Enterprise Analytics - Review

"Lectures, Meanders, Pontificates, But Does Not Educate"

"Enterprise Analytics: Optimize Performance, Process, and Decisions Through Big Data (FT Press Operations Management)", Thomas H. Davenport, et al.
2 stars
(AmazonMy Amazon review, Kindle US, Amazon UK, Kindle UK, Flipkart)

Or, how a book on Big Data, Enterprise Analytics, and technology can neatly skirt any meaningful discussion of Big Data, Enterprise Analytics, and technology.

While a few chapters stand out for their reasoning and clarity, what is jarringly absent from this book is any meaningful, technical discussion about Big Data itself. Without such a discussion, most of the book's content can be recycled with minimum effort ten years from now and applied to the next big thing in technology. Even assuming that this book is targeted at decision makers and so-called C-level executives, an absence of the nuances and complexities of Big Data mean that executives will be as clueless on that dimension of Big Data knowledge after reading the book as before. If you are responsible for selling sausages, you had jolly well get a look at the sausage factory, if not work there a day.

Big Data, Unstructured Data, the Cloud - if these three buzzwords were not enough, you can add the salsa-ish phrase SoLoMo - i.e. Social, Local, and Mobile, to the mix. Businesses, consultants, enterprises, anyone who is anything in technology wants to know more about what this buzzword alphabet soup is, and how to make sense of it before their competitor does, or worse - a disruptor.

The hope is that if the decision makers, the corner-room occupiers can understand this, they will be better able to drive a coherent process and structure within their organizations to take advantage and benefit. Hence this book.

The book is a collection of eighteen chapters, divided into five parts - the first is an "Overview", "Application", "Technologies", "Human Side", and "Case Studies" of Analytics. Each chapter is written by a different author, with a total of fourteen authors in the fray. A few chapters have been written by the editor himself, Thomas Davenport, and these are among the standout chapters - for their clarity and organization.

Since I work in the technology of analytics, I should be excused for either taking too technical a view of things, or for being too harsh in my criticisms. Having said that, there are at least quibbles that in my opinion leave this book only a middling, mediocre effort, and not one that will be remembered or consulted, much, if at all.

- The close and financial collaboration of at least a few technology vendors with the International Institute of Analytics means that most of the specific examples cited in this book are where the technology vendor's solution was used. Fair enough, but it leaves an aftertaste of an advertorial in the reader's mind.

- When discussing web analytics, there is a lot of ink devoted to the topic of "page views". Page Views are still relevant, but they are becoming increasingly obsolete in the world of AJAX - where parts of a page and its contents can be updated without having to reload the entire page. Web analytics metrics operate at the least granular level of pages, and hence cannot capture a significant chunk of user interactions and engagement that occur on pages and sites that make heavy use of such asynchronous page content refreshes (AJAX does stand for "Asynchronous JavaScript and XML", and no - it does not contain the buzzword "Agile"). More sophisticated measures of user engagement are being built that track more than simple page views. When the chapter's author fixates on page-views without once mentioning the inaccuracies of measurement that AJAX can inject, the credibility of the chapter suffers.

- The chapter on "NBOs", i.e. "Next Best Offers" takes several cheap shots at Amazon (see page 90), which left me wondering whether Amazon had not turned down six-figure consulting offers from either of the authors to warrant this broadside.

- There is a laboured chapter on "engagement" - an attempt to define a compound measure based on essentially a summation of basically arbitrary weighted base measures. For instance, a measure of online engagement is a summation of eight different indices. Putting a "sigma" symbol in the equation makes it look impressive, but in the end it is more arbitrary than methodical. Because decision makers need to have information supplied to them in a simple manner, it is often supplied to them in a simplistic manner, cloaked in technical-sounding phrases.

- Privacy is becoming an increasingly sensitive and relevant issue as more data is collected from customers and users, often without their knowledge, and sometimes without their consent, almost always without providing users a clear picture of what is done with that user-data so collected. Privacy is an important topic in this discussion on enterprise analytics. And it is given short shrift in the book. After a cursory nod, almost as an afterthought, to privacy concerns, there are examples cited that are almost creepy in the extent they suggest the invasion of a user's privacy. Sample these: "The next generation of video game offers could have pictures of your friends or your tastes and interests built right in." Or "A company called Sense Networks has developed an application to help infer a person's lifestyle based on his or her location history." Harvesting a user's location and web-click information should require explicit opt-in - it is basic respect for human decency. Take the section where the authors talk about collecting data by anonymizing data. It has been proven that even after anonymizing data, it is possible to individually identify users with a very high degree of accuracy based on only a few attributes. There is no such thing as truly anonymous user-tracking on the web.

- Then there is this most curious statement that states - "Despite all the hype around the unstructured data component of "big data", it seems that structured data still rules the in predictive analytics." Well, yes! Unstructured data is fairly recent, especially when compared with structured data, that has been around for literally decades. It is but natural that the use of unstructured data in predictive analytics will take time to gain traction, especially as the technology and means of blending structured and unstructured data evolve.

- Even the chapter, "Predictive Analytics in the Cloud" contains phrases that make absolutely no sense, other than to bump up the chapter's jargon-index. Sample this: "These cloud-based solutions inject predictive analytics into other software that is cloud-based or delivered as SaaS." Is a specific example too much to ask? "inject" is an impressive-sounding word, especially if you have heard of the phrase "sql-injection" in the context of hacking attacks, but just what does the word "inject" mean in the context of predictive analytics and the cloud? And why does this injection require the cloud? Can it not be done with more traditional, hosted solutions? And what exactly are "cloud-based dashboards"??? Is any dashboard served via a browser "cloud-based"?

I could go on an on, but a short summary of the book would be this: each chapter suggests and promises value, but falls short.
Kindle Excerpt:

  © 2012, Abhinav Agarwal (अभिनव अग्रवाल). All rights reserved.