|
|
Business Management Book Store > Business Management books beginning with S
|
Super Crunchers: Why Thinking-by-Numbers Is the New Way to Be Smart |
Author: Ian Ayres
Published: 2007-08-28 |
List price: $25.00
Our price: $16.50
|
Usually ships in 24 hours
As of: November 21st, 2008 01:41:50 PM
|
|
|
Customer comments on this selection.
 Intuition vs. Data...and the winner is...
Neither. I think. Super Crunchers is a fascinating study of the ascension of data analysis and statistics in decision and policy-making in all realms of life, from business to government to health. Ayres shows us how the ability to collect millions upon millions of data points and number crunch them to study trends, analyze relationships and make predictions, has created a schism between professionals (lawyers, educators and doctors) for example, who use their experience and expertise with intution to come to decisions, and data crunchers, who us the power of computers to do the same. Ayres uses significant case studies to demonstrate how super crunching consistently beats intuition.
The caveat is that data crunchers are also human, ultimately, and may make errors in selecting the inputs and variables that are analyzed. In addition, Ayres warns against the possible misuse of super crunching to, for example, create pricing systems where every consumer would be charge a customized price for a product based on a super crunching analysis that predicts how much each unique consumer is willing to pay.
Nevertheless, Ayres demonstrates the decision making and analysis power of super crunching with examples from medicine, law enforcement and education. The message is that, in sum, super crunching is good and is something that we should all become familiar and comfortable with because it will not go away and will, indeed, be a strong complement to intuition.
The end of the expert?
The gimmick in the TV show Numbers--and all crime shows have to have some sort of gimmick--is that a genius mathematician is able to help the FBI solve crimes. He particularly does so by finding patterns amongst the haze of large data sets. Ian Ayres's book Super Crunchers is a non-fiction look at a similar idea: the trend to find patterns and make predictions from analysis of large amounts of data.
The two principal ways that this is done--outlined in the first two chapters--is through regression and randomized trials. Simply put, regression sees how things have performed in the past and tries to extrapolate into the future; it's like plotting points in a graph and finding the best line or curve that fits that line. With randomized trials, you take a random sample from a population and see how these sample members react to a certain situation, such as a new drug or advertising slogan.
Of course, as Mark Twain is alleged to have said, there are lies, damn lies and statistics, and the same holds true with this "Super Crunching" of data. It only works properly if you know what you're doing and you're doing it properly. A well-known example (also cited in Freakonomics, which is kind of a companion piece to this book) deals with John Lott, the author of More Guns, Less Crime. According to Ayres (and the Freakonomics writers), Lott's crunching of the numbers had some serious errors, which when corrected, showed that the data actually contradicted his conclusion that the more guns owned in a community, the less crime took place. Did Lott purposely skew information to satisfy his own agenda or were these accidental mistakes (or did Ayres)? The reader can make his own judgment.
Beyond any issue with a particular study, the more relevant issue is the increasing prevalence of Super Crunching data and the effect it is having in determining business decisions and government policy. As cited early in the book, if the numbers are worked properly, a computer can do a better job predicting the performance of a wine or baseball player than a wine expert or a baseball scout.
In fact, often times the databases will provide better insight than experts. What does this mean? Will so-called experts go away? Will more and more decisions be based solely by computer with little or no human input? Is this a bad thing? Once again, the reader can make his own judgment as to whether this is a good or bad trend, though it is certainly an increasing one.
All this would mean little to the reader, however, if the writing is not good; fortunately, Ayres is successful at describing what Super Crunching is and what its possible ramifications are. You do not need to be a math or business expert to understand this book. And if Super Crunching is the future (even more than the present), it makes sense to read this book and get a glimpse of things to come.
The End of Intuition
The author explains that he originally intended to title the book, "The End of Intuition". I think that would have been the better title. Interestingly, the name change resulted in a clever use of Google's adwords. The problem with the title Super Crunchers is that it makes the impression that the book is about data mining huge data sets. I have read one scathing review in the KDD community that makes this point, but data mining is not the focus of the book. If you are looking for a book about data mining, it is a poor choice.
The theme of the book is how data, and data analysis, is making inroads in industries and firms where it has not been emphasized in the past and how the use of data to make decisions is often fiercely resisted. The techniques will seem mundane to professional analysts. I won't be recommending this to many of my data mining colleagues - there is nothing really new here. However, this book does have its place. I often meet analytically oriented managers that are trying to introduce advanced analytics into departments (or entire firms) that are not accustomed to data based decision making. It doesn't matter that Ayres' examples center on ordinary multiple regression, or standard randomized trials. This is where the battle is being fought and won in many organizations. The reluctance of some doctors to embrace "evidence based medicine" made for compelling reading.
The examples Ayres uses are diverse and unusual: wine rating, medicine, welfare reform, prison sentencing, dating web sites, Capital One and others. Some are much better than others. I didn't learn anything about data mining or analysis while reading this book. But I did learn something about people, and it confirmed some things I have learned about trying to deploy models in organizations. So while this book is not without flaws, if you want to better understand why there is often so much reluctance to use data to drive decisions it is worth a read.
 Entertaining, but far from super
This is an easy and mostly entertaining read. The author uses many anecdotes to
persuade us that statistics can be a useful tool for decision making. Some of
the described applications use lots of data and multiple regression. Those are
easier to do now than they used to be, because more data is collected and kept.
Some are trivial. If your company hurts a customer, apologize. You might get
some ideas of thing to do that might help your organization. You will not get
any detailed help about how to implement the improvement, but there is a good
chance there is enough information that some systems person can figure out what
other skills are needed to make the idea work.
There is some discussion of limitations on the methods, and some warnings about
potential abuse, but not enough. Ayres seems to confuse correlation with causation.
He also frequently assumes the sample is representative of the population.
Even when trying to make the sample representative, it often is not. He also
assumes the answer is in the data. Sometimes it is not. Ayres reports a study
concluding widespread point shaving in college basketball because a distribution
at game end did not match the distribution five minutes earlier when a highly
favored team was ahead by about the spread. I have no opinion about the conclusion,
but the simpler explanation of the coach thinking it was late enough to safely
let the weaker players participate more was not considered.
Regression is a powerful tool, but it is easy to misuse. For an ongoing
survey of misuses, see junkfoodscience dot com, a blog. Many of the entries show
the flaws in statistical claims of medical trials. Also try stats dot org.
What you can do with large datasets
The answer is of course: a lot.
And Ian Ayres' book will tell you a little about it.
Supercrunchers are those who use lage datasets
to find patterns in human behaviour, and
predict the future based on these large datasets.
The book informs us that super crunching is on the verge of being
used all over. E.g.
Chess grandmaster Kasparov was no match
for IBMs Deep Blue chess computer,
that stored some 700.000 grandmaster chess games to help find the
winning move.
The IRS could use its data to tell a small business,
if it is spending too much or too little on advertising.
Indeed, the IRS probably has enough data to
make good estimates on whether business, marriages, etc. etc.
will fail - based only on comparison with its existing dataset.
For the paranoid, it is a horror that supermarkets could map your life cycle and predict your next purchases pretty accurately (based on
what other similar customers did).
For the optimist data mining is a good thing and we'll all lead better lives because of it.
Want to write a bestseller about it? Compare your title and some key words with data from a database of books, titlescore.com, containing millions of bestsellers and flops, and you will get your answer.
It all seems pretty straight forward, and the book has some nice examples of what we can expect in the coming years.
-Simon
|
|
Our Business Management book picks:
|
|
Search the Business Management Products Store
LCS Amazon Store 2.5 © 2008
|
|
|