An executives guide to machine learning | McKinsey & Company

1 di 6

http://www.mckinsey.com/Insights/High_Tech_Telecoms_Internet/An_e...

Article | McKinsey Quarterly

Its no longer the preserve of artificial-intelligence researchers and born-digital compa

Google, and Netflix.
June 2015 | by Dorian Pyle and Cristina San Jose

Machine learning is based on algorithms that can learn from data without relying on rule
It came into its own as a scientific discipline in the late 1990s as steady advances in digitizat
computing power enabled data scientists to stop building finished models and instead train
The unmanageable volume and complexity of the big data that the world is now swimming
potential of machine learningand the need for it.
In 2007 Fei-Fei Li, the head of Stanfords Artificial Intelligence Lab, gave up trying to progr
recognize objects and began labeling the millions of raw images that a child might encounte
feeding them to computers. By being shown thousands and thousands of labeled data sets w
cat, the machine could shape its own rules for deciding whether a particular set of digital pi
Last November, Lis team unveiled a program that identifies the visual elements of any pictu
of accuracy. IBMs Watson machine relied on a similar self-generated scoring system among
answers to crush the worlds best Jeopardy! players in 2011.
Dazzling as such feats are, machine learning is nothing like learning in the human sense (ye
does extraordinarily welland will get better atis relentlessly chewing through any amou
combination of variables. Because machine learnings emergence as a mainstream managem
recent, it often raises questions. In this article, weve posed some that we often hear and ans
we hope will be useful for any executive. Now is the time to grapple with these issues, becau
significance of business models turbocharged by machine learning is poised to surge. Indee
Ram Charan suggests that any organization that is not a math house now or is unable to be
already a legacy company.

1. How are traditional industries using machine learning to gather fresh busine
Well, lets start with sports. This past spring, contenders for the US National Basketball Asso
relied on the analytics of Second Spectrum, a California machine-learning start-up. By digit

22/07/2015 14:05

An executives guide to machine learning | McKinsey & Company

2 di 6

http://www.mckinsey.com/Insights/High_Tech_Telecoms_Internet/An_e...

seasons games, it has created predictive models that allow a coach to distinguish between, a
Maheswaran puts it, a bad shooter who takes good shots and a good shooter who takes bad
his decisions accordingly.
You cant get more venerable or traditional than General Electric, the only member of the o
Industrial Average still around after 119 years. GE already makes hundreds of millions of do
data it collects from deep-sea oil wells or jet engines to optimize performance, anticipate br
streamline maintenance. But Colin Parris, who joined GE Software from IBM late last year
software research, believes that continued advances in data-processing power, sensors, and
will soon give his company the same sharpness of insight into the individual vagaries of a je
has into the online behavior of a 24-year-old netizen from West Hollywood.

2. What about outside North America?

In Europe, more than a dozen banks have replaced older statistical-modeling approaches w
techniques and, in some cases, experienced 10 percent increases in sales of new products, 2
capital expenditures, 20 percent increases in cash collections, and 20 percent declines in ch
achieved these gains by devising new recommendation engines for clients in retailing and in
medium-sized companies. They have also built microtargeted models that more accurately
cancel service or default on their loans, and how best to intervene.
Closer to home, as a recent article in McKinsey Quarterly notes, our colleagues have been a
to the soft stuff of talent management. Last fall, they tested the ability of three algorithms d
vendors and one built internally to forecast, solely by examining scanned rsums, which of
potential recruits the firm would have accepted. The predictions strongly correlated with th
Interestingly, the machines accepted a slightly higher percentage of female candidates, whi
using analytics to unlock a more diverse range of profiles and counter hidden human bias.
As ever more of the analog world gets digitized, our ability to learn from data by developing
will only become more important for what are now seen as traditional businesses. Google ch
Varian calls this computer kaizen. For just as mass production changed the way products
continuous improvement changed how manufacturing was done, he says, so continuous [
experimentation will improve the way we optimize business processes in our organizations

