Creating a Data-Driven Organization: Practical Advice from the Trenches

by Carl Anderson

And there are great impacts and gains to be had. One report,1 controlling for other factors, found that data-driven organizations have a 5%–6% greater output and productivity that their less data-driven counterparts. They also had higher asset utilization, return on equity, and market value. Another report2 claims that analytics pays back $13.01 for every dollar spent. Being data-driven pays!

I often hear that data scientists spend 80% of their time obtaining, cleaning,

and preparing data, and only 20% of their time building models, analyzing, visualizing, and drawing conclusions from that data (for example, http://bit.ly/nyt-janitor and http://bit.ly/im-data-sci).

Prerequisite #2: Data must be accessible and queryable.

Shareable There must be a data-sharing culture within the organization so that data can be joined, such as combining customers’ clickstream with their transactional history.

Queryable There must be appropriate tools to query and slice and dice the data.

So, for an organization to be data-driven, there have to be humans in the loop, humans who ask the right questions of the data, humans who have the skills to extract the right data and metrics, and humans who use that data to inform next steps. In short, data alone is not going to save your organization.

Alerting

Four times more likely to capture information very well Nine times more likely to aggregate information very well Eight times more likely to analyze information very well Ten times more likely to disseminate information and insights very well 63% more likely to use a centralized analytics unit as the primary source of analytics (analytics organizational structures are covered in Chapter 4

“When working at scale, always remember issues that are one in a million can happen every second!”

shared responsibility.

Reminds we have a product I was thinking of called the equity did it score. And now I think this might be a good topic but needs to be tuned a little to address the idea that there is no true score

46% of medication errors (which had a base rate of 11% of all orders) were attributed to transcription mistakes.

:

That gives you the Unix command

Cgywin

the “5 whys method” from Toyota and Six Sigma, is a good example.

Edward Tufte (Envisioning Information, Visual Explanations, and The Visual Display of Quantitative Information

Stephen Few’s Now You See It (Analytics Press),

Show Me The Numbers

For web-based visualizations, try Scott Murray’s Interactive Data Visualization (O’Reilly). This

Dona Wong’s The

Wall Street Journal Guide to

Information Graphics (W. W. Norto...

Visualization Checklist

times.) Compare the charts in /r/dataisbeautiful/ versus r/dataisugly. What makes them beautiful and ugly, respectively?

Quote

Drive others

the viewer think as little as possible. Importantly, this doesn’t mean dumb down the content. First, start with a very clear understanding of the question that you are trying to answer, and a clear expectation of the audience, including their expectations and needs. Second, choose your visual carefully so that it is both appropriate to the data and to maximize the signal that shines through.

Greg was lucky. Not so much because it worked (that is, of course, significant), but because even back then Amazon had enough testing infrastructure and data-driven culture for him to be able to roll out a test. He was able to prove its value, roll it out, and impact the bottom line. In many situations, especially those that are novel, our intuition or gut instinct is poor.

This is why well-designed experimentation is so valuable. It switches the conversation from “I think...” to “The data shows...” Thus, this is an invaluable component of a data-driven organization.

HiPPOs (highest paid person’s opinion; covered in more detail in the next chapter) ruling the roost, there is a more democratic shift from higher-level decisions to lower-level hypothesis generation. He suggests that “you enable your most junior people to test their best ideas.”

“When people feel comfortable saying, ‘I don’t know, but let’s run an experiment,’

you eliminate the need to generate theories and instead focus on ideas, ideas that might just work and make a difference.

it: A/B testing actually stands for always be testing.

What’s the most data-driven animal? An adder. (Insert groan here.) What’s the least data-driven animal? A HiPPO. This is much more serious. HiPPO, “highest paid person’s opinion”

Fogg behavioral model.

In yet another survey of more than 700 senior executives, 61% agree that when making decisions, human insights must precede hard analytics, and 62% say it is often necessary or even preferable to rely on gut feelings and soft factors.

Rolf Dobelli’s The Art of Thinking Clearly or view this list

Readlate

We are not as good as we think we are

The Hidden Brain

Although Einstein joked, “If the facts don’t fit the theory, change the facts,”

Intuition can have immense value if used as a “gut check.” If data doesn’t match what one might expect, then it can be a clear signal that one should double check the data.

Surprisingly, however, it turns out that in tasks that require complex or creative thinking, not only is money not a poor motivator, but it actually leads to decreased performance.

Accenture found that 58% of executives see “outcome from data” as a key analytics challenge. “Establishing the linkage between data collection and analysis on the one hand, and the actions and outcomes predicated by analytics is proving to be a more difficult task for many than data collection or data integration.” Moreover, they found that only 39% of execs found that the data they generate is “relevant to the business strategy.”