Balanced Scorecard

[Jonathan Siegel <jmsiegel@yahoo.com>]

Steve,

This issue relates to what I'm doing right now -- I
went back to being a statistical programmer (currently
on the Avastin team at Genentech) after leaving
graduate school, and before that I made a living
building division-sized data warehouses in a software
package called SAS and using them for analysis. (This
also means there's a potential conflict of interest in
what I discuss below) 

Bill Inmon criticized the whole datamart/Star Schema
approach virtually from the first time Ralph Kimball
began publicizing it. One of his criticisms was that
it over-summarizes and oversimpliies. For one thing,
it only shows data in fixed pre-existing ways, and one
has to be able to support reorganizing data in new
ways on-the-fly to be be able to detect patterns or do
something innovative and valuable. For another, it
tends to have too large a granularity -- it is
impossible to detect issues that require individual
transactions to see (e.g. the order of data over
time), since they are averaged out into a summary. 

Bill Inmon's "Exploration Warehousing" book provides
an alternative theory on how to organize data to
support problem solving and innovation. Parts of it
(by no means all) can be a little technical, but it's
well worth a read. 

SAS has become the tenth largest software company in
the country virtually under the radar screen. It's
doubtful many on the DEN have heard of it. It
essentially has a monopoly on the market for people
doing serious research involving large amounts of data
from diverse sources. That market may be less well
known than splashier marketing-database purveyors, but
it's quite large. In addition, it's a closely held
company -- two founders own all the stock. And it has
very little employee turnover due to extremely
paternalistic employee policies, from in-house doctors
to in-house Montessori schools, and for the janitors
as well as the programmers. 

SAS's problem has always been that it's expensive and
it requires expertise to use it. So it's never been
much of a hit with people who insist on instant
pudding. This difficulty is often a weakness of the
product. I've long felt that its products are
unecessarily unweldy and could be made clearer and
easier to use. But it is also an inherent part of the
problems the product is trying to solve. Any toolkit
(as opposed to a plug-it-in appliance) has similar
problem. 

I should add that SAS Institute's marketing group has
recently began pushing its ability to calculate a
balanced scorecard as a selling point. The statistical
package often takes a distinctly non-Deming approach
to statistics, including the way its QC package
interprets control charts. I'm certainly not
suggesting that everything in the product, or
everything the company does, is something Deming would
approve of. 

Jonathan Siegel