Customer Reviews

The Numerati

5 star:		(22)
4 star:		(12)
3 star:		(5)
2 star:		(9)
1 star:		(4)

 

 

Stephen Baker, a technology writer for Business Week, takes us into the world of data miners, forecasters, and matchmakers. The math whizzes who analyze our blogs for trends, create the ads that make us eager to buy, and analyze the chatter that could conceal signs of criminal activity--these are the Numerati. Baker gives us a chapter each on work, shopping, politics, spy vs. spy, healthcare, and even [...] (What does the length of your ring finger have to do with the kind of person you're attracted to? Read and find out.)

Some of it is "house-of-the-future" stuff--imagine, for instance, a floor tile that will alert the doctor when your aging parent's gait seems more hesitant than usual. According to Baker, experts watching old reruns of Michael J. Fox shows can detect characteristic signs years before he was diagnosed with Parkinson's disease.

And then there's the political game. With ever-more-insightful analysis, political math mavens have found that (thank god!) America is nowhere near as polarized as you would expect. Many a liberal Democrat lurks in the McMansion suburbs, and vice versa. But politics is tough--your grocery basket doesn't lie, but nobody wants to give the time of day to a pollster. How they craft the exact political messages that will get you to the voting booth might, oddly enough, be related to your shopping habits.

Shopping--now this is a chapter that should be of interest to every die-hard Amazon fan. Sophisticated algorithms designed to deduce your taste in novels or music can be frighteningly accurate (or, as my Quick Picks occasionally remind me, maddeningly stupid, but that's the topic for a different book). After finishing this chapter, I could think of half a dozen things my grocery store knows about me that I never told them. If they chose to sell their data to magazine publishers, say, we would surely be targeted for the cooking mags ("Look, this family buys at least four units of different fresh herbs a week, and their weight in extra-virgin olive oil every month"). They can tell we have a teenager in the house ("Lots of Clean&Clear products") and could probably guess how old within a year or two ("Look it up--when did they quit buying diapers?"). Any health insurer would be interested in knowing that we spend a lot in produce and seafood, and very little at the meat counter--but what about those frequent trips to the candy aisle? It's a false positive, I swear--they're for the snack bar at my office!

You should be a little frightened, and more than a little fascinated, by The Numerati.

[Edited to add: For a more detailed look at the doings of one of the Numerati, take a look at Click: What Millions of People Are Doing Online and Why it Matters, by Bill Tancer of Hitwise.]

I became interested in this book after reading the companion cover story in BusinessWeek. Although the stories and interviews were interesting, I thought the book fell short on connecting the math beyond the most basic concepts.

Baker admits he was a liberal arts major in college and doesn't pretend to fully understand the math behind the analysis. Obviously, an in-depth mathematical discussion would have been beyond the grasp of most readers and presumably the author. However, a little more detail on the methodologies beyond the simplistic descriptions would have given the book more substance and utility.

Data Mining and Data Warehousing have been around for many years. Retailers have used it extensively to understand their customers. Yet, Baker fails to discuss these established practices and compare them with this new emerging area.

Baker spends most of his book describing the people he interviews in a series of stories. The book is an easy read and is entertaining. If you read for entertainment and are interested in this subject, you will probably like this book. However, if you read for knowledge and are looking for a good, informative business book on this subject, it may disappoint you.

I would highly recommend reading Baker's book immediately before or after reading How to Measure Anything: Finding the Value of "Intangibles" in Business by Douglas Hubbard. Baker would probably consider Hubbard one of the "numerati". Both authors talk about some of the specifics of the analysis methods (but moreso Hubbard) and both talk about the general trends and impacts (but moreso Baker).

Like his table of contents (which is simply worker, shopper, voter, blogger, terrorist, patient, lover), Baker's book is sweeping if a bit terse in places. As a quant, I find Numerati an easy read with virtually no math but still enlightening even for the most quantitatively adept reader. There were several examples in Baker's book where I already knew of the mathod but had not heard of that application. He did some great research and covered a lot of topics in this giant and elaborate field of work.

My main concern for many management-level readers of this book is that in some cases Baker gives a reader just enough information to think they can apply it to a similar problem they have, falling into the "a little knowledge is a dangerous thing" trap. Again, this can be offset with a read of Hubbard's book. It might also have been helpful to talk about the rise of "crackpot rigour" in a world with lots of data and relatively few competent mathematical analysts (various "data mining" experts come to mind).

In all, its one of my favorite reads of the year. I felt like someone was finally casting light on my own obscure field.

