Page 84 - 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

Putting Data Science, and Data Scientists, in the Right Spots 75

In contrast, Deming proposed putting
Data Science Centers of Excellence: Can statisticians in the line: “There will be in
Data Science Be Outsourced?
each division a statistician whose job it is
Many companies group their data scientists to find problems in that division, and
together in “centers of excellence.” Reasons work on them. He has the right and obliga-
for doing so include providing a better envi- tion to ask questions about any activity of
ronment to help data scientists grow into the division, and he is entitled to responsible
their roles, cultivate their craft, and learn answers” (Deming 1982).
from one another; to resolve funding issues; While we concur with Deming’s thinking,
and to create a critical mass of talent. Data it needs to be adapted to current needs and
scientists assume roles much like internal technologies. We actually see a continuum
consultants (although, of course, they are of data science opportunities. On one
not just advising but doing the work). This end are opportunities for basic process
is a good option for some companies and improvements. For such problems, putting
problems. data scientists “in the line” is clearly appro-
Extending this logic, the center of excel- priate and in accord with Deming (see also
lence could just as easily be a separate Hahn 2007).
company that decision‐makers hire on a On the other end of the spectrum are
fee‐for‐service basis. And given the more speculative opportunities (e.g.
(relative) newness of data science, it is only rethinking credit decisions based on social
natural for senior leaders to ask if they media data) that require fundamental inno-
should do so and outsource their data sci- vation. These must be performed in a “data
ence efforts. After all, learning how to man- lab” (Redman and Sweeney 2013a).
age data science is a tall order. Further, a And of course, as Figure 15.1 suggests,
cottage industry of companies offering there are opportunities that occupy a mid-
data science services is emerging. Such dle ground in this continuum. Each requires
companies already know how to manage its own structure. For example, fine‐tuning
the effort and already have a cadre of sea- a sophisticated algorithm might be best
soned data scientists and plenty of experi- done in a “center of excellence,” which
ence on hand. We see some merit in the some companies adopt (see “Data Science
approach, particularly for companies that Centers of Excellence: Can Data Science
are just getting started, and the problem at Be Outsourced?”).
hand demands expertise they simply do not There is no “one‐size‐fits‐all” solution.
have. But we do not think it is a good long‐ Rather, CAOs must balance the needs for
term solution. Whether in‐house or not, data science to be close to the decision‐
companies still have to manage the effort. makers they support, the company’s ability
More importantly, data science is increas- to manage data scientists (few line managers
ingly becoming a source of competitive really know how to treat data scientists), and
advantage. Sooner or later, companies must practical political realities.
learn how to grow and retain the talent. It is Further, CAOs are unlikely to have full
never a good idea to outsource your com- control over all the data scientists in even a
petitive advantage. relatively small company – as managers
are free to hire their own data science
teams (plenty of data scientists report into marketing, for example). Thus, the watchword for
CAOs is ensuring that decision‐makers are supported, through a mix of embedded, outsourced,
and laboratory‐based data scientists, while building a community among them.