3. What were the early foundations of machine learning?

Machine learning is based on a number of earlier building blocks, starting with classical stat
inference does form an important foundation for the current implementations of artificial i
important to recognize that classical statistical techniques were developed between the 18th
centuries for much smaller data sets than the ones we now have at our disposal. Machine le
unconstrained by the preset assumptions of statistics. As a result, it can yield insights that h

22/07/2015 14:05

An executives guide to machine learning | McKinsey & Company

3 di 6

http://www.mckinsey.com/Insights/High_Tech_Telecoms_Internet/An_e...

see on their own and make predictions with ever-higher degrees of accuracy.
More recently, in the 1930s and 1940s, the pioneers of computing (such as Alan Turing, wh
abiding interest in artificial intelligence) began formulating and tinkering with the basic tec
networks that make todays machine learning possible. But those techniques stayed in the la
many technologies did and, for the most part, had to await the development and infrastruct
computers, in the late 1970s and early 1980s. Thats probably the starting point for the mac
curve. New technologies introduced into modern economiesthe steam engine, electricity,
computers, for exampleseem to take about 80 years to transition from the laboratory to w
cultural invisibility. The computer hasnt faded from sight just yet, but its likely to by 2040
take much longer for machine learning to recede into the background.

4. What does it take to get started?

C-level executives will best exploit machine learning if they see it as a tool to craft and impl
vision. But that means putting strategy first. Without strategy as a starting point, machine le
a tool buried inside a companys routine operations: it will provide a useful service, but its l
probably be limited to an endless repetition of cookie cutter applications such as models f
stimulating, and retaining customers.
We find the parallels with M&A instructive. That, after all, is a means to a well-defined end
rushes into a flurry of acquisitions or mergers and then just sits back to see what happens. C
on machine learning should make the same three commitments companies make before em
commitments are, first, to investigate all feasible alternatives; second, to pursue the strategy
C-suite level; and, third, to use (or if necessary acquire) existing expertise and knowledge in
the application of that strategy.
The people charged with creating the strategic vision may well be (or have been) data scien
the problem and the desired outcome of the strategy, they will need guidance from C-level
other crucial strategic initiatives. More broadly, companies must have two types of people t
of machine learning. Quants are schooled in its language and methods. Translators can
of data, machine learning, and decision making by reframing the quants complex results as
that generalist managers can execute.
Access to troves of useful and reliable data is required for effective machine learning, such a
tests, to predict oncological outcomes better than physicians or Facebooks recent success te
identify specific human faces nearly as accurately as humans do. A true data strategy starts
the data, determining the time and money required to fill those gaps, and breaking down sil
departments hoard information and politicize access to itone reason some companies hav
of chief data officer to pull together whats required. Other elements include putting respon
data in the hands of frontline managers.

22/07/2015 14:05

An executives guide to machine learning | McKinsey & Company

4 di 6

http://www.mckinsey.com/Insights/High_Tech_Telecoms_Internet/An_e...

Start smalllook for low-hanging fruit and trumpet any early success. This will help recruit
and reinforce the changes in individual behavior and the employee buy-in that ultimately d
organization can apply machine learning effectively. Finally, evaluate the results in the light
criteria for success.

5. Whats the role of top management?

Behavioral change will be critical, and one of top managements key roles will be to influenc
Traditional managers, for example, will have to get comfortable with their own variations o
technique digital companies use to see what will and will not appeal to online consumers. F
armed with insights from increasingly powerful computers, must learn to make more decisi
top management setting the overall direction and zeroing in only when exceptions surface.
of analyticsproviding the front line with the necessary skills and setting appropriate incen
sharingwill require time.
C-level officers should think about applied machine learning in three stages: machine learn
3.0or, as we prefer to say, description, prediction, and prescription. They probably dont n
about the description stage, which most companies have already been through. That was all
in databases (which had to be invented for the purpose), a development that gave managers
past. OLAPonline analytical processingis now pretty routine and well established in mo
Theres a much more urgent need to embrace the prediction stage, which is happening righ
cutting-edge technology already allows businesses not only to look at their historical data bu
behavior or outcomes in the futurefor example, by helping credit-risk officers at banks to
customers are most likely to default or by enabling telcos to anticipate which customers are
churn in the near term (exhibit).
Exhibit