Quantitative profiling of human behavior ranges from the beneficial (recommendation engines for books and movies) to the scary (employer and police monitoring), and everything in between. "Numerati" provides a journalistic introduction to this topic, that is easy to read and understand. I found it way too simplified, though:

1. The author treats this technology as a "black box" which makes it seem almost miraculous to the uninitiated reader. The first requirement in writing about any technology is to explain what it can and can't do; the book does not provide enough information about this.

2. Like all technology it has both good and bad uses (and most uses are good in some ways and bad in other ways), but the book does not provide enough information about the social and policy tradeoffs inherent in its development, use, and regulation.

In summary, the book provides a readers with a very basic introduction to the brave new world of statistical profiling, but doesn't explain enough about the technology or its consequences to be really satisfying.

It's extremely rare that I find a book so bad, so devoid of value that I can't even finish it. Even the worst book has some redeeming merit -- even if it's only to understand the perspective of its author. Numerati, then, joins a list so short I can count its members on one hand: books that are simultaneously so annoying and so devoid of value that I couldn't get through them.

The biggest reason I couldn't finish this book is that I was never convinced of Baker's fundamental thesis: that the "numerati" are a class apart, different in kind from the rest of us. This flawed idea seems to go hand in hand with the idea, commonly perpetuated by the mass media, that mathematicians in general possess some mysterious skill denied to others. I disagree. I think that anybody can be numerate and, with the tools and data that the Web increasingly make universally accessible, anybody can be their own "numeratus".

The people in this book are different from the average internet user in degree, not in kind. They may be as different from us as Tiger Woods is from the average weekend hacker... but, like Tiger, they are playing the same game with the same tools. Baker's awe-filled, almost worshipful descriptions makes it seem like the "numerati" are separated from "normal people" by an uncrossable chasm... and to me that is not merely wrongheaded, it's dangerous. It encourages the rest of us to abdicate control of our futures to this mythical over-class.

Even those flaws wouldn't be enough by themselves to make me set this book aside -- it would still be interesting to understand this apparently common attitude that I don't share. The last straw for me was Baker's irritating insistence on inserting himself into the narrative at every possible moment. I really don't care what block of what street he had brunch on, or how many times the interviewee's phone rang, or how long he had to wait for the elevator afterwards. The whole breathless narrative as if he were participating in a cross between Raiders of the Lost Ark and the Da Vinci Code, rather than conducting interviews with people who happen to be busy multitaskers, just struck me as annoyingly pompous and self-centered.

Yes, I know I'm going to get a lot of "unhelpful" votes -- negative reviews always do (for some reason, people who liked a book find it hard to tolerate people who didn't). But it's worth it to let people know: if you are in any degree numerate yourself, you're probably going to find "Numerati" intolerable.

The science of data-mining is gaining in sophistication, but don't look to this book for any real understanding. Baker has written a book containing very little actual information content. He does not even attempt to convey how these techniques work or what their limitations are. Instead he paints a picture of a sinister and not-too-human "Numerati" that is handling our data while spurning basic social skills. It's a comic book plot that takes the place of any actual factual information. All you come away with is the idea that Baker is scared of what mathematicians are doing. 90% of the book is fluff.

I am a Data Miner and a Data Mining software trainer. I really enjoyed this book. My audience is not made of full time data miners (usually), but rather data analysts who are new to data mining. They are not experts who develop their own technologies, but rather are purchasers of software who use the software to Mine data with the most established of the techniques. They report to managers, and have co-workers, that are also new to data mining, curious about what it is, anxious to clarify its definition, wondering about possible application areas, working to effect the bottom line.

I disagree with those reviewers that suggest that Baker does not understand the methods. He tries very hard at explaining them to a general audience. I have tried to do the same, and he does an admirable job. I would be less than candid, if I didn't admit that I winced once or twice. But it is tough to strike the right balance.

I think he has does a brilliant job at choosing a structure of his book. His chapters are not structured to be on the different methods. One negative review claims that he `mentions' Support Vector Machines, but never mentions cluster analysis. It is simply not true. The "Voters" chapter is dedicated almost in its entirely to cluster analysis and market segmentation. I have taught this topic to thousands of software users, and I think it is a great chapter. He avoids technical jargon including the names of the methods because he audience is not me. His audience is everyone who does not do data mining, but is potentially effected by it, curious about it. So, if you want a chapter on how Support Vector Machines work, one can barely imagine a less appropriate choice. If you want a gentle introduction to many techniques like Neural Nets and Classification Trees go with Berry and Linoff's Data Mining Techniques. Data Mining Techniques: For Marketing, Sales, and Customer Relationship Management For an advanced discussion, many of the negative reviews have additional suggestions. I give LaRose's trilogy mixed reviews, but you can consider them. Discovering Knowledge in Data: An Introduction to Data Mining is the first.

