Creating A Data-Driven Organization

Author: Carl Anderson
Publisher: O'Reilly
Pages: 302
ISBN: 978-1491916919
Kindle: B012UDK3KG
Audience: IT managers
Rating: 4
Reviewer: Kay Ewbank

If you need to explain to someone how data can be useful, this book shows how organizations should use data to make decisions. 

Most developers working with data will have experienced managers (or even entire companies) who think that generating a lot of reports or having lots of data dashboards means they've cracked the whole big data thing. In this book Carl Anderson looks instead at how data can be used more effectively.


The book is essentially a conversation with people who don't understand, showing them how reports are backward looking, often without context or any attempt to explain why things happened (or didn't happen). Instead, the author gives examples of how an organization can work towards producing models that make recommendations and predictions for the future.





Anderson starts by explaining what he means by data-driven, looking at data collection, contrasting reporting and analysis, and looking in more depth at the different types of analysis.

The thorny topic of data quality is covered next. Anderson correctly observes that data has to be timely, relevant and trustworthy. He gives some good advice about different aspects of data quality, and discusses ways to improve data entry, and how to mitigate errors once they've crept into the data. 

Working out what data is the 'right' data to collect is the topic of the next chapter, with discussions about potential useful data sources, how to collect the data, and how to work out its value. A chapter on the different types of analyst comes next, essentially giving half a page each to various roles. I can see why it was included, but it seemed too slight to be useful.

The next three chapters are perhaps the most useful of the book, covering data analysis, metric design, and how to use the data. The chapter on data analysis discusses what an analyst should be trying to achieve when they are analyzing the data, along with the tools you can use to gain insights. Anderson gives overviews of various types of analysis - descriptive, exploratory, inferential, predictive and causal - and for the first time in the book there's some (very gentle) technical content. 

The chapter on metric design looks at how to work out what key performance indicators are worth tracking, and the best way to go about choosing or designing a metric. The third chapter in this set has the title 'storytelling with data', but is more useful than the title sounds. In reality, this is where many techies need to concentrate, because having really useful information is worthless unless you can make the right people listen and act on your info. Anderson says this is about the why and the what of communicating data, as opposed to the how, and his advice throughout the chapter is good.

 A/B testing is the topic for the next chapter, and Anderson uses some genuinely interesting examples to illustrate how it works: take two versions of your website/email offer or whatever, send one version to half a test group, the second version to the other half, see which gets a better response. Anderson discusses best practices and problems of A/B testing along with other approaches.


The chapter on Decision Making starts with a discussion of HiPPO decision making, where the Highest Paid Person's Opinion is used to make the decision regardless of what the data says. Having been honest about how many companies make decisions, Anderson goes on with an interesting discussion of what makes decision making hard, and what solutions exist to the problems raised.

Anderson's main message throughout the book is that organizations need a data-driven culture, and that's the topic of the next chapter. He backs up his argument with some interesting facts, but I suspect actually changing the way people interact with data takes more than some examples of why it's important. The next chapter is equally aspirational, looking at the need to have chief data officers and chief analytics officers in addition to CIOs or CTOs. The book closes with a chapter on privacy, ethics and risk.

This isn't a book for developers, but would be a useful read for IT managers..


To keep up with our coverage of books for programmers, follow @bookwatchiprog on Twitter or subscribe to I Programmer's Books RSS feed for each day's new addition to Book Watch and for new reviews.


Modern Fortran

Author: Milan Curcic
Publisher: Manning
Date: November 2020
Pages: 416
ISBN: 978-1617295287
Print: 1617295280
Audience: Fortran programmers
Rating: 5
Reviewer: Mike James
Not your parents' Fortran?

Python Programming and Visualization for Scientists 2nd Ed

Author: Alex DeCaria and Grant Petty
Publisher: Sundog Publishing
Pages: 372
ISBN: 978-0972903356
Print: 0972903356
Audience: Scientists wanting to use Python
Rating: 2
Reviewer: Mike James
Visualization - a difficult topic and difficult to see how to explain the ideas in a book.

More Reviews

Related Reviews

Beyond Big Data

Big Data Analytics

Data Science And Big Data Analytics

Big Data

Last Updated ( Friday, 26 October 2018 )