22/07/2015 14:05

An executives guide to machine learning | McKinsey & Company

5 di 6

http://www.mckinsey.com/Insights/High_Tech_Telecoms_Internet/An_e...

A frequent concern for the C-suite when it embarks on the prediction stage is the quality of
often paralyzes executives. In our experience, though, the last decades IT investments have
companies with sufficient information to obtain new insights even from incomplete, messy
course that those companies choose the right algorithm. Adding exotic new data sources ma
benefit compared with what can be mined from existing data warehouses. Confronting that
the chief data scientist.
Prescriptionthe third and most advanced stage of machine learningis the opportunity o
therefore command strong C-suite attention. It is, after all, not enough just to predict what
do; only by understanding why they are going to do it can companies encourage or deter th
future. Technically, todays machine-learning algorithms, aided by human translators, can a
example, an international bank concerned about the scale of defaults in its retail business re
group of customers who had suddenly switched from using credit cards during the day to u
middle of the night. That pattern was accompanied by a steep decrease in their savings rate
branch managers, the bank further discovered that the people behaving in this way were als
recent stressful event. As a result, all customers tagged by the algorithm as members of that
automatically given a new limit on their credit cards and offered financial advice.
The prescription stage of machine learning, ushering in a new era of manmachine collabo
biggest change in the way we work. While the machine identifies patterns, the human trans
will be to interpret them for different microsegments and to recommend a course of action.
be directly involved in the crafting and formulation of the objectives that such algorithms at

22/07/2015 14:05

An executives guide to machine learning | McKinsey & Company

6 di 6

http://www.mckinsey.com/Insights/High_Tech_Telecoms_Internet/An_e...

6. This sounds awfully like automation replacing humans in the long run. Are w
knowing whether machines will replace managers?
Its true that change is coming (and data are generated) so quickly that human-in-the-loop i
decision making is rapidly becoming impractical. Looking three to five years out, we expect
of artificial intelligence, as well as the development of distributed autonomous corporations
self-motivating, self-contained agents, formed as corporations, will be able to carry out set o
autonomously, without any direct human supervision. Some DACs will certainly become sel
One current of opinion sees distributed autonomous corporations as threatening and inimic
by the time they fully evolve, machine learning will have become culturally invisible in the s
inventions of the 20th century disappeared into the background. The role of humans will be
the algorithms as they attempt to achieve the objectives that they are given. That is one less
trading algorithms which wreaked such damage during the financial crisis of 2008.
No matter what fresh insights computers unearth, only human managers can decide the ess
as which critical business problems a company is really trying to solve. Just as human collea
reviews and assessments, so these brilliant machines and their works will also need to be
refinedand, who knows, perhaps even fired or told to pursue entirely different pathsby
experience, judgment, and domain expertise.
The winners will be neither machines alone, nor humans alone, but the two working togeth

7. So in the long term theres no need to worry?

Its hard to be sure, but distributed autonomous corporations and machine learning should
agenda. We anticipate a time when the philosophical discussion of what intelligence, artific
be will end because there will be no such thing as intelligencejust processes. If distributed
corporations act intelligently, perform intelligently, and respond intelligently, we will cease
high-level intelligence other than the human variety exists. In the meantime, we must all th
want these entities to do, the way we want them to behave, and how we are going to work w
About the authors
Dorian Pyle is a data expert in McKinseys Miami office, and Cristina San Jose is a principal in the Madrid office.

22/07/2015 14:05