DM: RE: Cost Savings with Data Mining?

Rebecca Buchheit asked:

> I'm preparing a grant proposal and would like to cite examples where
the
> use of KDD/data mining was able to produce significant cost savings.
I
> have done extensive searching on the web, but have been unable to find
> much that specifically cites how much (in dollar amounts, rather than
> just saying "we estimate that we saved a significant amount of time
and
> money") was saved.  Can anyone point me to a paper or resource about a
> case study that actually reports cost savings?  

I've had a go at doing this for a few projects I've completed (for the
same
reasons as Rebecca's - to build a firmer case for future proposals). I'm
going a bit beyond Rebecca's brief, but the point is the same...

While the information gain - in terms of both NEW and more PRECISE
signal from DM is (fairly) easy to identify, you have to go a few
further steps
up the organisation's chain of management and planning to quantify
savings. 

Firstly, there's the issue of implementation:  Where was the DM derived
signal factored into decision making, planning and measurement within
the client's business processes (including revision of business
processes).

Secondly, there is a serious issue of how business costs are attributed.
On one project (for example), there was such a diversity of ways of
costing, and of recording cashflows, that profitability calculations 
(immediate, or in the sense of net present value) could make ANYTHING
look good, to ANYTHING look dismal. Eg, status quo and downsizing
can be easily made to look attractive (status quo incurs low costs and 
downsizing generates a net cash flow as less new projects start - less
upfront costs - and more 'old' projects complete - more final payments).
Various NPV and opportunity valuations make new innovations look
attractive, depending on how risk factors into calculations. Where you
you put various overhead and infrastructure costs. ETC.

Certainly, one project we completed was audited independently, and
it was worth (in NPV terms) an estimated $AUD140M..$AUD200M,
which was MUCH more than the costs of the project. This was for
a FS client, and details cannot be released.

One other selling point for Rebecca would be that good predictive
and exploratory DM required good data quality. The project's value
has a multiplier effect in ensuring better data quality within the
organisation. However, there is often resistance to seriously
biting the bullet on DQ in many organisations (Oh God, they
may have to change some business processes and reporting/
measurement practices!!!).

I'm not saying it can't be done - we do it at times - but there are
many issues (not directly related to DM) that get in the way.

I'm very interested to hear other views.

Tom.


Dr Tom Osborn
Director of Modelling
NTF
Decision Support Consultants
Level 7, 1 York Street
SYDNEY NSW 2000
AUSTRALIA
phone:	+61 2 9252 0600
fax:	+61 2 9251 9894

• Prev by Date: Re: DM: Cost Savings with Data Mining?
• Next by Date: No Subject
• Prev by thread: DM: Administrative Note: Datamine-l Hosting is Changing
• Next by thread: DM: Postdoc-position in a european project at University Paris 13
• Index(es):
  • Date
  • Thread