Page 71 - The Real Work Of Data Science Turning Data Into Information, Better Decisions, And Stronger Organizations by Ron S. Kenett, Thomas C. Redman (z-lib.org)_Neat

60 The Real Work of Data Science

the data used in the analysis and the quality of the generated information. You need to make
sure they’ve not done so in unfair ways. The clarity and completeness of the answer corre-
lates with the weight you should give to their conclusions.
Third, data science is essentially about developing a deeper understanding of how the
world works. False assumptions are crippling. For example, the assumption that home
prices were uncorrelated across markets was a major contributor to the financial crisis
that began in 2007 (Silver 2012). You should insist that data scientists state their assump-
tions in your language – don’t look away sheepishly if one states, “We’ve assumed
‘homoskedasticity.’”
6. Will your conclusions stand up to the scrutiny of our markets, moderately changing condi-
tions, and a “worst‐case scenario”? Don’t confuse data science with classical physics.
Verifying conclusions is not as simple as repeatedly dropping uneven weights from a tower.
You want your data scientists to be skeptical; to challenge each other; to test, test, and test
again; and to quantify, or at least fully describe, the uncertainty in their conclusions under
normal situations and to make clear when uncertainty explodes! This is critical because
operationalization may well be beyond the purview of the data scientist.
7. Who will be impacted and how? For example, privacy is a very touchy subject – both inside
and outside of your organization. The line between helpful and creepy is gray and thin
(Duhigg 2012), different societies think about privacy very differently, individuals vary and
change their mind, and legal frameworks (e.g. GDPR; see Appendix D) are only now under
development. Data scientists can produce startling insights, but they are not fully equipped
to think through the implications. It’s important to ask this question not only to the data
scientists you work with but your colleagues, privacy specialists, and those charged with
protecting the company’s brand as well.
8. What can I do to help? Quite obviously, there is no need to ask this question if the answers
to the first seven questions don’t satisfy. Bear in mind here that any important discovery
will have implications across the organization. We’re particularly concerned about change
management. All change is difficult, and resistance to counterintuitive results will prove far
too strong for most data scientists.

The starter kit, as the name implies, is
broad and not particularly deep, although it
HBR Guide to Data Analytics Basics
for Managers will almost certainly lead to deep discus-
sions. It facilitates discussions on the range
We are understandably reluctant to tell of topics that both data scientists and
anyone to “go read something,” although decision‐makers consider. At the same time,
we make one exception for this book. A decision‐makers should bring in other
compilation of digital articles by luminaries experts on topics such as privacy as the
such as Tom Davenport, D. J. Patil, and situation demands.
Michael Schrage; journalists such as Amy
Gallo; and a host of others, it is simply out- Implications
standing. The chapters are short, focused,
and extremely well written. Managers We’ve previously opined (Chapter 10) that
should scan it and put it on their shelves for helping individual decision‐makers and the
reference as the need arises. And data sci- entire organization become increasingly
entists should study it to understand where data‐driven is in the data scientists’ and
decision‐makers are coming from. CAOs’ best long‐term interests. The
waist measurement with rope aims to help