to me since these four cost items depend on different drivers.
Let me give you a few ideas on cost drivers:

depends mainly on amount of data, complexity of real-time calculations and
number of concurrent users. Make this item a minimum of 5k Eur. With
today's hardware I would not want to spend more than 40k Eur on hardware,
including some peanuts on hardware for ETL.

Which software? Strictly speaking, a few licensed seats of your favourite
relational database system will do. How much this will cost you largely
depends on which database you pick and who you are. In case of Oracle, it
will also depend on the phase of the moon. Another 1k to 15k?
Depending on where your data come from, a decent ETL tool might be well
worth the 2-5k it's gonna cost you.
What next? Is it sufficient to have some fine clean data in a relational
pot, or do we want multidimensional OLAP bells and whistles? Need to slice,
dice & drill around in the guts of your accounting? Throw in another 2-25k
on OLAP software. The better the sales rep the more youl'll end up paying.
After all, you need some front-end client software for your users. You'll
probably want to mix a web front end for the masses (3k) plus a few
managing, power & design-enabled workstation desktops at 1k each.

This one's easy. It will cost you 50% more than you planned for. It will
cost you just 5% less than you could pay without going broke immediately.

This is the key to success. Start here before you start to think of
Hardware, Software or external consultants. Define and put down on paper
what you want to see in the warehouse, how you want it presented and where
the information should come from. Build a prototype in MS Access. Yes,
Access. You don't need no whizz-bang mainframe to find out your data is
garbage. And you don't need an external consultant to host a discussion
between marketing and sales about how a "customer" is to be definded.
Of course the amount of manpower needed depends largely on the issues you
want to target with your warehouse. Some basic accounting stuff should be
covered with some templates plus a couple of days of customisation. If your
accounting is done in a particularly creative manner or if you want to
cover some process figures from production, logistics or CRM, you might
need some more effort as you won't find any templates that really suit your
case. (According to some vendors, their ready-made business templates fit
any case. Go figure!)
So, labour can be from 15 to 1500 days of which as much as possible should
be spent inhouse, depending on skills available.

As you can see, most of these items are highly variable and depend on the
individual circumstances. They all scale to different factors, so any
statistical percentage ratios seem rather meaningless to me.

However, if you are mainly interested in some figures for a presentation,
maybe Gartner has some, ask Google.

cheers
Heimo

--
l'essentiel est invisible pour les yeux.

Michael Weimar <miki_wei2000 (AT) yahoo (DOT) de> wrote:
That's the wrong way to start thinking about a data warehouse.
What quantifyable benefits can you achieve with a datawarehouse?

Stephan