Instead, and wisely, he structures his book as a handful of chapters on application areas. Each chapter involves interviews with experts, the Numerati of the title. There is usually one primary lead character in each chapter. Does he `gush'? Is he informal? A little perhaps, perhaps more than I would have liked. I like my Data Mining books technical, succinct, algorithm based, and cooly academic in style. But does a general audience want that? I like my astronomy books to be gentle on me. I like my physics books to be gentle on me. Why wouldn't a non-data miner want a gentle introduction?

As one critic noted, he does "insert himself into the narrative". He chats about the room in which he is interviewing. He mentioned Starbucks, and the physical description of his subjects. Some would leave that to Dickens and leave it out of this book. I might have cut back on that a bit, but incompetent? No. Not by a long shot. This is a great non-technical introduction to an important societal trend. I actually learned a fair amount. For instance, the concept of `next friend' and its use in social networks was new to me. I have also ranked this book higher than "Supercrunchers". Super Crunchers: Why Thinking-by-Numbers Is the New Way to Be Smart I believe the author of that book, while a fine analyst, just doesn't get it when it comes to Data Mining and where it is taking us. Davenport's book Competing on Analytics: The New Science of Winning, reads too much like Harvard Business Review for a general audience, and it's kinda dry. Baker's book, in contrast, nails it. I recognized my future in his book. Read it, and you will understand what Data Mining is attempting to do, and the challenges it faces.

As the chief marketing officer of Fair Isaac, I work daily with many of the Numerati that Stephen Baker so elegantly describes. However, I do not subscribe to the idea that all quantification and normalization of consumer data carries a nefarious purpose. Indeed, the societal value and impact of a given technology is generally determined by how this technology is applied and not by the technology itself.

As an example, Fair Isaac applies statistical analysis to the granting and administration of loans by establishing a standard credit scoring service (www.myfico.com). This quantitative approach, as pointed out by Stephen Baker, enables equal opportunity banking by not discriminating on the basis of anything but numbers. Another benefit is the empowerment of consumers who can now directly monitor the health of their credit by subscribing to this service. It may even entice someone to go on a credit diet!

In a global and connected economy the usage of well designed algorithms applied across vast data sets can help greatly in improving transparency and accountability.

"Numerati" is an unfocsed discussion of how different speciliasts use mathematical techniques in order to characterize people for purposes of identifying customers/ terrorists/ productive workers, etc.

This book could have been written in two ways that would have been effective (and the author chose neither). The first would have been to choose some limited number of mathematical techniques/ subspecialties and see how they were used to solve some practical problems. The second way would have been to look at some practical problems of some companies and then go backward and see how they used certain mathematical techniques in order to solve them.

As it is, this book tried to read like a novel (detective or otherwise), with detailed descriptions of the people to whom it was referring. Somehow, it just didn't work.

Another thing that I noticed was that the notes/ references for this book fit fully onto 9 pages (and the references were such that almost all of every footnote was a commentary and not something that pointed to an actual journal paper). The sources *and* further reading fit onto just over one half of one page. The conclusion that I came away with was that this book was fairly thinly researched. I counted one single journal reference, and that was from a 1957 publication.

An important question that is not answered is: What does this book describe/ demonstrate that is NEW? Are the techniques that he is (not really) introducing us to things that mathematicians/ industrial engineers have been using all along and that they are only refining? Or are they totally different techniques?

There really might have been enough here to make a good book, but the author just didn't know what he was doing in order to achieve that aim. He could use some help from a Thomas Sowell or Malcolm Gladwell. For something that is similar to this, but a lot better, read Super Crunchers: Why Thinking-by-Numbers Is the New Way to Be Smart by Ian Ayres.

It's worth about the price of a secondhand book, but not of the time that it takes to read it.

As a Data Miner I enjoyed reading this book. It was fun for me to identify the actual techniques being discussed before he named them. Then again Baker didn't name many of the techniques and kept all the math out to appeal to the audience.

However, I think he seriously overestimates our ability to predict mass and individual behavior, now and in the future. I've built models with two thousand observations and some with over 20 million. I used 8 to 10 variables to starting with a couple thousand.

The fact is, statistical modeling will never have the level of precision Baker suggests is comming. By its very nature a statistical model requires an error term.

Short of taking courses in multivariate analytics, you can get an idea of why more data won't increase our predictive power in laymen's terms by reading this blog post.

http://www.broadcastthoughts.net/2009/04/predictiveness-versus-parsimony.html

Even with that critique I think Baker did provide a good view to the many areas data mining can help inform intelligent decisions.